Code import - branch release/SIENTIAPDE-1646

.env.example

@@ -0,0 +1,51 @@
+POSTGRES_HOST="paradedb-rw.paradedb.svc.cluster.local"
+POSTGRES_PORT="5432"
+POSTGRES_USER="sientia"
+POSTGRES_PASSWORD="password"
+POSTGRES_DBNAME="sientia"
+POSTGRES_MIN_CONNECTIONS="10"
+POSTGRES_MAX_CONNECTIONS="30"
+
+MLFLOW_HOST="http://sientia-tracker-mlflow-tracking.sientia-tracker.svc.cluster.local"
+MLFLOW_PORT="80"
+MLFLOW_USERNAME="aignosi"
+MLFLOW_PASSWORD="mlflow_password"
+
+# Worker runtime name for PluginStore.install_runtime (PredictionsBatch / MinimalRetrain workers).
+RUNTIME="single"
+
+# Plugin store (Git-backed catalog + runtime install).
+STORE_BASE_URL="http://gitea.sientia.svc.cluster.local:3000"
+STORE_OWNER="sientia"
+STORE_REPO="model-library-store"
+STORE_BRANCH="main"
+STORE_USERNAME=""
+STORE_PASSWORD=""
+STORE_CACHE_TTL_SECONDS=""
+
+PYPI_SERVER="http://library-distribution-server.library.svc.cluster.local:5000"
+PYPI_USERNAME=""
+PYPI_PASSWORD=""
+
+OPC_ID="1"
+OPC_URL="opc.tcp://sientia-opc-simulator-opc.sientia.svc.cluster.local:4840"
+
+LOG_LEVEL="DEBUG"
+HTTP_METRICS_PORT="9090"
+HTTP_SDK_METRICS_PORT="9091"
+PROJECT_NAME="sientia-laborious"
+
+TEMPORAL_HOST="temporal-frontend.temporal.svc.cluster.local:7233"
+TEMPORAL_NAMESPACE="laborious"
+
+MONGODB_USERNAME="mongo_user"
+MONGODB_PASSWORD="mongo_db_password"
+MONGODB_URL="my-release-mongodb.mongodb.svc.cluster.local:27017"
+MONGODB_DATABASE="sientia"
+MONGODB_TTL_INDEX_HOURS="1"
+
+MINIO_ENDPOINT_URL="http://localhost:9000"
+MINIO_ACCESS_KEY="sientia"
+MINIO_SECRET_KEY="sientia"
+MINIO_REGION_NAME="sa-east-1"
+MINIO_DEFAULT_BUCKET="sientia"

.github/workflows/quality-gate.yml

@@ -0,0 +1,17 @@
+name: Quality gate
+
+on:
+  pull_request:
+    branches:
+      - main
+    types: [ opened, synchronize, reopened ]
+
+jobs:
+  quality-gate:
+    uses: Aignosi/github_workflow_templates/.github/workflows/dataops-module-quality-gate.yml@main
+    permissions: write-all
+    with:
+      project_name: 'laborious'
+      repositories: 'sientia-dataops-library, sientia-mlops-library'
+      requirements_file: 'requirements-light.txt'
+    secrets: inherit

.github/workflows/release.yml

@@ -0,0 +1,25 @@
+name: Create Release on Merge to Main
+
+on:
+  pull_request:
+    types: [closed]
+    branches:
+      - main
+  workflow_dispatch:
+    inputs:
+      version:
+        description: 'Version to release'
+        required: false
+        type: string
+
+jobs:
+  release:
+    if: |
+      (github.event_name == 'pull_request' && github.event.pull_request.merged == true) ||
+      github.event_name == 'workflow_dispatch'
+    uses: Aignosi/github_workflow_templates/.github/workflows/dataops-module-release.yml@main
+    permissions: write-all
+    with:
+      project_name: 'laborious'
+      release_version: ${{ github.event.inputs.version || '' }}
+    secrets: inherit

.gitignore

@@ -0,0 +1,58 @@
+# Ignorar volumes do Docker
+docker-compose.override.yml
+**/db_data/
+**/kafka-volume/
+**/zookeeper-volume/
+**/mage_data/
+**/minio_data/
+**/venv/
+**/certs/*.pem
+**/certs/*.der
+**/certs/*.csr
+**/deploy/*.yaml
+scouter/.file_versions/
+scouter/pipelines/**/triggers.yaml
+**/postgres_data/**
+# Ignorar arquivos e diretórios de cache do Python
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+
+# Ignorar logs
+*.log
+
+# Ignorar arquivos de configuração locais
+.vscode/
+.pytest_cache/
+.idea/
+*.swp
+
+# Ignorar arquivos temporários
+*.tmp
+*.bak
+*.old
+.secret
+
+# Ignorar coverage
+htmlcov/
+.coverage
+coverage.xml
+
+# git keys
+git_key*
+
+git_log
+
+.env
+
+tmp/
+catboost_info/
+
+.ruff_cache/
+.mypy_cache/
+mlruns/
+
+relatorio*
+openspec/*
+.cursor/*

Makefile

@@ -0,0 +1,7 @@
+VERSION = 1.0.8
+name = sientia-laborious
+# ENVIRONMENT = production
+
+docker-hub:
+	@docker build --no-cache -t aignosi.azurecr.io/$(name):$(VERSION) .
+	@docker push aignosi.azurecr.io/$(name):$(VERSION)

docs/E2E_TEST_REPORT.md

@@ -0,0 +1,149 @@
+# E2E test run report
+
+**Date:** 2026-05-08  
+**Command:** `source venv/bin/activate && rtk pytest e2e/ -v --tb=short`  
+**Environment:** Linux, Python 3.11.15, pytest 9.0.3  
+
+## Summary
+
+| Metric | Count |
+|--------|------:|
+| Collected | 43 |
+| **Passed** | **37** |
+| **Failed** | **6** |
+
+Full pytest output (compressed by `rtk`) was written to:
+
+`~/.local/share/rtk/tee/1778268193_pytest.log`
+
+---
+
+## Failed tests (6)
+
+1. `e2e/test_drift.py::test_drift_happy_path_persists_all_columns_with_reference_data`
+2. `e2e/test_drift.py::test_drift_uses_30pct_fallback_when_reference_unavailable`
+3. `e2e/test_drift.py::test_drift_chunk_period_seconds_preserves_seconds_in_chunk_start_date`
+4. `e2e/test_predictions_batch_prediction_process.py::test_scenario_2_1_3_input_gate_triggers_repeat`
+5. `e2e/test_predictions_batch_prediction_process.py::test_scenario_2_2_3_transform_gate_triggers_repeat`
+6. `e2e/test_predictions_batch_prediction_process.py::test_scenario_2_3_3_predict_gate_triggers_repeat`
+
+**Follow-up:** Jensen–Shannon NULLs for the drift happy-path scenario are fully traced (histogram out-of-range + `density=True`, vs NannyML’s leftover bin) in [DRIFT_JS_NULL_INVESTIGATION.md](./DRIFT_JS_NULL_INVESTIGATION.md).
+
+---
+
+## Failure group A — Drift: `drift_metrics.value` NOT NULL (2 tests)
+
+### Error
+
+`psycopg2.errors.NotNullViolation`: null value in column `value` of relation `sientia_data.drift_metrics` violates not-null constraint.
+
+Example failing row (from logs): `feature=sensor_2`, `method=jensen_shannon`, `value=null`, with `kolmogorov_smirnov` / `wasserstein` populated for the same chunk.
+
+The bulk INSERT built by `Activities.export_data_to_postgres` includes parameters such as `'value__4': None` for `jensen_shannon` on a given chunk.
+
+### Root cause
+
+`calculate_drift` (real `DriftAnalysis` + `ModelMetrics.get_drift_metrics`) can emit **NaN / missing** values for some metric methods (here **Jensen–Shannon**) on some features/chunks. Pandas/SQLAlchemy turns that into SQL `NULL`, while the E2E schema (mirroring production) defines:
+
+```sql
+value numeric NOT NULL
+```
+
+in `e2e/db_schema.sql` for `sientia_data.drift_metrics`.
+
+### Recommended fixes (pick one consistent with product rules)
+
+1. **Application layer (preferred if NULLs are never valid in production):** Before export, sanitize the drift dataframe — e.g. drop rows where `value` is null/NaN, or replace with a defined sentinel (only if product agrees), or skip emitting that method row when the statistic is undefined.
+2. **Analytics layer:** Harden the Jensen–Shannon (and similar) paths so they always return a finite float for the supported inputs, or explicitly map “undefined” to an agreed numeric convention.
+3. **Schema (only if product allows missing metrics):** Align DDL with reality by making `value` nullable — **only** if production and downstream consumers already expect missing metrics; the E2E comment in `test_drift.py` suggests `feature`/`timestamp` nullable cases exist, but `value` is still listed in `NON_NULL_DRIFT_COLUMNS`.
+
+### Tests affected
+
+- `test_drift_happy_path_persists_all_columns_with_reference_data`
+- `test_drift_uses_30pct_fallback_when_reference_unavailable`
+
+---
+
+## Failure group B — Drift: chunk period seconds (1 test)
+
+### Error
+
+`AssertionError: expected at least one drift row to be persisted` (`e2e/test_drift.py:493`).
+
+The workflow run completed without failing the test via `WorkflowFailureError`, but **no rows** were found in `sientia_data.drift_metrics` for the model.
+
+### Likely cause
+
+In `Drift.run`, persistence runs only when `if drift_data:` is truthy (`laborious/workflows/drift.py`). An **empty** drift result skips `export_data_to_postgres`, so the table stays empty.
+
+Probable reasons:
+
+- With **`chunk_period='s'`** and only **three** target rows (30 s spacing), `calculate_drift` / `DriftAnalysis` may produce **no output rows** (insufficient data per chunk or internal filters).
+- Less likely here: time-window mismatch — timestamps are built from `datetime.now(UTC)` with `interval: 60` minutes from `drift_base.json`, so data should still fall in the window.
+
+### Recommended fixes
+
+1. **Test data:** Increase the number of second-spaced points (and/or span multiple chunk boundaries) so the analyzer reliably emits at least one chunk row.
+2. **Product code:** If sub-minute chunking is required to always produce metrics when any data exists, adjust `ModelMetrics` / `DriftAnalysis` integration for small-N second buckets.
+3. **Diagnostics:** Run the same scenario with `pytest -s` and confirm logs for “empty `drift_data`” vs export errors.
+
+Solução a ser aplicada:
+Dividir em dois testes:
+
+1. Teste com dados suficientes para produzir pelo menos uma linha de drift
+2. Teste com dados insuficientes para produzir pelo menos uma linha de drift, mas ja esperando os erros e validando que nao foi persistido nada
+
+### Test affected
+
+- `test_drift_chunk_period_seconds_preserves_seconds_in_chunk_start_date`
+
+---
+
+## Failure group C — Prediction REPEAT path: unique constraint on predictions (3 tests)
+
+### Error
+
+`psycopg2.errors.UniqueViolation`: duplicate key value violates unique constraint **`unique_model_id_timestamp`** on `sientia_data.predictions` (`model_id`, `timestamp`).
+
+### Root cause
+
+REPEAT is handled by `Activities.repeat_last_prediction` (registered from `sientia_do.temporal.activities.postgres_sync.Postgres`, via `Storage` inheritance in `laborious/activities/storage.py`). The E2E tests seed a prior row with a **fixed** timestamp:
+
+```python
+# e2e/test_predictions_batch_prediction_process.py — insert_sample_prediction
+VALUES ({model_id}, '2024-01-01 12:00:00+00:00', ...)
+```
+
+The helper `assert_repeat` expects **two** rows with the same `(model_id, prediction, prediction_confidence, prediction_status)` but compares **only** those columns — not `timestamp` (`e2e/helpers.py`). So the intended behavior is: **duplicate business payload**, not necessarily **duplicate primary unique key** `(model_id, timestamp)`.
+
+If `repeat_last_prediction` **INSERT**s a copy using the **same** `timestamp` as the last prediction, Postgres correctly rejects the second insert.
+
+### Recommended fixes
+
+1. **Implement REPEAT as UPSERT:** Use `ON CONFLICT (model_id, timestamp) DO UPDATE` (or the project’s existing `export_data_to_postgres` `on_conflict` pattern used in `FormatAndExportPrediction`) when writing the repeated prediction — if the product definition of REPEAT is “refresh same logical slot.”
+2. **Insert with the current batch timestamp:** Copy numeric/status fields from the last row but set `timestamp` to the **new** batch instant (e.g. the workflow’s `last_timestamp` / slice timestamp). This matches `assert_repeat`, which does not assert on `timestamp`.
+3. **Test-only change (weakest):** Relax assertions or change seed data — only if production behavior is “duplicate key is expected” (unlikely).
+
+Solução a ser aplicada:
+Alterar o teste para usar o timestamp da ultima execucao do batch, ao inves do timestamp do primeiro batch.
+2. **Insert with the current batch timestamp:** Copy numeric/status fields from the last row but set `timestamp` to the **new** batch instant (e.g. the workflow’s `last_timestamp` / slice timestamp). This matches `assert_repeat`, which does not assert on `timestamp`.
+
+### Tests affected
+
+- `test_scenario_2_1_3_input_gate_triggers_repeat`
+- `test_scenario_2_2_3_transform_gate_triggers_repeat`
+- `test_scenario_2_3_3_predict_gate_triggers_repeat`
+
+---
+
+## Passing areas (sanity check)
+
+All scenarios in `test_child_workflows_e2e.py`, `test_minimal_retrain.py`, `test_minio_offload.py`, `test_predictions_batch_format_export.py`, `test_predictions_batch_main_workflow.py` (except the three REPEAT cases above), and `test_simple_metrics.py` **passed** in this run.
+
+---
+
+## Suggested order of work
+
+1. Fix **drift `value` NULL** — unblocks two high-value drift E2Es and may clarify the chunk-seconds scenario if exports start succeeding consistently.
+2. Fix **`repeat_last_prediction` uniqueness** — unblocks three prediction-process E2Es; implementation likely lives in **`sientia_do`** Postgres activities, not in this repo.
+3. Revisit **`test_drift_chunk_period_seconds`** data volume / expectations after drift export is stable.

docs/opc-communication.md

@@ -0,0 +1,164 @@
+# OPC UA communication (Laborious)
+
+Laborious exports predictions to OPC UA servers through `OpcRepository` ([`laborious/utils/repository/opc_repository.py`](../laborious/utils/repository/opc_repository.py)) and the synchronous Temporal activity layer in [`laborious/activities/opc.py`](../laborious/activities/opc.py). The repository uses `asyncua.sync.Client` (asyncio on a background thread) so activities remain blocking without `async def`.
+
+OPC reconnect, write error classification (`opc_error_kind`), and activity confidence/comment behavior are converted from the **async** implementation on `main` at `fcc8920a8be4` (`asyncua.Client` + `asyncio` reconnect task → `threading` reconnect thread). Re-convert with `scripts/convert_opc_async_to_sync.py` when `main` OPC files change.
+
+Implementation plan for session/channel recovery on Tier-1 `Bad*` errors: [`.cursor/plans/opc_bad_reconnect_ac4c6045.plan.md`](../.cursor/plans/opc_bad_reconnect_ac4c6045.plan.md).
+
+## Architecture
+
+```text
+Worker (long-lived)
+  └── OpcRepository per OPC server id (from OPC_CONFIG / env)
+        ├── connect / disconnect / validate_connection (read-only)
+        ├── _connect_locked / _reconnect_locked (under _connection_lock)
+        ├── write_data (single attempt per call)
+        └── background reconnect on Tier-1 Bad*, closed protocol, or stale session
+
+Temporal activity write_opc_data
+  └── OPC.manage_output_tags → write_data per tag (sequential per activity)
+```
+
+One worker process holds one `OpcRepository` instance per configured server. Multiple Temporal activities can call `write_data` concurrently on the same repository.
+
+## Connection lifecycle
+
+| Phase | Behavior |
+|-------|----------|
+| Startup | `init_opc()` creates repositories and calls `connect()` → `_connect_locked()` |
+| Steady state | `validate_connection()` is read-only (`protocol.state` only); `_session_ready` is checked in `write_data` |
+| Tier-1 Bad* / protocol closed / session not ready | `_start_reconnect(reason)` → `_run_reconnect` (thread) → `_reconnect_locked()` (respects `reconnection_interval`) |
+| Write | `write_data()` checks in-flight reconnect thread, `_session_ready`, validates, then one `get_node` + `write_value` |
+| Shutdown | `disconnect()` sets `_allow_reconnect = False`, then tears down session |
+
+### Session and channel timeouts
+
+Requested session and secure-channel lifetime: **10 minutes** (`OPC_UA_SESSION_AND_CHANNEL_TIMEOUT_MS` in `opc_repository.py`). The server may revise these values; negotiated values are logged after connect and exposed as `opc_session_revised_timeout_milliseconds`.
+
+### Reconnection interval
+
+`OPC_RECONNECTION_INTERVAL` is in **seconds** (default `120`). It gates **background** reconnect after Tier-1 `Bad*` (`last_reconnection_time` is updated only in `_reconnect_locked()`). It limits load on the OPC server when many workflows fail at once.
+
+## Concurrency: connection lock and session readiness
+
+To allow **multiple concurrent writes** when the session is healthy, but **block all writes** while the connection is being torn down or re-established:
+
+| Primitive | Role |
+|-----------|------|
+| `_connection_lock` (`threading.Lock`) | Held for the entire `disconnect` → `connect` path. Only one connection-maintenance task at a time. |
+| `_session_ready` (`threading.Event`) | Set when a session is ready for writes; cleared before reconnect starts and set again after a successful connect. |
+| `_allow_reconnect` | Cleared in `disconnect()` so shutdown does not spawn reconnect threads |
+
+**Connection methods (caller holds `_connection_lock` for `_*_locked` helpers):**
+
+| Method | Role |
+|--------|------|
+| `_create_client()` | Create asyncua `Client` + optional `set_security`; raises if `client` already exists |
+| `_open_session()` | `client.connect()` + metrics; raises if session already open or client missing |
+| `_connect_locked()` | `_create_client()` (when needed) + `_open_session()`; raises if already connected |
+| `_disconnect_locked()` | Teardown session and clear `client` |
+| `_reconnect_locked()` | `_disconnect_locked()` + `_connect_locked()`; sets `last_reconnection_time` |
+
+Public `connect()` / `disconnect()` acquire the lock and call `_connect_locked()` / `_disconnect_locked()`.
+
+**Write path (`write_data`):**
+
+1. If a reconnect **thread** is alive → `reconnect_in_progress`.
+2. If `_session_ready` is cleared → schedule `SessionNotReady` reconnect; return `reconnect_in_progress` or `connection_lost`.
+3. If `validate_connection()` fails (protocol closed) → schedule `ProtocolClosed` reconnect; return `connection_lost`.
+4. Single `get_node` + `write_value` (no retry in the same call).
+
+**Reconnect path (`_run_reconnect`):**
+
+1. `_start_reconnect` clears `_session_ready` and starts a daemon thread when the interval allows and `_allow_reconnect` is true.
+2. `with _connection_lock:` → `_reconnect_locked()`.
+3. `_session_ready` is set on successful `_open_session()`.
+
+A second `_connect_locked()` while a session is already open raises `OpcSessionAlreadyConnectedError` (disconnect first).
+
+**asyncua note:** Concurrent `write_value` on the same session is only safe if the stack tolerates it. If production shows issues, serialize writes while keeping the connection lock semantics above.
+
+## Tier-1 `Bad*` errors and reconnect
+
+When the server invalidates the session (e.g. `BadSessionIdInvalid`) but the client still sees transport as open, `write_data` fails once, records the OPC status in metrics, and **schedules** reconnect if:
+
+- The exception is a `UaStatusCodeError` whose name is in `RECONNECTABLE_OPC_BAD_NAMES` (see plan), and
+- `reconnection_interval` has elapsed since `last_reconnection_time`, and
+- No reconnect task is already running.
+
+There is **no write retry**: the failed export is not sent again in the same activity.
+
+## Prediction confidence and PostgreSQL comments
+
+| `prediction_confidence` | Meaning |
+|-------------------------|---------|
+| (unchanged) | Successful OPC export |
+| **12** | Generic OPC write failure (`OPC_WRITTING_ERROR_CONFIDENCE`) |
+| **14** | Tier-1 session/channel `Bad*` on export (`OPC_SESSION_BAD_CONFIDENCE`) |
+| **14** | Write while reconnect in progress (`OPC_SESSION_BAD_CONFIDENCE`, comment `OPC UA reconnect in progress`) |
+| **13** | PI Web API write failure (separate path) |
+
+Session/channel errors use a stable comment for counting:
+
+```text
+OPC UA session/channel error: BadSessionIdInvalid
+```
+
+Reconnect-in-progress exports use:
+
+```text
+OPC UA reconnect in progress
+```
+
+Example SQL:
+
+```sql
+SELECT count(*) FROM predictions WHERE prediction_confidence = 14;
+SELECT count(*) FROM predictions WHERE comments LIKE 'OPC UA session/channel error:%';
+```
+
+## Prometheus metrics (`opc_*`)
+
+Defined in [`laborious/metrics.py`](../laborious/metrics.py). Do not rename in production without a dashboard migration.
+
+| Metric | Purpose |
+|--------|---------|
+| `opc_connections_initiated_total` | Connection attempts |
+| `opc_connections_failed_total` | Failed connects |
+| `opc_connection_status` | Gauge 1=connected, 0=disconnected |
+| `opc_session_created_total` | Session established after connect |
+| `opc_session_closed_total` | Disconnect initiated |
+| `opc_session_revised_timeout_milliseconds` | Negotiated session timeout (ms) |
+| `opc_write_attempts_total` | Per write; label `result` = `OK` or exception name |
+| `opc_write_inter_arrival_over_session_timeout_total` | Successful writes spaced longer than revised session timeout |
+
+Legacy activity metrics: `laborious_prediction_opc_writing_count`, `laborious_prediction_opc_writing_response_time_monitor`.
+
+## Environment variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `OPC_CONFIG` | — | JSON map of server configs (overrides single-server env) |
+| `OPC_ID` | `1` | Server id |
+| `OPC_URL` | `opc.tcp://localhost:4840` | Endpoint |
+| `OPC_SERVER_NAME` | `default_server` | Label for metrics/logs |
+| `OPC_SERVER_URI` | same as URL | Application URI / cert SAN |
+| `OPC_CERT_PATH` | — | Client certificate (secure mode) |
+| `OPC_PRIVATE_KEY_PATH` | — | Client private key |
+| `OPC_SERVER_CERT_PATH` | — | Server certificate |
+| `OPC_RECONNECTION_INTERVAL` | `120` | Minimum seconds between reconnects |
+
+## Operations checklist
+
+- Correlate `BadSessionIdInvalid` in `opc_write_attempts_total` with `opc_session_closed_total` / `opc_session_created_total` (reconnect may finish after the row is stored with confidence 14).
+- Use confidence **14** and comment prefix for session invalidation rates; use **12** for other OPC failures.
+- Respect `OPC_RECONNECTION_INTERVAL` under parallel load; bursts of confidence 14 are expected until the next successful cycle.
+
+## Related tests
+
+- Unit: [`tests/laborious/utils/repository/test_opc_repository.py`](../tests/laborious/utils/repository/test_opc_repository.py)
+- Unit: [`tests/laborious/activities/test_opc.py`](../tests/laborious/activities/test_opc.py)
+- E2E (mock OPC): [`e2e/test_predictions_batch_format_export.py`](../e2e/test_predictions_batch_format_export.py)
+- E2E (in-process asyncua server + real `OpcRepository`): [`e2e/test_opc_real_server.py`](../e2e/test_opc_real_server.py) — scenarios 3.1.2, 3.2.2, 3.2.4, 3.2.5
+- Scenarios: [`e2e/scenarios.md`](../e2e/scenarios.md)

docs/sientia_model_drift_jensen_shannon_change.md

@@ -0,0 +1,55 @@
+# Specification: `sientia_model` — Jensen–Shannon drift (`DriftAnalysis`)
+
+This document describes what **`sientia_model.analytics.drift_analysis.DriftAnalysis`** should change so downstream consumers (e.g. Laborious `calculate_drift` → Postgres `drift_metrics.value NOT NULL`) no longer receive **NaN** for Jensen–Shannon on valid finite data.
+
+## Scope
+
+- **File:** `sientia_model/analytics/drift_analysis.py`
+- **Method:** `_jensen_shannon_distance(self, ref: np.ndarray, cur: np.ndarray, bins: int = 20) -> float`
+- **Callers:** `detect_univariate_drift` uses this for the `jensen_shannon` method; results are written to `value` in the drift dataframe.
+
+## Problem
+
+The current implementation:
+
+1. Builds bin edges from **`ref`** only: `np.histogram(ref, bins=bins, density=True)`.
+2. Builds the chunk histogram with **`density=True`** on the same edges: `np.histogram(cur, bins=edges, density=True)`.
+
+When **every** value in **`cur`** falls **outside** the closed support implied by those edges (typical case: production chunk drifted above the reference max or below the reference min), NumPy yields **all-zero counts** for `cur`. With **`density=True`**, normalization does **0/0**, producing **NaN** for the whole histogram, which propagates to **`float('nan')`** in `detect_univariate_drift` → **SQL NULL** where `value` is `NOT NULL`.
+
+This appears in Laborious when:
+
+- `model_config.target` excludes the main target column from univariate features, so another feature (e.g. `sensor_2`) is compared chunk-by-chunk against the full reference series for that feature.
+- Reference and current ranges do not overlap for some chunks (strong drift or different scaling).
+
+Other methods (`kolmogorov_smirnov`, `wasserstein`) do not use this histogram+density path, so they can stay finite while **Jensen–Shannon** alone becomes null.
+
+## Required behavior
+
+1. **Finite output** for finite `ref` and `cur` after removing non-finite values, whenever both sides have **at least one** usable sample.
+2. **Explicit handling of out-of-range chunk mass:** probability mass from `cur` that does not fall into any bin defined from `ref` must still be represented (so the chunk distribution sums to 1), analogous to NannyML’s continuous JS approach (tail / “leftover” mass).
+3. **Missing values:** drop `NaN` from `ref` and `cur` before computing. If either side is **empty** after that, return **`float('nan')`** (callers may filter or map; schema may still forbid null — product decision outside this spec).
+
+## Recommended algorithm (replace current body)
+
+1. `ref = np.asarray(ref, float); cur = np.asarray(cur, float)`.
+2. `ref = ref[~np.isnan(ref)]; cur = cur[~np.isnan(cur)]`.
+3. If `ref.size == 0` or `cur.size == 0`: return `float('nan')`.
+4. `hist_ref, edges = np.histogram(ref, bins=bins)` — **counts**, not `density=True`.
+5. `p = hist_ref.astype(float) / ref.size` (reference bin probabilities).
+6. `hist_cur, _ = np.histogram(cur, bins=edges)`; `q = hist_cur.astype(float) / cur.size`.
+7. `leftover = 1.0 - float(np.sum(q))`. If `leftover > 1e-15` (tolerance for float noise), append **`leftover`** to `q` and **`0.0`** to `p` so both remain proper discrete distributions over the same extended support.
+8. Apply small smoothing (existing module constant `EPSILON` is fine): add `EPSILON` to `p` and `q`, renormalize each to sum 1.
+9. `m = 0.5 * (p + q)`; compute symmetric JS via KL terms as today, e.g. `inner = 0.5 * (sum(p*log(p/m)) + sum(q*log(q/m)))`.
+10. Return `sqrt(max(inner, 0.0))` to guard against tiny negative `inner` from floating-point error.
+
+## Non-goals / notes
+
+- **Numerical parity** with the old `density=True` implementation is not required; parity with **NannyML** or **scipy** JS is desirable but optional. The priority is **finite, interpretable** drift when the chunk is outside the reference histogram range.
+- **Multivariate** drift in the same file is unchanged by this spec.
+- **Tests** in `sientia_model` should cover: (a) chunk entirely above reference max, (b) entirely below reference min, (c) overlapping range, (d) `ref` or `cur` all-NaN after cleaning.
+
+## Reference (external)
+
+- NannyML continuous JS uses count-based bin probabilities and a **leftover** mass bin; see `ContinuousJensenShannonDistance` in `nannyml/drift/univariate/methods.py` (`_calculate`, `leftover = 1 - np.sum(...)`).
+- Historical NaN-in-reference issue: [NannyML#339](https://github.com/NannyML/nannyml/issues/339) / [#340](https://github.com/NannyML/nannyml/pull/340) (orthogonal to out-of-range mass, but relevant for input cleaning).

e2e/__init__.py

@@ -0,0 +1,3 @@
+"""
+End-to-end tests for laborious temporal workflows.
+"""

e2e/conftest.py

@@ -0,0 +1,603 @@
+"""Pytest configuration and fixtures for E2E tests."""
+
+import asyncio
+import time
+from concurrent.futures import ThreadPoolExecutor
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pandas as pd
+import pytest
+import pytest_asyncio
+from sqlalchemy import create_engine
+from testcontainers.core.container import DockerContainer
+from testcontainers.minio import MinioContainer
+from testcontainers.postgres import PostgresContainer
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.opc_test_server import OpcE2ETestServer
+from laborious.activities.activities import Activities
+from laborious.workflows.drift import Drift
+from laborious.workflows.minimal_retrain import MinimalRetrain
+from laborious.workflows.predictions_batch import PredictionsBatch
+from laborious.workflows.simple_metrics import SimpleMetrics
+from laborious.workflows.sub_workflows.format_and_export_prediction import (
+    FormatAndExportPrediction,
+)
+from laborious.workflows.sub_workflows.prediction_process import PredictionProcess
+from sientia_do.notifications.handlers import CoreNotificationHandler
+from sientia_do.observability.logger import Logger
+from sientia_do.observability.metrics_controller import MetricsController
+
+# Single source of truth for the test database schema. Mirrors the production
+# DDL for ``sientia_data`` so any production change can be pasted directly into
+# this file (see ``e2e/db_schema.sql``) without touching Python.
+DB_SCHEMA_SQL_PATH = Path(__file__).parent / 'db_schema.sql'
+
+
+@pytest_asyncio.fixture(scope='session')
+def postgres_container():
+    """PostgreSQL testcontainer used by all E2E tests."""
+    postgres = PostgresContainer('postgres:15')
+    postgres.start()
+    yield postgres
+    postgres.stop()
+
+
+@pytest_asyncio.fixture(scope='session')
+def minio_container():
+    """MinIO testcontainer used by E2E offload and payload retrieval paths."""
+    minio = MinioContainer()
+    minio.start()
+    yield minio
+    minio.stop()
+
+
+@pytest_asyncio.fixture(scope='session')
+def mongo_container():
+    """MongoDB testcontainer used by real CoreNotificationHandler."""
+    mongo = DockerContainer('mongo:7').with_exposed_ports(27017)
+    mongo.start()
+    yield mongo
+    mongo.stop()
+
+
+@pytest_asyncio.fixture
+def postgres_engine(postgres_container):
+    """SQLAlchemy engine bound to the PostgreSQL testcontainer."""
+    engine = create_engine(postgres_container.get_connection_url())
+    yield engine
+    engine.dispose()
+
+
+def _create_schema_and_tables(engine):
+    """
+    Create all schemas/tables required by workflow and activity paths.
+
+    Loads the DDL from ``e2e/db_schema.sql`` (single source of truth that
+    mirrors the production schema). The SQL file is executed via the raw
+    DBAPI cursor so multi-statement DDL is supported.
+    """
+    sql_text = DB_SCHEMA_SQL_PATH.read_text(encoding='utf-8')
+    with engine.begin() as conn:
+        conn.exec_driver_sql(sql_text)
+
+
+@pytest_asyncio.fixture(autouse=True)
+def setup_postgres_schema_and_tables(postgres_engine):
+    """Ensure required schema and tables exist before each E2E test."""
+    _create_schema_and_tables(postgres_engine)
+    yield
+
+
+@pytest_asyncio.fixture
+def mock_logger():
+    """Logger double with readable console output for E2E runs."""
+    logger = MagicMock(spec=Logger)
+    logger.info = MagicMock(side_effect=lambda msg: print(f'[LOG] {msg}'))
+    logger.debug = MagicMock(side_effect=lambda msg: print(f'[LOG] {msg}'))
+    logger.error = MagicMock(side_effect=lambda msg: print(f'[LOG] {msg}'))
+    logger.warning = MagicMock(side_effect=lambda msg: print(f'[LOG] {msg}'))
+    logger.custom_info = MagicMock(side_effect=lambda msg, _meta=None: print(f'[LOG] {msg}'))
+    logger.custom_debug = MagicMock(side_effect=lambda msg, _meta=None: print(f'[LOG] {msg}'))
+    logger.custom_error = MagicMock(side_effect=lambda msg, _meta=None: print(f'[LOG] {msg}'))
+    logger.custom_warning = MagicMock(side_effect=lambda msg, _meta=None: print(f'[LOG] {msg}'))
+    return logger
+
+
+@pytest_asyncio.fixture
+def metrics_controller(mock_logger):
+    """Real metrics controller for E2E observability paths."""
+    return MetricsController(logger=mock_logger)
+
+
+@pytest_asyncio.fixture
+def notification_handler(mock_logger, mongo_container):
+    """Real notification handler using MongoDB testcontainer."""
+    mongo_port = mongo_container.get_exposed_port(27017)
+    handler = CoreNotificationHandler(
+        connection_string=f'mongodb://localhost:{mongo_port}',
+        database='test_db',
+        logger=mock_logger,
+        project_name='laborious',
+    )
+    try:
+        yield handler
+    finally:
+        handler.shutdown()
+
+
+@pytest.fixture
+def notification_inserts(notification_handler):
+    """Spy on real Mongo insert calls issued by notification handler."""
+    collection = notification_handler.mongo_collection
+    original_insert_one = collection.insert_one
+    spy = MagicMock(wraps=original_insert_one)
+    collection.insert_one = spy
+    try:
+        yield spy
+    finally:
+        collection.insert_one = original_insert_one
+
+
+class _FakeModelWrapper:
+    """External MLflow wrapper double used by repository stub."""
+
+    def __init__(self):
+        self.transform = MagicMock(side_effect=self._default_transform)
+        self.predict = MagicMock(side_effect=self._default_predict)
+
+    @staticmethod
+    def _default_transform(data: pd.DataFrame):
+        result = pd.DataFrame(
+            {
+                'feature_1': [0.234] * len(data),
+                'feature_2': [0.783] * len(data),
+            }
+        )
+        result.index = data.index
+        return result, {}
+
+    @staticmethod
+    def _default_predict(_params: dict, data: pd.DataFrame):
+        pred = pd.DataFrame([0.5] * len(data), columns=['placeholder'])
+        pred.index = data.index
+        return pred, {}
+
+
+@pytest_asyncio.fixture
+def mlflow_repository_stub():
+    """External MLflow repository stub."""
+    repo = MagicMock()
+    wrapper = _FakeModelWrapper()
+    repo.stub_wrapper = wrapper
+    repo.get_cached_model = MagicMock(return_value=wrapper)
+    repo._client = MagicMock()
+    return repo
+
+
+class _FakePIWebAPIClient:
+    """External PI Web API client stub with deterministic responses."""
+
+    def __init__(self):
+        self._responses = None
+        self.write_value = MagicMock(side_effect=self._write_value)
+        self.close = MagicMock()
+
+    def set_side_effect(self, side_effect):
+        self._responses = side_effect
+
+    def _write_value(self, web_ids, value, metadata=None, **kwargs):
+        if isinstance(self._responses, Exception):
+            raise self._responses
+        if isinstance(self._responses, list):
+            item = self._responses.pop(0)
+            if isinstance(item, Exception):
+                raise item
+            return item
+        if callable(self._responses):
+            return self._responses(web_ids=web_ids, value=value, metadata=metadata, **kwargs)
+        return [{'WebId': wid, 'Errors': []} for wid in web_ids]
+
+
+@pytest_asyncio.fixture
+def pi_web_api_client_stub():
+    """PI Web API stub fixture."""
+    return _FakePIWebAPIClient()
+
+
+@pytest_asyncio.fixture
+def opc_repository_stub():
+    """OPC external dependency stub."""
+    repo = MagicMock()
+    repo.write_data = MagicMock(return_value=(True, {'response_time': 0.1}))
+    repo.disconnect = MagicMock()
+    return repo
+
+
+@pytest_asyncio.fixture
+def plugin_store_stub():
+    """Plugin store external dependency stub."""
+    return MagicMock()
+
+
+@pytest_asyncio.fixture(scope='function')
+async def test_activities(
+    postgres_container,
+    minio_container,
+    mock_logger,
+    notification_handler,
+    metrics_controller,
+    mlflow_repository_stub,
+    plugin_store_stub,
+    pi_web_api_client_stub,
+    opc_repository_stub,
+):
+    """Activities with real infra and external-system stubs only."""
+    minio_client = minio_container.get_client()
+    if not minio_client.bucket_exists('test-bucket'):
+        minio_client.make_bucket('test-bucket')
+    minio_port = minio_container.get_exposed_port(9000)
+
+    activities = Activities(
+        postgres_config={
+            'host': 'localhost',
+            'port': int(postgres_container.get_exposed_port(5432)),
+            'user': postgres_container.username,
+            'password': postgres_container.password,
+            'dbname': postgres_container.dbname,
+            'min_connections': 1,
+            'max_connections': 5,
+        },
+        plugin_store=plugin_store_stub,
+        minio_config={
+            'endpoint_url': f'localhost:{minio_port}',
+            'access_key': 'minioadmin',
+            'secret_key': 'minioadmin',
+            'default_bucket': 'test-bucket',
+            'retention_hours': 24,
+            'secure': False,
+        },
+        opc_config={},
+        pi_web_api_config={
+            'base_url': 'http://localhost:8080',
+            'auth_type': 'bearer',
+            'auth_token': 'test_token',
+        },
+        logger=mock_logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+        mlflow_repository=mlflow_repository_stub,
+    )
+    activities.pi_web_api_client = pi_web_api_client_stub
+    activities.opc_repository = {'1': opc_repository_stub}
+    try:
+        yield activities
+    finally:
+        activities.shutdown()
+
+
+@pytest_asyncio.fixture(scope='function')
+async def test_activities_real_minio(
+    postgres_container,
+    minio_container,
+    mock_logger,
+    notification_handler,
+    metrics_controller,
+    mlflow_repository_stub,
+    plugin_store_stub,
+    pi_web_api_client_stub,
+    opc_repository_stub,
+):
+    """Compatibility alias for offload tests."""
+    minio_client = minio_container.get_client()
+    if not minio_client.bucket_exists('test-bucket'):
+        minio_client.make_bucket('test-bucket')
+    minio_port = minio_container.get_exposed_port(9000)
+
+    activities = Activities(
+        postgres_config={
+            'host': 'localhost',
+            'port': int(postgres_container.get_exposed_port(5432)),
+            'user': postgres_container.username,
+            'password': postgres_container.password,
+            'dbname': postgres_container.dbname,
+            'min_connections': 1,
+            'max_connections': 5,
+        },
+        plugin_store=plugin_store_stub,
+        minio_config={
+            'endpoint_url': f'localhost:{minio_port}',
+            'access_key': 'minioadmin',
+            'secret_key': 'minioadmin',
+            'default_bucket': 'test-bucket',
+            'retention_hours': 24,
+            'secure': False,
+        },
+        opc_config={},
+        pi_web_api_config={
+            'base_url': 'http://localhost:8080',
+            'auth_type': 'bearer',
+            'auth_token': 'test_token',
+        },
+        logger=mock_logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+        mlflow_repository=mlflow_repository_stub,
+    )
+    activities.pi_web_api_client = pi_web_api_client_stub
+    activities.opc_repository = {'1': opc_repository_stub}
+    try:
+        yield activities
+    finally:
+        activities.shutdown()
+
+
+def _worker_activity_list(test_activities: Activities):
+    """List of registered activity callables used by Temporal worker in E2E."""
+    return [
+        test_activities.load_query_with_minio_offload,
+        test_activities.cleanup_minio_objects_expired,
+        test_activities.input_gate,
+        test_activities.request_transform,
+        test_activities.mlflow_response_gate,
+        test_activities.mlflow_content_gate,
+        test_activities.request_predict,
+        test_activities.repeat_last_prediction,
+        test_activities.format_prediction,
+        test_activities.format_transformed_data,
+        test_activities.format_default_prediction,
+        test_activities.write_pi_web_api_data,
+        test_activities.write_opc_data,
+        test_activities.export_data_to_postgres,
+        test_activities.export_payload_to_postgres,
+        test_activities.write_metrics,
+    ]
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_test_env():
+    """Temporal test environment with time-skipping."""
+    env = await WorkflowEnvironment.start_time_skipping()
+    async with env:
+        yield env
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_worker(temporal_test_env, test_activities):
+    """Temporal worker for full predictions-batch and child workflows."""
+    with ThreadPoolExecutor(max_workers=32) as activity_executor:
+        async with Worker(
+            temporal_test_env.client,
+            task_queue='test-queue',
+            workflows=[PredictionsBatch, PredictionProcess, FormatAndExportPrediction],
+            activities=_worker_activity_list(test_activities),
+            activity_executor=activity_executor,
+        ) as worker:
+            yield worker
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_worker_real_minio(temporal_test_env, test_activities_real_minio):
+    """Temporal worker alias for tests that emphasize MinIO behavior."""
+    with ThreadPoolExecutor(max_workers=32) as activity_executor:
+        async with Worker(
+            temporal_test_env.client,
+            task_queue='test-queue',
+            workflows=[PredictionsBatch, PredictionProcess, FormatAndExportPrediction],
+            activities=_worker_activity_list(test_activities_real_minio),
+            activity_executor=activity_executor,
+        ) as worker:
+            yield worker
+
+
+def _drift_worker_activity_list(test_activities: Activities):
+    """Activity callables registered on the drift Temporal worker."""
+    return [
+        test_activities.load_custom_query,
+        test_activities.get_reference_data,
+        test_activities.calculate_drift,
+        test_activities.export_data_to_postgres,
+    ]
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_worker_drift(temporal_test_env, test_activities):
+    """Temporal worker registered with the Drift workflow and its activities."""
+    with ThreadPoolExecutor(max_workers=32) as activity_executor:
+        async with Worker(
+            temporal_test_env.client,
+            task_queue='test-queue',
+            workflows=[Drift],
+            activities=_drift_worker_activity_list(test_activities),
+            activity_executor=activity_executor,
+        ) as worker:
+            yield worker
+
+
+def _simple_metrics_worker_activity_list(test_activities: Activities):
+    """Activity callables registered on the simple-metrics Temporal worker."""
+    return [
+        test_activities.load_custom_query,
+        test_activities.calculate_simple_metrics,
+        test_activities.export_data_to_postgres,
+    ]
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_worker_simple_metrics(temporal_test_env, test_activities):
+    """Temporal worker registered with the SimpleMetrics workflow and its activities."""
+    with ThreadPoolExecutor(max_workers=32) as activity_executor:
+        async with Worker(
+            temporal_test_env.client,
+            task_queue='test-queue',
+            workflows=[SimpleMetrics],
+            activities=_simple_metrics_worker_activity_list(test_activities),
+            activity_executor=activity_executor,
+        ) as worker:
+            yield worker
+
+
+def _minimal_retrain_worker_activity_list(test_activities: Activities):
+    """Activity callables registered on the minimal-retrain Temporal worker."""
+    return [
+        test_activities.load_query_with_minio_offload,
+        test_activities.retrain_model,
+        test_activities.update_production_model,
+        test_activities.format_retrain_report,
+        test_activities.export_data_to_postgres,
+    ]
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_worker_minimal_retrain(temporal_test_env, test_activities):
+    """Temporal worker registered with the MinimalRetrain workflow and its activities."""
+    with ThreadPoolExecutor(max_workers=32) as activity_executor:
+        async with Worker(
+            temporal_test_env.client,
+            task_queue='test-queue',
+            workflows=[MinimalRetrain],
+            activities=_minimal_retrain_worker_activity_list(test_activities),
+            activity_executor=activity_executor,
+        ) as worker:
+            yield worker
+
+
+def _connect_activities_to_opc_server(activities: Activities, server_url: str) -> None:
+    """
+    Initialize OPC repositories and block until the E2E server session is ready.
+
+    Runs synchronously (typically via ``asyncio.to_thread``) so the asyncua test
+    server event loop is not blocked during ``Client.connect()``.
+
+    Args:
+        activities (Activities): Worker activities under test.
+        server_url (str): ``opc.tcp://`` URL from ``OpcE2ETestServer``.
+    """
+    activities.init_opc()
+    repo = activities.opc_repository['1']
+    deadline = time.monotonic() + 30.0
+    while time.monotonic() < deadline:
+        if repo._session_ready.is_set():
+            return
+        connected, _ = repo.connect()
+        if connected:
+            return
+        time.sleep(0.5)
+    raise RuntimeError(f'Could not connect OpcRepository to OPC E2E server at {server_url}')
+
+
+def _build_e2e_opc_config(server_url: str) -> dict[str, dict]:
+    """
+    OPC server config for E2E Activities pointing at an in-process asyncua server.
+
+    Args:
+        server_url (str): ``opc.tcp://`` endpoint from ``OpcE2ETestServer``.
+
+    Return:
+        dict: ``opc_config`` payload for ``Activities`` (server id ``1``).
+    """
+    return {
+        '1': {
+            'id': '1',
+            'server_name': 'e2e_opcua',
+            'url': server_url,
+            'server_uri': server_url,
+            'cert_path': None,
+            'private_key_path': None,
+            'server_cert_path': None,
+            'reconnection_interval': 0,
+        }
+    }
+
+
+@pytest_asyncio.fixture
+async def opc_e2e_server():
+    """In-process asyncua server with writable prediction/confidence nodes."""
+    server = OpcE2ETestServer()
+    await server.start()
+    await asyncio.sleep(0.5)
+    try:
+        yield server
+    finally:
+        await server.stop()
+
+
+@pytest_asyncio.fixture(scope='function')
+async def test_activities_real_opc(
+    postgres_container,
+    minio_container,
+    mock_logger,
+    notification_handler,
+    metrics_controller,
+    mlflow_repository_stub,
+    plugin_store_stub,
+    pi_web_api_client_stub,
+    opc_e2e_server: OpcE2ETestServer,
+):
+    """Activities with real OpcRepository connected to the in-process OPC UA server."""
+    minio_client = minio_container.get_client()
+    if not minio_client.bucket_exists('test-bucket'):
+        minio_client.make_bucket('test-bucket')
+    minio_port = minio_container.get_exposed_port(9000)
+
+    activities = Activities(
+        postgres_config={
+            'host': 'localhost',
+            'port': int(postgres_container.get_exposed_port(5432)),
+            'user': postgres_container.username,
+            'password': postgres_container.password,
+            'dbname': postgres_container.dbname,
+            'min_connections': 1,
+            'max_connections': 5,
+        },
+        plugin_store=plugin_store_stub,
+        minio_config={
+            'endpoint_url': f'localhost:{minio_port}',
+            'access_key': 'minioadmin',
+            'secret_key': 'minioadmin',
+            'default_bucket': 'test-bucket',
+            'retention_hours': 24,
+            'secure': False,
+        },
+        opc_config=_build_e2e_opc_config(opc_e2e_server.url),
+        pi_web_api_config={
+            'base_url': 'http://localhost:8080',
+            'auth_type': 'bearer',
+            'auth_token': 'test_token',
+        },
+        logger=mock_logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+        mlflow_repository=mlflow_repository_stub,
+    )
+    activities.pi_web_api_client = pi_web_api_client_stub
+    await asyncio.to_thread(_connect_activities_to_opc_server, activities, opc_e2e_server.url)
+    try:
+        yield activities
+    finally:
+        await asyncio.to_thread(_teardown_real_opc_activities, activities)
+
+
+def _teardown_real_opc_activities(activities: Activities) -> None:
+    """Disconnect OPC sessions and shut down activities (sync, for asyncio.to_thread)."""
+    for opc_repo in activities.opc_repository.values():
+        opc_repo.disconnect()
+    activities.shutdown()
+
+
+@pytest_asyncio.fixture(scope='function')
+async def temporal_worker_real_opc(temporal_test_env, test_activities_real_opc):
+    """Temporal worker using real OpcRepository against the in-process OPC UA server."""
+    with ThreadPoolExecutor(max_workers=32) as activity_executor:
+        async with Worker(
+            temporal_test_env.client,
+            task_queue='test-queue',
+            workflows=[PredictionsBatch, PredictionProcess, FormatAndExportPrediction],
+            activities=_worker_activity_list(test_activities_real_opc),
+            activity_executor=activity_executor,
+        ) as worker:
+            yield worker
+
+

e2e/db_schema.sql

@@ -0,0 +1,109 @@
+-- =============================================================================
+-- E2E test database schema for the ``sientia_data`` namespace.
+--
+-- Mirrors the production DDL one-to-one so any change in production can be
+-- pasted directly into this file. The conftest fixture loads this SQL into the
+-- testcontainers Postgres before each test run.
+--
+-- Notes on differences from production:
+--  * Tables that are partitioned in production (e.g. ``simple_metrics``,
+--    ``transformed_data``, ``drift_metrics``) are created as plain tables
+--    here because the test suite does not exercise partition pruning.
+--  * Indexes are intentionally omitted; tests rely on functional behavior,
+--    not query plans.
+-- =============================================================================
+
+CREATE SCHEMA IF NOT EXISTS sientia_data;
+
+-- -----------------------------------------------------------------------------
+-- sientia_data.laborious_data
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS sientia_data.laborious_data (
+    model_id    int4 NOT NULL,
+    variable    text NOT NULL,
+    value       numeric NULL,
+    "timestamp" timestamptz NOT NULL,
+    created_at  timestamptz DEFAULT CURRENT_TIMESTAMP NOT NULL,
+    CONSTRAINT unique_timestamp_variable
+        UNIQUE (model_id, "timestamp", variable)
+);
+
+-- -----------------------------------------------------------------------------
+-- sientia_data.predictions
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS sientia_data.predictions (
+    model_id              int4 NOT NULL,
+    prediction            numeric NULL,
+    prediction_confidence numeric NOT NULL,
+    response_time         numeric NOT NULL,
+    prediction_status     text NOT NULL,
+    "timestamp"           timestamptz NOT NULL,
+    created_at            timestamptz DEFAULT CURRENT_TIMESTAMP NOT NULL,
+    "comments"            text NULL,
+    CONSTRAINT unique_model_id_timestamp
+        UNIQUE (model_id, "timestamp")
+);
+
+-- -----------------------------------------------------------------------------
+-- sientia_data.transformed_data
+--   Production: PARTITION BY RANGE (created_at). Tests use a plain table.
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS sientia_data.transformed_data (
+    id          SERIAL NOT NULL,
+    model_id    int4 NOT NULL,
+    variable    text NOT NULL,
+    value       numeric NULL,
+    "timestamp" timestamptz NOT NULL,
+    created_at  timestamptz DEFAULT CURRENT_TIMESTAMP NOT NULL,
+    PRIMARY KEY (id, created_at)
+);
+
+-- -----------------------------------------------------------------------------
+-- sientia_data.drift_metrics
+--   Production: PARTITION BY RANGE (created_at). Tests use a plain table.
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS sientia_data.drift_metrics (
+    id               SERIAL NOT NULL,
+    model_id         text NOT NULL,
+    feature          text NULL,
+    method           text NOT NULL,
+    value            numeric NOT NULL,
+    alert            bool NOT NULL,
+    chunk_index      int4 NOT NULL,
+    chunk_start_date text NOT NULL,
+    chunk_end_date   text NOT NULL,
+    accurate         bool NOT NULL,
+    "timestamp"      timestamptz NULL,
+    created_at       timestamptz DEFAULT CURRENT_TIMESTAMP NOT NULL,
+    PRIMARY KEY (id, created_at)
+);
+
+-- -----------------------------------------------------------------------------
+-- sientia_data.simple_metrics
+--   Production: PARTITION BY RANGE (created_at). Tests use a plain table.
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS sientia_data.simple_metrics (
+    id               SERIAL NOT NULL,
+    model_id         text NOT NULL,
+    metric           text NOT NULL,
+    value            numeric NOT NULL,
+    "timestamp"      timestamptz NULL,
+    data_size        int4 NOT NULL,
+    interval_minutes int4 NOT NULL,
+    created_at       timestamptz DEFAULT CURRENT_TIMESTAMP NOT NULL,
+    PRIMARY KEY (id, created_at)
+);
+
+-- -----------------------------------------------------------------------------
+-- sientia_data.log_retrain
+--   No primary key in production; all columns nullable.
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS sientia_data.log_retrain (
+    mlflow_experiment_id int8 NULL,
+    mlflow_run_id        text NULL,
+    model_id             text NULL,
+    model_name           text NULL,
+    status               text NULL,
+    "timestamp"          timestamptz NULL,
+    "version"            text NULL
+);

e2e/helpers.py

@@ -0,0 +1,368 @@
+"""
+Shared helpers for E2E tests (Temporal workflows + PostgreSQL).
+"""
+
+import asyncio
+import json
+from datetime import datetime
+from decimal import Decimal
+from pathlib import Path
+from typing import Any
+
+from sqlalchemy import text
+from sqlalchemy.engine import Engine
+
+SCENARIO_INPUTS_DIR = Path(__file__).parent / 'scenario_inputs'
+
+
+def _replace_template_values(payload: Any, model_id: int) -> Any:
+    """
+    Replace string placeholders in scenario payloads with the concrete model id.
+
+    Args:
+        payload: JSON-like structure loaded from scenario input file.
+        model_id: Model id used to render template placeholders.
+
+    Return:
+        Any: Payload with ``{{MODEL_ID}}`` replaced where applicable.
+    """
+    if isinstance(payload, dict):
+        return {key: _replace_template_values(value, model_id) for key, value in payload.items()}
+    if isinstance(payload, list):
+        return [_replace_template_values(item, model_id) for item in payload]
+    if isinstance(payload, str):
+        if payload == '{{MODEL_ID}}':
+            return model_id
+        return payload.replace('{{MODEL_ID}}', str(model_id))
+    return payload
+
+
+def load_scenario_input(file_name: str, model_id: int | None = None) -> dict[str, Any]:
+    """
+    Load a scenario input JSON from ``e2e/scenario_inputs``.
+
+    Args:
+        file_name: JSON file name inside ``e2e/scenario_inputs``.
+        model_id: Optional model id used to render ``{{MODEL_ID}}`` placeholders.
+
+    Return:
+        dict[str, Any]: Input payload ready to be passed to workflow/activity calls.
+    """
+    file_path = SCENARIO_INPUTS_DIR / file_name
+    with file_path.open('r', encoding='utf-8') as f:
+        payload = json.load(f)
+
+    if model_id is not None:
+        return _replace_template_values(payload, model_id)
+    return payload
+
+
+async def start_and_await_workflow(client, workflow_run, input_data: dict, workflow_id: str, timeout: float = 60.0):
+    """
+    Start a workflow and wait for its result.
+
+    Args:
+        client: Temporal client from WorkflowEnvironment.
+        workflow_run: Workflow run method (e.g. PredictionsBatch.run).
+        input_data: Workflow input payload.
+        workflow_id: Unique workflow id.
+        timeout: Max seconds to wait for completion.
+
+    Return:
+        Workflow result value.
+    """
+    handle = await client.start_workflow(
+        workflow_run,
+        input_data,
+        id=workflow_id,
+        task_queue='test-queue',
+    )
+    return await asyncio.wait_for(handle.result(), timeout=timeout)
+
+
+DEFAULT_BATCH_TIMESTAMP = '2024-01-01 12:00:00+00:00'
+DEFAULT_PREDICTION_HISTORY_TIMESTAMP = '2024-01-01 12:00:00+00:00'
+
+
+def insert_sample_data(
+    postgres_engine: Engine,
+    model_id: int,
+    values: list[Any],
+    *,
+    data_timestamp: str = DEFAULT_BATCH_TIMESTAMP,
+) -> None:
+    """
+    Replace laborious_data rows for a model_id with one row per value (sensor_1..n).
+
+    Args:
+        postgres_engine: SQLAlchemy engine.
+        model_id: Model id column value.
+        values: Per-sensor values; use string 'NULL' for SQL NULL.
+        data_timestamp: Timestamp and created_at for every inserted row; drives
+            ``last_timestamp`` on the MinIO/query payload (max row time).
+    """
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+        values_sql = []
+        for i, value in enumerate(values):
+            values_sql.append(f"""
+                ({model_id}, 'sensor_{i + 1}', {value}, '{data_timestamp}', '{data_timestamp}')
+            """)
+        insert_sql = f"""
+            INSERT INTO sientia_data.laborious_data (model_id, variable, value, timestamp, created_at)
+            VALUES
+                {', '.join(values_sql)}
+        """
+        conn.execute(text(insert_sql))
+
+
+def insert_sample_prediction(
+    postgres_engine: Engine,
+    model_id: int,
+    *,
+    prediction_timestamp: str = DEFAULT_PREDICTION_HISTORY_TIMESTAMP,
+) -> tuple[int, Decimal, Decimal, str]:
+    """
+    Insert a single historical prediction row for REPEAT scenarios.
+
+    Args:
+        postgres_engine: SQLAlchemy engine.
+        model_id: Model id.
+        prediction_timestamp: Row ``timestamp`` (unique with model_id in tests).
+
+    Return:
+        tuple: (model_id, prediction, prediction_confidence, prediction_status) for assertions.
+    """
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.predictions WHERE model_id = {model_id}'))
+        insert_sql = f"""
+            INSERT INTO sientia_data.predictions (
+                model_id, timestamp, prediction, prediction_confidence, prediction_status, comments, response_time
+            )
+            VALUES (
+                {model_id}, '{prediction_timestamp}', 10, 0, 'Good', '', 0.1
+            )
+        """
+        conn.execute(text(insert_sql))
+    return (model_id, Decimal(10), Decimal(0), 'Good')
+
+
+def workflow_failure_message_chain(exc: BaseException) -> list[str]:
+    """
+    Collect ``str()`` / ``message`` from an exception and its ``__cause__`` chain.
+
+    Args:
+        exc: Root exception (e.g. from ``pytest.raises``).
+
+    Return:
+        list[str]: Messages from root to innermost cause.
+    """
+    messages: list[str] = []
+    current: BaseException | None = exc
+    while current is not None:
+        messages.append(getattr(current, 'message', None) or str(current) or repr(current))
+        current = current.__cause__
+    return messages
+
+
+def assert_postgres_unique_violation_in_chain(exc: BaseException) -> None:
+    """
+    Assert the exception chain mentions Postgres unique-constraint violation.
+
+    Args:
+        exc: Workflow or activity error from Temporal.
+
+    Raises:
+        AssertionError: If no link in the chain looks like UniqueViolation.
+    """
+    chain = ' | '.join(workflow_failure_message_chain(exc))
+    assert 'UniqueViolation' in chain or 'unique_model_id_timestamp' in chain, (
+        f'Expected unique constraint violation in error chain, got: {chain}'
+    )
+
+
+def assert_prediction_row_count(postgres_engine: Engine, model_id: int, expected: int) -> None:
+    """
+    Assert how many prediction rows exist for a model_id.
+
+    Args:
+        postgres_engine: SQLAlchemy engine.
+        model_id: Model id filter.
+        expected: Expected row count.
+    """
+    with postgres_engine.connect() as conn:
+        n = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.predictions WHERE model_id = :m'),
+            {'m': model_id},
+        ).scalar()
+    assert n == expected, f'Expected {expected} prediction rows, got {n}'
+
+
+def assert_prediction(
+    postgres_engine: Engine,
+    model_id: int,
+    prediction: float = 0.5,
+    prediction_confidence: int | Decimal = 0,
+    prediction_status: str = 'Good',
+    comments: str = '',
+    comments_contains: str | None = None,
+) -> None:
+    """
+    Assert exactly one prediction row exists for model_id with expected columns.
+
+    Args:
+        postgres_engine: SQLAlchemy engine.
+        model_id: Expected model_id.
+        prediction: Expected prediction value.
+        prediction_confidence: Expected confidence (int or Decimal for numeric column).
+        prediction_status: Expected status string.
+        comments: Expected exact comments string (ignored when ``comments_contains`` is set).
+        comments_contains: When set, assert this substring appears in comments.
+    """
+    with postgres_engine.connect() as conn:
+        result_query = conn.execute(
+            text(
+                f'SELECT model_id, prediction, prediction_confidence, prediction_status, comments '
+                f'FROM sientia_data.predictions WHERE model_id = {model_id} '
+                f'ORDER BY created_at ASC'
+            )
+        )
+        prediction_rows = result_query.fetchall()
+        assert len(prediction_rows) == 1, f'Expected one prediction record, got {len(prediction_rows)}'
+        row = prediction_rows[0]
+        assert row[0] == model_id, f'Expected model_id={model_id}, got {row[0]}'
+        assert row[1] == prediction or Decimal(str(row[1])) == Decimal(str(prediction)), (
+            f'Expected prediction={prediction}, got {row[1]}'
+        )
+        assert row[2] == prediction_confidence or Decimal(str(row[2])) == Decimal(
+            str(prediction_confidence)
+        ), f'Expected prediction_confidence={prediction_confidence}, got {row[2]}'
+        assert row[3] == prediction_status, f"Expected prediction_status='{prediction_status}', got {row[3]}"
+        actual_comments = row[4] or ''
+        if comments_contains is not None:
+            assert comments_contains in actual_comments, (
+                f"Expected comments to contain '{comments_contains}', got '{actual_comments}'"
+            )
+        else:
+            assert actual_comments == comments, f"Expected comments='{comments}', got '{actual_comments}'"
+
+
+def assert_continue(
+    postgres_engine: Engine,
+    model_id: int,
+    prediction_confidence: Decimal = Decimal(2),
+    comments: str = 'Input data with bad quality',
+) -> None:
+    """Assert one default-style prediction row after CONTINUE gate path."""
+    with postgres_engine.connect() as conn:
+        result_query = conn.execute(
+            text(
+                f'SELECT model_id, prediction, prediction_confidence, prediction_status, comments '
+                f'FROM sientia_data.predictions WHERE model_id = {model_id}'
+            )
+        )
+        prediction_rows = result_query.fetchall()
+        assert len(prediction_rows) == 1, 'Expected one prediction record despite warnings'
+        row = prediction_rows[0]
+        assert row[1] == 0, f'Expected prediction=0, got {row[1]}'
+        assert row[2] == prediction_confidence, (
+            f'Expected prediction_confidence={prediction_confidence}, got {row[2]}'
+        )
+        assert row[3] == 'Bad', f"Expected prediction_status='Bad', got {row[3]}"
+        assert row[4] == comments, f"Expected comments='{comments}', got {row[4]}"
+
+
+def assert_stop(postgres_engine: Engine, model_id: int) -> None:
+    """Assert no prediction rows for model_id."""
+    import pytest
+
+    with postgres_engine.connect() as conn:
+        result_query = conn.execute(
+            text(f'SELECT COUNT(*) FROM sientia_data.predictions WHERE model_id = {model_id}')
+        )
+        count = result_query.scalar()
+        assert count == 0, f'Expected no predictions, but found {count} records'
+
+
+def assert_repeat(postgres_engine: Engine, model_id: int, last_prediction: tuple) -> None:
+    """
+    Assert two prediction rows for model_id both match last_prediction.
+
+    Rows are compared in created_at order for stability.
+
+    Args:
+        postgres_engine: SQLAlchemy engine.
+        model_id: Model id.
+        last_prediction: Tuple (model_id, prediction, confidence, status) to match both rows.
+    """
+    import pytest
+
+    with postgres_engine.connect() as conn:
+        result_query = conn.execute(
+            text(
+                f'SELECT model_id, prediction, prediction_confidence, prediction_status '
+                f'FROM sientia_data.predictions WHERE model_id = {model_id} '
+                f'ORDER BY created_at ASC'
+            )
+        )
+        prediction_rows = result_query.fetchall()
+        assert len(prediction_rows) == 2, 'Expected two prediction records'
+        assert prediction_rows[0] == last_prediction, (
+            f'Expected first row {last_prediction}, got {prediction_rows[0]}'
+        )
+        assert prediction_rows[1] == last_prediction, (
+            f'Expected second row {last_prediction}, got {prediction_rows[1]}'
+        )
+
+
+def make_workflow_id(prefix: str) -> str:
+    """Build a unique workflow id using a prefix and current timestamp."""
+    return f'{prefix}-{datetime.now().timestamp()}'
+
+
+def insert_target_data_for_drift(
+    postgres_engine: Engine,
+    model_id: int,
+    timestamps: list[str],
+    variables_values: dict[str, list[float]],
+) -> None:
+    """
+    Insert one row per (timestamp, variable) pair into ``laborious_data``.
+
+    Used by drift scenarios that need wide-format input where the pivot keeps a
+    full row for every timestamp.
+
+    Args:
+     - postgres_engine: SQLAlchemy engine bound to the test container.
+     - model_id: Model id stamped on every row.
+     - timestamps: ISO-8601 strings used both as ``timestamp`` and ``created_at``.
+     - variables_values: Mapping of variable name to a list of values; each list
+       must be the same length as ``timestamps``.
+    """
+    for var_name, values in variables_values.items():
+        if len(values) != len(timestamps):
+            raise ValueError(
+                f"Variable '{var_name}' has {len(values)} values but {len(timestamps)} timestamps"
+            )
+
+    rows_sql = []
+    for index, ts in enumerate(timestamps):
+        for var_name, values in variables_values.items():
+            rows_sql.append(
+                f"({model_id}, '{var_name}', {values[index]}, '{ts}', '{ts}')"
+            )
+
+    with postgres_engine.begin() as conn:
+        conn.execute(
+            text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}')
+        )
+        if rows_sql:
+            conn.execute(
+                text(
+                    'INSERT INTO sientia_data.laborious_data '
+                    '(model_id, variable, value, "timestamp", created_at) VALUES '
+                    + ', '.join(rows_sql)
+                )
+            )
+
+

e2e/opc_test_server.py

@@ -0,0 +1,189 @@
+"""
+In-process OPC UA server for E2E tests (asyncua).
+
+Provides writable prediction/confidence nodes and optional write faults
+(Tier-1 BadSessionIdInvalid via PreWrite callback).
+"""
+
+from __future__ import annotations
+
+import socket
+from dataclasses import dataclass
+from typing import TYPE_CHECKING
+
+from asyncua import Server, ua
+from asyncua.common.callback import CallbackType
+from asyncua.common.utils import ServiceError
+
+if TYPE_CHECKING:
+    from asyncua.common.node import Node
+
+
+UNKNOWN_NODE_ID = 'ns=99;i=9999'
+
+
+@dataclass(frozen=True)
+class OpcE2ENodeIds:
+    """NodeId strings used in opc_output_config for E2E workflows."""
+
+    prediction: str
+    confidence: str
+    unknown: str = UNKNOWN_NODE_ID
+
+
+class OpcE2ETestServer:
+    """
+    Ephemeral asyncua server with Laborious E2E variables and controllable faults.
+
+    Args:
+        host: Bind address (default 127.0.0.1).
+    """
+
+    def __init__(self, host: str = '127.0.0.1') -> None:
+        self._host = host
+        self._server: Server | None = None
+        self._prediction_node: Node | None = None
+        self._confidence_node: Node | None = None
+        self._session_bad_on_write = False
+        self._url: str | None = None
+        self._node_ids: OpcE2ENodeIds | None = None
+
+    @property
+    def url(self) -> str:
+        if self._url is None:
+            raise RuntimeError('OPC E2E server is not started')
+        return self._url
+
+    @property
+    def node_ids(self) -> OpcE2ENodeIds:
+        if self._node_ids is None:
+            raise RuntimeError('OPC E2E server is not started')
+        return self._node_ids
+
+    def set_session_bad_on_write(self, enabled: bool) -> None:
+        """
+        When enabled, every client Write is rejected with BadSessionIdInvalid.
+
+        Args:
+            enabled (bool): Turn Tier-1 session fault injection on or off.
+        """
+        self._session_bad_on_write = enabled
+
+    async def start(self) -> OpcE2ENodeIds:
+        """
+        Start the OPC UA server on a free TCP port.
+
+        Return:
+            OpcE2ENodeIds: NodeId strings for prediction and confidence tags.
+        """
+        port = _free_port(self._host)
+        self._url = f'opc.tcp://{self._host}:{port}/freeopcua/server/'
+
+        server = Server()
+        server.set_endpoint(self._url)
+        await server.init()
+        server.iserver.callback_service.addListener(
+            CallbackType.PreWrite,
+            self._pre_write_callback,
+        )
+
+        idx = await server.register_namespace('http://sientia.test/laborious-e2e')
+        e2e_object = await server.nodes.objects.add_object(idx, 'LaboriousE2E')
+        prediction = await e2e_object.add_variable(
+            idx,
+            'Prediction',
+            ua.Variant(0.0, ua.VariantType.Float),
+        )
+        confidence = await e2e_object.add_variable(
+            idx,
+            'Confidence',
+            ua.Variant(0.0, ua.VariantType.Float),
+        )
+        await prediction.set_writable()
+        await confidence.set_writable()
+
+        await server.start()
+        self._server = server
+        self._prediction_node = prediction
+        self._confidence_node = confidence
+        self._node_ids = OpcE2ENodeIds(
+            prediction=prediction.nodeid.to_string(),
+            confidence=confidence.nodeid.to_string(),
+        )
+        return self._node_ids
+
+    async def stop(self) -> None:
+        """Stop the OPC UA server and release the listening port."""
+        if self._server is not None:
+            await self._server.stop()
+            self._server = None
+        self._prediction_node = None
+        self._confidence_node = None
+        self._url = None
+        self._node_ids = None
+        self._session_bad_on_write = False
+
+    async def read_prediction(self) -> float:
+        """
+        Read the current prediction variable value from the address space.
+
+        Return:
+            float: Stored prediction value.
+        """
+        if self._prediction_node is None:
+            raise RuntimeError('OPC E2E server is not started')
+        value = await self._prediction_node.read_value()
+        return float(value)
+
+    async def read_confidence(self) -> float:
+        """
+        Read the current confidence variable value from the address space.
+
+        Return:
+            float: Stored confidence value.
+        """
+        if self._confidence_node is None:
+            raise RuntimeError('OPC E2E server is not started')
+        value = await self._confidence_node.read_value()
+        return float(value)
+
+    async def _pre_write_callback(self, _event, _service) -> None:
+        if self._session_bad_on_write:
+            raise ServiceError(ua.StatusCodes.BadSessionIdInvalid)
+
+
+def _free_port(host: str) -> int:
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as sock:
+        sock.bind((host, 0))
+        return int(sock.getsockname()[1])
+
+
+def build_opc_output_config(
+    node_ids: OpcE2ENodeIds,
+    *,
+    prediction_tag: str | None = None,
+    confidence_tag: str | None = None,
+    prediction_only: bool = False,
+    server_key: str = '1',
+) -> dict[str, dict]:
+    """
+    Build opc_output_config for PredictionsBatch using real server NodeIds.
+
+    Args:
+        node_ids (OpcE2ENodeIds): Node ids from OpcE2ETestServer.
+        prediction_tag (str | None): Override prediction NodeId (default: node_ids.prediction).
+        confidence_tag (str | None): Override confidence NodeId (default: node_ids.confidence).
+        prediction_only (bool): When True, omit confidence_tags (single write per activity).
+        server_key (str): OPC server id key in opc_output_config.
+
+    Return:
+        dict: opc_output_config payload for workflow input.
+    """
+    pred = prediction_tag if prediction_tag is not None else node_ids.prediction
+    conf = confidence_tag if confidence_tag is not None else node_ids.confidence
+    server_config: dict = {
+        'prediction_tags': {pred: {'data_type': 'float'}},
+    }
+    if not prediction_only:
+        server_config['confidence_tags'] = {conf: {'data_type': 'float'}}
+    return {server_key: server_config}

e2e/scenario_inputs/drift_base.json

@@ -0,0 +1,14 @@
+{
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "schema": "sientia_data",
+  "source_table_name": "laborious_data",
+  "target_table_name": "drift_metrics",
+  "interval": 60,
+  "drift_metrics": ["kolmogorov_smirnov", "jensen_shannon", "wasserstein"],
+  "chunk_period": "min",
+  "model_config": {
+    "target": "sensor_1"
+  }
+}

e2e/scenario_inputs/format_export_base.json

@@ -0,0 +1,37 @@
+{
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT timestamp, variable, value, created_at FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data",
+  "input_filters": {
+    "EMPTY_DATA": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_transform_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_predict_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "path_priority": ["STOP", "CONTINUE", "REPEAT"],
+  "opc_output_config": {},
+  "pi_web_api_output_config": {},
+  "save_transform": true,
+  "prediction_store_policy": "lts:1",
+  "model_config": {
+    "retention_minutes": 0,
+    "target": "sensor_1"
+  },
+  "datetime_columns": ["timestamp", "created_at"]
+}

e2e/scenario_inputs/main_happy_path.json

@@ -0,0 +1,45 @@
+{
+  "metadata": {
+    "metadata": {
+      "model_id": "{{MODEL_ID}}",
+      "model_name": "test_model",
+      "schedule_name": "test-schedule",
+      "workflow_name": "predictions_batch"
+    }
+  },
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT timestamp, variable, value, created_at FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data",
+  "input_filters": {
+    "EMPTY_DATA": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_transform_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_predict_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "path_priority": ["STOP", "CONTINUE", "REPEAT"],
+  "opc_output_config": {},
+  "pi_web_api_output_config": {},
+  "save_transform": true,
+  "prediction_store_policy": "lts:1",
+  "model_config": {
+    "target": "sensor_1",
+    "retention_minutes": 0
+  },
+  "datetime_columns": ["timestamp", "created_at"]
+}

e2e/scenario_inputs/main_invalid_datetime.json

@@ -0,0 +1,45 @@
+{
+  "metadata": {
+    "metadata": {
+      "model_id": "{{MODEL_ID}}",
+      "model_name": "test_model",
+      "schedule_name": "test-schedule",
+      "workflow_name": "predictions_batch"
+    }
+  },
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT timestamp, variable, value FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data",
+  "input_filters": {
+    "EMPTY_DATA": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_transform_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_predict_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "path_priority": ["STOP", "CONTINUE", "REPEAT"],
+  "opc_output_config": {},
+  "pi_web_api_output_config": {},
+  "save_transform": true,
+  "prediction_store_policy": "lts:1",
+  "model_config": {
+    "target": "sensor_1",
+    "retention_minutes": 0
+  },
+  "datetime_columns": ["nonexistent_column"]
+}

e2e/scenario_inputs/main_missing_required.json

@@ -0,0 +1,16 @@
+{
+  "metadata": {
+    "metadata": {
+      "model_id": "{{MODEL_ID}}",
+      "model_name": "test_model",
+      "schedule_name": "test-schedule",
+      "workflow_name": "predictions_batch"
+    }
+  },
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data"
+}

e2e/scenario_inputs/main_sql_error.json

@@ -0,0 +1,44 @@
+{
+  "metadata": {
+    "metadata": {
+      "model_id": "{{MODEL_ID}}",
+      "model_name": "test_model",
+      "schedule_name": "test-schedule",
+      "workflow_name": "predictions_batch"
+    }
+  },
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT * FROM nonexistent_table WHERE invalid_syntax =",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data",
+  "input_filters": {
+    "EMPTY_DATA": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_transform_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_predict_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "path_priority": ["STOP", "CONTINUE", "REPEAT"],
+  "opc_output_config": {},
+  "pi_web_api_output_config": {},
+  "save_transform": true,
+  "prediction_store_policy": "lts:1",
+  "model_config": {
+    "target": "sensor_1",
+    "retention_minutes": 0
+  }
+}

e2e/scenario_inputs/minimal_retrain_base.json

@@ -0,0 +1,12 @@
+{
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT timestamp, variable, value, created_at FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "log_retrain",
+  "datetime_columns": ["timestamp", "created_at"],
+  "model_config": {
+    "target": "sensor_1"
+  }
+}

e2e/scenario_inputs/minio_offload_load_query.json

@@ -0,0 +1,11 @@
+{
+  "metadata": {
+    "schedule_name": "test-schedule",
+    "model_name": "test_model",
+    "model_id": "{{MODEL_ID}}",
+    "workflow_name": "predictions_batch"
+  },
+  "query": "SELECT timestamp, variable, value, created_at FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "model_name": "test_model",
+  "datetime_columns": ["timestamp", "created_at"]
+}

e2e/scenario_inputs/minio_offload_workflow.json

@@ -0,0 +1,37 @@
+{
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT timestamp, variable, value, created_at FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data",
+  "input_filters": {
+    "EMPTY_DATA": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_transform_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_predict_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "path_priority": ["STOP", "CONTINUE", "REPEAT"],
+  "opc_output_config": {},
+  "pi_web_api_output_config": {},
+  "save_transform": false,
+  "prediction_store_policy": "lts:1",
+  "model_config": {
+    "retention_minutes": 0,
+    "target": "sensor_1"
+  },
+  "datetime_columns": ["timestamp", "created_at"]
+}

e2e/scenario_inputs/prediction_process_base.json

@@ -0,0 +1,39 @@
+{
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "query": "SELECT timestamp, variable, value, created_at FROM sientia_data.laborious_data WHERE model_id = {{MODEL_ID}}",
+  "schema": "sientia_data",
+  "table_name": "predictions",
+  "transform_table_name": "transformed_data",
+  "input_filters": {
+    "SPECIFIC_VARIABLES_NULL_VALUES": {
+      "POLICY": "CONTINUE",
+      "CONFIG": {
+        "variables": ["sensor_1"]
+      }
+    }
+  },
+  "mlflow_transform_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "mlflow_predict_filters": {
+    "API_ERROR": {
+      "POLICY": "STOP",
+      "CONFIG": {}
+    }
+  },
+  "path_priority": ["STOP", "CONTINUE", "REPEAT"],
+  "opc_output_config": {},
+  "pi_web_api_output_config": {},
+  "save_transform": true,
+  "prediction_store_policy": "lts:1",
+  "model_config": {
+    "retention_minutes": 0,
+    "target": "sensor_1"
+  },
+  "datetime_columns": ["timestamp", "created_at"]
+}

e2e/scenario_inputs/simple_metrics_base.json

@@ -0,0 +1,14 @@
+{
+  "schedule_name": "test-schedule",
+  "model_name": "test_model",
+  "model_id": "{{MODEL_ID}}",
+  "schema": "sientia_data",
+  "predictions_table_name": "predictions",
+  "data_table_name": "laborious_data",
+  "target_table_name": "simple_metrics",
+  "interval_minutes": 60,
+  "metrics": ["rmse", "mse", "mae", "r2"],
+  "model_config": {
+    "target": "sensor_target"
+  }
+}

e2e/scenarios.md

@@ -0,0 +1,455 @@
+# E2E Scenario Documentation - Predictions Batch
+
+This document describes the end-to-end scenarios for `predictions_batch` and its child workflows:
+`prediction_process` and `format_and_export_prediction`.
+
+It is a functional reference of scenario behavior, inputs, and expected outcomes.
+
+## Execution Context
+
+- Tests run under `e2e/` and are marked with `@pytest.mark.integration`.
+- PostgreSQL and MinIO are provisioned with testcontainers.
+- `test_minio_offload.py` uses real MinIO I/O; other scenario suites may use stubs/mocks for optional outputs.
+- Real OPC UA scenarios use `@pytest.mark.opc` and an in-process asyncua server (`e2e/test_opc_real_server.py`).
+
+### Local validation
+
+Use the existing project virtualenv and the shared `validate` script for unit/quality gates; run E2E separately (Docker required).
+
+```bash
+source ./venv/bin/activate
+
+# Auto-fix + static checks (no pytest)
+validate --fix --project-name=laborious
+
+# Full unit + quality gate
+validate --project-name=laborious
+
+# E2E (integration)
+pytest e2e/ --override-ini testpaths=e2e -m integration
+
+# E2E (real OPC server only)
+pytest e2e/test_opc_real_server.py --override-ini testpaths=e2e -m opc
+```
+
+---
+
+## 1. Main Workflow Scenarios
+Source: `e2e/test_predictions_batch_main_workflow.py`
+
+### 1.1.1 Happy Path - Complete Success
+**Summary**: Full workflow succeeds with valid query and default gate behavior.
+
+**Description**:
+- Query returns rows for a model.
+- `prediction_process` runs transform and predict paths.
+- Final prediction and transformed data are persisted.
+
+**Expected Outcome**:
+- Exactly one prediction row is created.
+- Transform rows are created.
+- Confidence/status/comments are success values.
+
+### 1.2.1 SQL Query Execution Error
+**Summary**: Invalid SQL leads to no persisted prediction.
+
+**Description**:
+- Input query is invalid.
+- Load step fails and workflow follows error/short-circuit path.
+
+**Expected Outcome**:
+- No prediction rows for the model.
+- Workflow does not require retry-loop assumptions in assertions.
+
+### 1.2.2 Missing Required Parameters
+**Summary**: Missing required fields prevent workflow completion path.
+
+**Description**:
+- Required input key (e.g. `query`) is omitted.
+- Workflow fails to produce actionable input for child flow.
+
+**Expected Outcome**:
+- No prediction rows are persisted.
+- Workflow handle may require explicit terminate in E2E harness.
+
+### 1.2.3 Invalid Datetime Column Specification (de-prioritized)
+**Summary**: Legacy invalid datetime-column case is retained only as low-priority legacy coverage.
+
+**Description**:
+- `datetime_columns` references non-existing columns.
+- Behavior may vary by query shape and parser fallback.
+
+**Expected Outcome**:
+- No predictions persisted in the covered legacy assertion path.
+- Scenario is not considered primary behavior coverage.
+
+---
+
+## 2. Prediction Process Scenarios
+Source: `e2e/test_predictions_batch_prediction_process.py`
+
+### 2.1 Input Gate Path Decisions
+
+#### 2.1.1 CONTINUE
+**Summary**: Input filter flags quality issue but allows continuation via default path.
+
+**Description**:
+- Input gate returns `CONTINUE`.
+- MLFlow transform/predict are skipped.
+- Export path persists default-style prediction with warning context.
+
+#### 2.1.2 STOP
+**Summary**: Input filter blocks processing.
+
+**Description**:
+- Input gate returns `STOP`.
+- Workflow exits without export.
+
+#### 2.1.3 REPEAT with history
+**Summary**: Prior prediction is reused.
+
+**Description**:
+- Input gate returns `REPEAT`.
+- `repeat_last_prediction` path is executed using existing historical row.
+
+#### 2.1.4 REPEAT without history
+**Summary**: Repeat requested but no previous prediction exists.
+
+**Description**:
+- Input gate returns `REPEAT`.
+- No prior row is available to duplicate.
+
+**Expected Outcome**:
+- No new prediction rows are created for the model.
+
+### 2.2 Transform Gate Decisions
+
+#### 2.2.1 CONTINUE on transform response error
+**Summary**: Transform response is degraded, but workflow continues.
+
+#### 2.2.2 STOP on transform response error
+**Summary**: Transform response error blocks downstream processing.
+
+#### 2.2.3 REPEAT on transform response error
+**Summary**: Transform response error triggers repeat-last-prediction path.
+
+#### 2.2.4 STOP on transform content NaN
+**Summary**: Content gate (`NAN_VALUES`) blocks on all-NaN transform payload.
+
+### 2.3 Predict Gate Decisions
+
+#### 2.3.1 CONTINUE on predict response error
+**Summary**: Predict response degraded; workflow exports with degraded metadata.
+
+#### 2.3.2 STOP on predict response error
+**Summary**: Predict response error blocks export.
+
+#### 2.3.3 REPEAT on predict response error
+**Summary**: Predict response error routes to repeat-last-prediction.
+
+### 2.4.1 Priority Conflict Resolution
+**Summary**: Deterministic selection when multiple filters produce different flags.
+
+**Description**:
+- Multiple filters may produce `STOP`, `CONTINUE`, and/or `REPEAT`.
+- `path_priority` defines precedence.
+
+**Expected Outcome**:
+- Highest-priority flag is applied consistently.
+- Executed branch matches configured priority ordering.
+
+---
+
+## 3. Format and Export Scenarios
+Source: `e2e/test_predictions_batch_format_export.py`
+
+### 3.1 Output Combination Scenarios
+
+#### 3.1.1 Default prediction export
+**Summary**: Non-`None` path flag uses `format_default_prediction`.
+
+**Description**:
+- Default prediction is generated.
+- Transform export is skipped.
+- Optional outputs (PI/OPC) still execute when configured.
+
+#### 3.1.2 OPC only
+**Summary**: Postgres + OPC writes, PI Web API disabled.
+
+#### 3.1.3 PI Web API only
+**Summary**: Postgres + PI writes, OPC disabled.
+
+#### 3.1.4 Postgres only
+**Summary**: Both optional outputs disabled; only Postgres persistence and metrics.
+
+#### 3.1.5 No transformed data export
+**Summary**: Prediction is persisted; transformed table is not written.
+
+### 3.2 Degraded-but-successful Completion
+
+#### 3.2.1 PI Web API write error
+**Summary**: PI write failure does not fail workflow.
+
+**Expected Outcome**:
+- Workflow completes.
+- Prediction persisted with degraded confidence/comments (PI error semantics).
+
+#### 3.2.2 OPC write error
+**Summary**: OPC write failure does not fail workflow.
+
+**Expected Outcome**:
+- Workflow completes.
+- Prediction persisted with OPC degraded confidence/comments.
+
+#### 3.2.3 PI Web API partial write error
+**Summary**: Partial PI acknowledgement is treated as degraded success.
+
+**Expected Outcome**:
+- Workflow completes.
+- Prediction persisted with PI error confidence and descriptive comment.
+
+#### 3.2.4 OPC session / channel error (confidence 14)
+**Summary**: Tier-1 `BadSessionIdInvalid` (or equivalent session error) degrades the prediction without failing the workflow.
+
+**Sources**:
+- Mock: `e2e/test_predictions_batch_format_export.py::test_scenario_3_2_4_opc_session_bad_mock`
+- Real server: `e2e/test_opc_real_server.py::test_scenario_3_2_4_opc_session_bad_real_server` (`@pytest.mark.opc`)
+
+**Expected Outcome**:
+- Workflow completes.
+- `prediction_confidence` is 14.
+- Comments contain `OPC UA session/channel error: BadSessionIdInvalid`.
+
+#### 3.2.5 OPC write blocked during reconnect (confidence 14)
+**Summary**: While reconnect holds the repository connection lock, writes fail fast with `reconnect_in_progress`.
+
+**Sources**:
+- Mock: `e2e/test_predictions_batch_format_export.py::test_scenario_3_2_5_opc_reconnect_in_progress_mock`
+- Real server: `e2e/test_opc_real_server.py::test_scenario_3_2_5_opc_write_blocked_during_reconnect_real_server` (`@pytest.mark.opc`)
+
+**Expected Outcome**:
+- Workflow completes.
+- `prediction_confidence` is 14.
+- Comments contain `OPC UA reconnect in progress`.
+
+### 3.3.1 Combined Optional Outputs (PI + OPC)
+**Summary**: Both external output channels are enabled together.
+
+**Description**:
+- PI Web API and OPC configs are both present.
+- Output mutation order matters for final persisted payload.
+
+**Expected Outcome**:
+- PI write executes before OPC write in workflow sequence.
+- Final Postgres payload reflects any confidence/comment updates.
+- OPC metrics are emitted when tag writes return response times.
+
+---
+
+## 4. MinIO Offload Scenarios
+Source: `e2e/test_minio_offload.py`
+
+### 4.1.1 Forced offload to MinIO
+**Summary**: Very low threshold forces parquet upload.
+
+**Description**:
+- Payload is offloaded (`object_key` present, inline data absent/empty).
+- Object is present in MinIO under `prediction_datasets/...`.
+- Retrieval reconstructs the dataframe.
+
+### 4.1.2 Full workflow with offloaded load payload
+**Summary**: Offload path works during full `predictions_batch` execution.
+
+**Expected Outcome**:
+- Workflow completes.
+- Prediction row is persisted.
+
+### 4.2.1 Inline payload below threshold
+**Summary**: Data remains inline when threshold is not exceeded.
+
+**Expected Outcome**:
+- Payload stores inline `data`.
+- `object_key` is `None`.
+- Downstream persistence behavior matches offload scenario semantics.
+
+---
+
+## 5. Drift Workflow Scenarios
+Source: `e2e/test_drift.py`
+
+The drift suite drives the **real** `sientia_model.analytics.drift_analysis.DriftAnalysis`
+analyzer (no stubs / mocks). Each scenario exercises the full pipeline:
+
+```
+laborious_data (Postgres) -> load_custom_query
+                          -> calculate_drift (DriftAnalysis univariate + multivariate)
+                          -> export_data_to_postgres (sientia_data.drift_metrics)
+```
+
+The `mlflow_repository_stub` provides the reference-data CSV via
+`download_artifacts`, and tests assert postgres rows in
+`sientia_data.drift_metrics` against this canonical schema:
+
+`id, model_id, feature, method, value, alert, chunk_index, chunk_start_date, chunk_end_date, accurate, timestamp, created_at`.
+
+Tests assert behavioral / structural properties (column presence, NOT NULL
+constraints, business-key invariants like uniform `timestamp` and stamped
+`model_id`) rather than exact numeric drift scores, since those depend on
+the real analyzer implementation and the synthetic data fed in.
+
+### 5.1 Happy paths
+
+#### D.1.1 Full pipeline persists all columns with reference data
+**Summary**: 10 minutes of target data are inserted; a 10-row reference CSV
+is configured via the MLflow stub. The `DriftAnalysis` runs end-to-end.
+
+**Expected Outcome**:
+- One row per `(chunk_index, feature, method)` plus a `multivariate` block
+  per chunk is persisted.
+- Every column in the DDL is populated; `feature` is the only nullable column
+  per the new schema.
+- `accurate=True` for every row (reference path).
+- All three default univariate methods reach the analyzer.
+- `model_id` is stamped as `text` and uniform across rows.
+- `timestamp` equals `max(target_data.timestamp)` and is uniform across rows.
+- `chunk_start_date` / `chunk_end_date` are persisted as ISO text and ordered.
+- `p_value` is dropped before persistence.
+
+#### D.1.2 30% fallback when reference data is unavailable
+**Summary**: MLflow alias resolution is forced to fail so
+`get_reference_data` returns `None`; `calculate_drift` falls back to the
+first 30% of target rows as reference.
+
+**Expected Outcome**:
+- All persisted rows carry `accurate=False`.
+- A `MODEL_METRICS_REFERENCE_DATA_WARNING` notification is emitted to MongoDB.
+
+### 5.2 Failure paths
+
+#### D.3.1 Empty target data short-circuits the workflow
+**Summary**: `load_custom_query` returns no rows.
+
+**Expected Outcome**:
+- The workflow returns early and writes nothing to `sientia_data.drift_metrics`.
+
+### 5.3 Configuration paths
+
+#### D.4.2 Invalid `chunk_period` raises ValueError
+**Summary**: Anything other than `min` / `s` is rejected by `calculate_drift`.
+
+**Expected Outcome**:
+- The workflow surfaces the `ValueError` ("Invalid chunk period: ...").
+- No rows are persisted.
+
+#### D.4.3 `chunk_period='s'` preserves seconds in `chunk_start_date`
+**Summary**: Target data spans two minutes with samples at second-30
+boundaries; the activity is configured with `chunk_period='s'`.
+
+**Expected Outcome**:
+- At least one persisted `chunk_start_date` carries `seconds=30`, proving
+  that the analyzer chunked at sub-minute granularity and the ISO-text
+  serialization preserved the boundary.
+
+---
+
+## 6. Simple Metrics Workflow Scenarios
+Source: `e2e/test_simple_metrics.py`
+
+Validates `sientia_data.simple_metrics` columns:
+`id, model_id, metric, value, timestamp, data_size, interval_minutes, created_at`.
+Note: ``timestamp`` is now nullable per the new DDL and ``model_id`` is ``text``.
+
+### 6.1 Happy paths
+
+#### S.1.1 rmse/mse/mae/r2 happy path
+**Summary**: Prediction/target pairs are inserted; the activity computes all
+four metrics with closed-form expected values.
+
+**Expected Outcome**:
+- One row per metric is persisted; all columns populated.
+- `data_size` matches the joined row count and `interval_minutes=60`.
+
+#### S.1.2 Subset metrics
+**Summary**: Requesting `metrics=['rmse']` writes only the rmse row.
+
+### 6.2 Edge cases
+
+#### S.2.1 Zero-variance target returns r2=0
+**Summary**: When all targets are equal, `ss_tot=0`; the activity must guard
+against division by zero and return `r2=0`.
+
+### 6.3 Failure paths
+
+#### S.3.1 No overlapping data short-circuits persistence
+**Summary**: With no `laborious_data` rows for the configured target variable
+the workflow exits before `calculate_simple_metrics` and writes nothing.
+
+---
+
+## 7. Minimal Retrain Workflow Scenarios
+Source: `e2e/test_minimal_retrain.py`
+
+The MLflow registry is fully mocked (no real artifacts in test container).
+Validates `sientia_data.log_retrain` columns:
+`mlflow_experiment_id, mlflow_run_id, model_id, model_name, status, timestamp, version`.
+Note: the new DDL drops the legacy ``id`` and ``created_at`` columns,
+``mlflow_experiment_id`` is now ``int8`` and ``model_id`` is ``text``.
+
+### 7.1 Happy path
+
+#### MR.1.1 Successful retrain + promotion
+**Summary**: Training data loads via MinIO offload, `wrapper.retrain` succeeds,
+the new version is promoted to the `production` alias.
+
+**Expected Outcome**:
+- Report row has success status, `version='7'`, `mlflow_run_id='retrain-run-id'`,
+  `mlflow_experiment_id=4242` (`int8`).
+- `mlflow.log_artifact` is called with the input CSV.
+- `promote_to_alias` is called once with the resolved version and alias.
+
+### 7.2 Failure paths
+
+#### MR.2.1 Wrapper retrain raises
+**Summary**: `wrapper.retrain` raises `RuntimeError`. The activity returns
+`success=False`, `update_production_model` is NOT invoked.
+
+**Expected Outcome**:
+- Report row carries the error message and `version`/`mlflow_*` columns are NULL.
+
+#### MR.2.2 Missing `model_config.target`
+**Summary**: Empty model config short-circuits before any MLflow call.
+
+**Expected Outcome**:
+- Report row carries the explicit guard message.
+- `get_cached_model` is never invoked.
+
+#### MR.3.1 No training data
+**Summary**: The training query returns no rows; the workflow does not
+persist any report row. The current code raises plain `ValueError` from the
+workflow function, which Temporal treats as a workflow-task failure (see
+`CODE_ISSUES.md` issue MR-1).
+
+---
+
+## Input Contract Reference
+
+Common scenario input fields:
+- `schedule_name`
+- `model_name`
+- `model_id`
+- `query`
+- `schema`
+- `table_name`
+- `transform_table_name`
+- `input_filters`
+- `mlflow_transform_filters`
+- `mlflow_predict_filters`
+- `path_priority` (default order: `STOP`, `CONTINUE`, `REPEAT`)
+- `save_transform`
+- `prediction_store_policy`
+- `model_config.target`
+- `datetime_columns` (when query returns temporal fields)
+
+Optional outputs:
+- `opc_output_config`
+- `pi_web_api_output_config`

e2e/test_child_workflows_e2e.py

@@ -0,0 +1,74 @@
+"""
+Direct E2E execution of child workflows (smaller surface than PredictionsBatch).
+"""
+
+from decimal import Decimal
+
+import pytest
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import make_workflow_id, start_and_await_workflow
+from laborious.workflows.sub_workflows.format_and_export_prediction import FormatAndExportPrediction
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_format_and_export_prediction_default_path_e2e(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    postgres_engine,
+):
+    """
+    Run FormatAndExportPrediction with path_flag set (format_default_prediction path).
+    """
+    client = temporal_test_env.client
+    model_id = 401
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.predictions WHERE model_id = {model_id}'))
+
+    metadata = {
+        'metadata': {
+            'model_id': model_id,
+            'model_name': 'test_model',
+            'schedule_name': 'test-schedule',
+            'workflow_name': 'subworkflow.format_and_export_prediction',
+        }
+    }
+    input_data = {
+        'metadata': metadata,
+        'path_flag': 'CONTINUE',
+        'data': {'last_timestamp': '2024-01-01 12:00:00+00:00'},
+        'prediction_confidence': 2,
+        'timestamp': '2024-01-01 12:00:00+00:00',
+        'model_id': model_id,
+        'model_name': 'test_model',
+        'schema': 'sientia_data',
+        'table_name': 'predictions',
+        'transform_table_name': 'transformed_data',
+        'comment': 'e2e child workflow default path',
+        'opc_output_config': {},
+        'pi_web_api_output_config': {},
+        'prediction_store_policy': 'lts:1',
+    }
+
+    await start_and_await_workflow(
+        client,
+        FormatAndExportPrediction.run,
+        input_data,
+        make_workflow_id('e2e-format-export-child'),
+    )
+
+    with postgres_engine.connect() as conn:
+        row = conn.execute(
+            text(
+                f'SELECT prediction, prediction_confidence, prediction_status, comments '
+                f'FROM sientia_data.predictions WHERE model_id = {model_id}'
+            )
+        ).fetchone()
+    assert row is not None
+    assert row[0] == 0
+    assert row[1] == Decimal(2)
+    assert row[2] == 'Bad'
+    assert row[3] == 'e2e child workflow default path'

e2e/test_drift.py

@@ -0,0 +1,600 @@
+"""
+End-to-end tests for the Drift workflow.
+
+The drift suite drives the **real** ``sientia_model.analytics.drift_analysis.DriftAnalysis``
+analyzer (no mocking). Each scenario exercises the full pipeline:
+
+    laborious_data (Postgres)
+        -> load_custom_query
+        -> calculate_drift (DriftAnalysis univariate + multivariate)
+        -> export_data_to_postgres (sientia_data.drift_metrics)
+
+Coverage focus:
+
+- Happy path persists every column required by ``sientia_data.drift_metrics``
+  with a valid reference dataset downloaded from MLflow.
+- 30% fallback path activates when the MLflow reference is unavailable and
+  emits the ``MODEL_METRICS_REFERENCE_DATA_WARNING`` notification.
+- Empty target data short-circuits the workflow without persisting anything.
+- Invalid ``chunk_period`` is rejected by ``calculate_drift``.
+- ``chunk_period='s'`` preserves second-level precision in
+  ``chunk_start_date``.
+
+Tests assert behavioral / structural properties (column presence, NOT NULL
+constraints, business-key invariants) rather than exact numeric values, since
+those depend on the real analyzer implementation and synthetic data.
+"""
+
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pandas as pd
+import pytest
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import (
+    insert_target_data_for_drift,
+    load_scenario_input,
+    make_workflow_id,
+    start_and_await_workflow,
+)
+from laborious.activities.activities import Activities
+from laborious.activities.model_metrics import ModelMetrics
+from laborious.workflows.drift import Drift
+from sientia_model.analytics.drift_analysis import DriftAnalysis
+
+# Drift columns persisted on every row in ``sientia_data.drift_metrics`` —
+# mirrors the production DDL.
+EXPECTED_DRIFT_COLUMNS = [
+    'id',
+    'model_id',
+    'feature',
+    'method',
+    'value',
+    'alert',
+    'chunk_index',
+    'chunk_start_date',
+    'chunk_end_date',
+    'accurate',
+    'timestamp',
+    'created_at',
+]
+
+# Columns the DDL marks as NOT NULL. ``feature`` and ``timestamp`` are
+# nullable in the production schema (multivariate rows do not bind to a
+# single feature; ``timestamp`` is allowed to be empty when upstream data has
+# no usable instant).
+NON_NULL_DRIFT_COLUMNS = {
+    'id',
+    'model_id',
+    'method',
+    'value',
+    'alert',
+    'chunk_index',
+    'chunk_start_date',
+    'chunk_end_date',
+    'accurate',
+    'created_at',
+}
+
+DEFAULT_DRIFT_METHODS = ['kolmogorov_smirnov', 'jensen_shannon', 'wasserstein']
+
+
+def _chunk_dataframe_skip_empty_groups(
+    self: DriftAnalysis,
+    df: pd.DataFrame,
+    timestamp_col: str,
+    chunk_period: str,
+) -> list[tuple[int, pd.DataFrame]]:
+    """
+    Same as ``DriftAnalysis._chunk_dataframe`` but omit empty time buckets.
+
+    ``pd.Grouper(freq='s')`` yields every second between min and max timestamp;
+    empty buckets still appear in the groupby iterator and produce invalid
+    drift rows (e.g. NaT timestamps) that ``calculate_drift`` later filters out
+    entirely. Production fix belongs in ``sientia_model``; this shim keeps the
+    e2e honest about second-level chunk boundaries with sparse samples.
+    """
+    grouped = df.groupby(pd.Grouper(key=timestamp_col, freq=chunk_period), dropna=True)
+    chunks: list[tuple[int, pd.DataFrame]] = []
+    idx = 0
+    for _, chunk in grouped:
+        if chunk.empty:
+            continue
+        chunks.append((idx, chunk.copy()))
+        idx += 1
+    return chunks
+
+
+def _drift_input(model_id: int, **overrides) -> dict:
+    """Load the base drift scenario JSON and apply ad-hoc overrides."""
+    input_data = load_scenario_input('drift_base.json', model_id=model_id)
+    input_data.update(overrides)
+    return input_data
+
+
+def _recent_minute_timestamps(count: int, offset_minutes: int = 6) -> list[str]:
+    """
+    Build ``count`` consecutive UTC minute timestamps placed in the recent past.
+
+    The Drift workflow filters target rows with ``timestamp > NOW() - INTERVAL``,
+    so timestamps must be recent for tests to retrieve any data. Snapping to
+    minute precision keeps the helper deterministic regardless of clock skew.
+
+    Args:
+     - count (int): How many consecutive minute timestamps to generate.
+     - offset_minutes (int): Minutes ago for the EARLIEST generated timestamp.
+
+    Return:
+     list[str]: ISO strings with ``+0000`` offset, one per minute.
+    """
+    base = datetime.now(timezone.utc).replace(second=0, microsecond=0) - timedelta(
+        minutes=offset_minutes
+    )
+    return [
+        (base + timedelta(minutes=i)).strftime('%Y-%m-%d %H:%M:%S%z') for i in range(count)
+    ]
+
+
+def _configure_reference_csv(mlflow_repository_stub, reference_rows: pd.DataFrame) -> None:
+    """
+    Wire ``mlflow_repository_stub`` so ``get_reference_data`` returns
+    ``reference_rows`` by writing them to ``dst_path/retrain_input.csv``.
+
+    Args:
+     - mlflow_repository_stub: External MLflow repository fixture.
+     - reference_rows (pd.DataFrame): Rows to expose as the production reference.
+    """
+
+    def _download(run_id: str, artifact_path: str, dst_path: str, metadata=None):
+        target = Path(dst_path) / artifact_path
+        target.parent.mkdir(parents=True, exist_ok=True)
+        reference_rows.to_csv(target, index=False)
+
+    mlflow_repository_stub._client.get_model_version_by_alias.return_value = MagicMock(
+        run_id='fake-reference-run'
+    )
+    file_info = MagicMock()
+    file_info.path = 'retrain_input.csv'
+    mlflow_repository_stub._client.list_artifacts.return_value = [file_info]
+    mlflow_repository_stub.download_artifacts.side_effect = _download
+
+
+def _force_reference_unavailable(mlflow_repository_stub) -> None:
+    """Make ``get_reference_data`` return ``None`` by failing alias resolution."""
+    mlflow_repository_stub._client.get_model_version_by_alias.side_effect = Exception(
+        'no production alias registered'
+    )
+
+
+def _select_drift_rows(postgres_engine, model_id: int) -> list[dict]:
+    """Read every persisted drift row for ``model_id`` ordered by chunk/feature/method."""
+    with postgres_engine.connect() as conn:
+        rows = (
+            conn.execute(
+                text(
+                    'SELECT * FROM sientia_data.drift_metrics '
+                    'WHERE model_id = :m '
+                    'ORDER BY chunk_index, feature, method'
+                ),
+                {'m': str(model_id)},
+            )
+            .mappings()
+            .all()
+        )
+    return [dict(row) for row in rows]
+
+
+def _assert_required_columns_populated(rows: list[dict]) -> None:
+    """Validate column presence and NOT NULL constraints on every row."""
+    assert rows, 'expected at least one drift row to be persisted'
+    seen_columns = set(rows[0].keys())
+    for column in EXPECTED_DRIFT_COLUMNS:
+        assert column in seen_columns, f'Missing drift column in postgres: {column}'
+    for row in rows:
+        for column in NON_NULL_DRIFT_COLUMNS:
+            assert row[column] is not None, f"Column '{column}' is NULL in {row}"
+        assert 'p_value' not in row, 'p_value must not be persisted to drift_metrics'
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_drift_happy_path_persists_all_columns_with_reference_data(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_drift: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario D.1.1: Happy path with reference data downloaded from MLflow.
+
+    Drives the full pipeline against the real ``DriftAnalysis``. Asserts:
+
+    - One row is persisted per ``(chunk_index, feature, method)`` combination
+      plus the multivariate row block, with every column required by
+      ``sientia_data.drift_metrics`` populated.
+    - The three default univariate methods are forwarded to the analyzer.
+    - ``model_id`` and ``timestamp`` are stamped by the activity (not by the
+      analyzer); ``timestamp`` equals ``max(target_data.timestamp)`` and is
+      identical on every persisted row.
+    - ``chunk_start_date`` / ``chunk_end_date`` are persisted as ISO text so
+      the analyzer's nanosecond-precision boundaries survive the ``text``
+      column type.
+    - ``accurate=True`` because the reference dataset was available.
+    """
+    client = temporal_test_env.client
+    model_id = 411
+
+    target_timestamps = _recent_minute_timestamps(count=10)
+    insert_target_data_for_drift(
+        postgres_engine,
+        model_id=model_id,
+        timestamps=target_timestamps,
+        variables_values={
+            'sensor_1': [10.0 + i * 0.1 for i in range(10)],
+            'sensor_2': [20.0 + i * 0.5 for i in range(10)],
+        },
+    )
+
+    reference_df = pd.DataFrame(
+        {
+            'timestamp': [
+                f'2023-12-31 11:{minute:02d}:00+00:00' for minute in range(10)
+            ],
+            'sensor_1': [9.0 + i * 0.05 for i in range(10)],
+            'sensor_2': [18.0 + i * 0.25 for i in range(10)],
+        }
+    )
+    _configure_reference_csv(mlflow_repository_stub, reference_df)
+
+    input_data = _drift_input(model_id)
+    await start_and_await_workflow(
+        client, Drift.run, input_data, make_workflow_id('test-drift-happy-path')
+    )
+
+    rows = _select_drift_rows(postgres_engine, model_id)
+
+    exported_csv_path = '/tmp/test_drift_happy_path_exported.csv'
+    pd.DataFrame(rows).to_csv(exported_csv_path, index=False)
+    print(
+        f'\n[test_drift_happy_path] Exported drift dataframe '
+        f'({len(rows)} rows) -> {exported_csv_path}'
+    )
+
+    _assert_required_columns_populated(rows)
+
+    # The activity drops the target column from the feature list, so only
+    # ``sensor_2`` participates in univariate analysis (``sensor_1`` is the
+    # configured target). Multivariate produces one row per chunk regardless.
+    univariate_rows = [r for r in rows if r['feature'] != 'multivariate']
+    multivariate_rows = [r for r in rows if r['feature'] == 'multivariate']
+    assert univariate_rows, 'expected univariate drift rows for non-target features'
+    assert multivariate_rows, 'expected one multivariate drift row per chunk'
+
+    # All three default methods must reach the analyzer.
+    assert {r['method'] for r in univariate_rows} == set(DEFAULT_DRIFT_METHODS)
+    assert all(r['method'] == 'multivariate' for r in multivariate_rows)
+    assert {r['feature'] for r in univariate_rows} == {'sensor_2'}
+
+    # ``timestamp`` is stamped uniformly with ``max(target_data.timestamp)``.
+    expected_timestamp = pd.to_datetime(max(target_timestamps), utc=True)
+    persisted_timestamps = {pd.to_datetime(r['timestamp'], utc=True) for r in rows}
+    assert len(persisted_timestamps) == 1, (
+        'timestamp must be uniform across all drift rows '
+        f'(got {len(persisted_timestamps)} distinct values)'
+    )
+    assert pd.Timestamp(persisted_timestamps.pop()) == expected_timestamp, (
+        'timestamp must equal max(target_data.timestamp)'
+    )
+
+    # ``model_id`` is stamped by ``calculate_drift`` (not produced by the analyzer).
+    assert all(r['model_id'] == str(model_id) for r in rows), (
+        'model_id must be stamped on every drift row'
+    )
+
+    # Reference path → accurate=True.
+    assert all(r['accurate'] is True for r in rows), (
+        'reference path should mark all rows as accurate'
+    )
+
+    # ISO text serialization preserves ordering between start/end of each chunk.
+    for row in rows:
+        assert 'T' in row['chunk_start_date'], (
+            f"chunk_start_date should be ISO text, got {row['chunk_start_date']!r}"
+        )
+        assert 'T' in row['chunk_end_date'], (
+            f"chunk_end_date should be ISO text, got {row['chunk_end_date']!r}"
+        )
+        assert row['chunk_start_date'] <= row['chunk_end_date'], (
+            f'chunk_start_date must precede chunk_end_date '
+            f"(start={row['chunk_start_date']}, end={row['chunk_end_date']})"
+        )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_drift_uses_30pct_fallback_when_reference_unavailable(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_drift: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+    notification_inserts,
+):
+    """
+    Scenario D.1.2: ``get_reference_data`` returns ``None`` (production alias
+    missing), so ``calculate_drift`` falls back to using the first 30% of
+    target rows as reference. Persisted rows must report ``accurate=False``
+    and a ``MODEL_METRICS_REFERENCE_DATA_WARNING`` notification must be
+    emitted to mongo.
+    """
+    client = temporal_test_env.client
+    model_id = 412
+
+    target_timestamps = _recent_minute_timestamps(count=10)
+    insert_target_data_for_drift(
+        postgres_engine,
+        model_id=model_id,
+        timestamps=target_timestamps,
+        variables_values={
+            'sensor_1': [10.0 + i * 0.1 for i in range(10)],
+            'sensor_2': [20.0 + i * 0.5 for i in range(10)],
+        },
+    )
+
+    _force_reference_unavailable(mlflow_repository_stub)
+
+    input_data = _drift_input(model_id)
+    await start_and_await_workflow(
+        client, Drift.run, input_data, make_workflow_id('test-drift-fallback')
+    )
+
+    rows = _select_drift_rows(postgres_engine, model_id)
+    _assert_required_columns_populated(rows)
+
+    assert all(r['accurate'] is False for r in rows), (
+        'fallback path must mark all rows as inaccurate'
+    )
+
+    fallback_warnings = [
+        call
+        for call in notification_inserts.call_args_list
+        if call.args
+        and isinstance(call.args[0], dict)
+        and call.args[0].get('notification_id') == 'MODEL_METRICS_REFERENCE_DATA_WARNING'
+    ]
+    assert len(fallback_warnings) >= 1, 'expected reference fallback warning notification'
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_drift_empty_target_data_short_circuits_workflow(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_drift: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario D.3.1: When ``load_custom_query`` returns no rows the workflow
+    must return early without invoking the analyzer or writing any drift rows.
+    """
+    client = temporal_test_env.client
+    model_id = 431
+
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+
+    _force_reference_unavailable(mlflow_repository_stub)
+
+    input_data = _drift_input(model_id)
+    await start_and_await_workflow(
+        client, Drift.run, input_data, make_workflow_id('test-drift-empty-target')
+    )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.drift_metrics WHERE model_id = :m'),
+            {'m': str(model_id)},
+        ).scalar()
+    assert count == 0
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_drift_invalid_chunk_period_raises_value_error(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_drift: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario D.4.2: ``calculate_drift`` validates ``chunk_period`` and rejects
+    anything other than ``min`` / ``s``. The workflow must surface the
+    ``ValueError`` and persist nothing.
+    """
+    client = temporal_test_env.client
+    model_id = 442
+
+    target_timestamps = _recent_minute_timestamps(count=5)
+    insert_target_data_for_drift(
+        postgres_engine,
+        model_id=model_id,
+        timestamps=target_timestamps,
+        variables_values={
+            'sensor_1': [1.0, 2.0, 3.0, 4.0, 5.0],
+            'sensor_2': [10.0, 20.0, 30.0, 40.0, 50.0],
+        },
+    )
+    _force_reference_unavailable(mlflow_repository_stub)
+
+    input_data = _drift_input(model_id, chunk_period='hour')
+
+    with pytest.raises(Exception) as excinfo:
+        await start_and_await_workflow(
+            client,
+            Drift.run,
+            input_data,
+            make_workflow_id('test-drift-bad-chunk-period'),
+        )
+
+    # Temporal wraps the activity ValueError in WorkflowFailureError; the
+    # message may live on ``.message`` or ``str(exc)`` depending on the SDK
+    # error class, so walk the cause chain looking for the guard text.
+    cause_descriptions = []
+    current: BaseException | None = excinfo.value
+    while current is not None:
+        cause_descriptions.append(
+            getattr(current, 'message', None) or str(current) or repr(current)
+        )
+        current = current.__cause__
+    assert any('Invalid chunk period' in msg for msg in cause_descriptions), (
+        f'Expected ValueError about chunk period in chain, got: {cause_descriptions}'
+    )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.drift_metrics WHERE model_id = :m'),
+            {'m': str(model_id)},
+        ).scalar()
+    assert count == 0
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_drift_empty_merge_skips_export_without_insufficient_notification(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_drift: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario D.4.3a: When the analyzer returns an empty merged frame (no metric rows),
+    ``calculate_drift`` yields ``[]``; the workflow skips export. Real insufficient-data
+    cases are signaled by ``DriftInsufficientDataError`` inside ``sientia_model``, not by
+    empty output alone.
+    """
+    client = temporal_test_env.client
+    model_id = 444
+
+    target_timestamps = _recent_minute_timestamps(count=5)
+    insert_target_data_for_drift(
+        postgres_engine,
+        model_id=model_id,
+        timestamps=target_timestamps,
+        variables_values={
+            'sensor_1': [1.0 + i * 0.1 for i in range(5)],
+            'sensor_2': [10.0 + i * 0.5 for i in range(5)],
+        },
+    )
+    _force_reference_unavailable(mlflow_repository_stub)
+
+    input_data = _drift_input(model_id, chunk_period='min')
+    empty_merge = pd.DataFrame(
+        columns=[
+            'timestamp',
+            'feature',
+            'method',
+            'value',
+            'alert',
+            'chunk_index',
+            'chunk_start_date',
+            'chunk_end_date',
+            'threshold',
+            'drift_type',
+        ]
+    )
+    with patch.object(ModelMetrics, 'get_drift_metrics', return_value=empty_merge):
+        await start_and_await_workflow(
+            client,
+            Drift.run,
+            input_data,
+            make_workflow_id('test-drift-empty-merge'),
+        )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.drift_metrics WHERE model_id = :m'),
+            {'m': str(model_id)},
+        ).scalar()
+    assert count == 0
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_drift_chunk_period_seconds_sufficient_data_preserves_seconds_in_chunk_start_date(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_drift: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario D.4.3b: With enough sub-minute samples and ``chunk_period='s'``, drift rows
+    persist and ``chunk_start_date`` keeps second-level precision (incl. second=30).
+
+    ``DriftAnalysis._chunk_dataframe`` is patched to skip empty ``pd.Grouper(freq='s')``
+    buckets so sparse seconds between samples do not flood the pipeline with NaT rows;
+    the durable fix belongs in ``sientia_model``.
+    """
+    client = temporal_test_env.client
+    model_id = 443
+
+    base = datetime.now(timezone.utc).replace(second=0, microsecond=0) - timedelta(minutes=6)
+    target_timestamps = []
+    sensor_1_vals = []
+    sensor_2_vals = []
+    for minute_offset in range(6):
+        t0 = base + timedelta(minutes=minute_offset)
+        t1 = t0 + timedelta(seconds=30)
+        target_timestamps.append(t0.strftime('%Y-%m-%d %H:%M:%S%z'))
+        target_timestamps.append(t1.strftime('%Y-%m-%d %H:%M:%S%z'))
+        v0 = 1.0 + minute_offset * 0.1
+        v1 = v0 + 0.05
+        sensor_1_vals.extend([v0, v1])
+        sensor_2_vals.extend([10.0 + v0, 10.0 + v1])
+
+    insert_target_data_for_drift(
+        postgres_engine,
+        model_id=model_id,
+        timestamps=target_timestamps,
+        variables_values={
+            'sensor_1': sensor_1_vals,
+            'sensor_2': sensor_2_vals,
+        },
+    )
+    _force_reference_unavailable(mlflow_repository_stub)
+
+    input_data = _drift_input(model_id, chunk_period='s')
+    with patch.object(DriftAnalysis, '_chunk_dataframe', _chunk_dataframe_skip_empty_groups):
+        await start_and_await_workflow(
+            client,
+            Drift.run,
+            input_data,
+            make_workflow_id('test-drift-chunk-seconds-sufficient'),
+        )
+
+    with postgres_engine.connect() as conn:
+        rows = (
+            conn.execute(
+                text(
+                    'SELECT chunk_start_date FROM sientia_data.drift_metrics '
+                    'WHERE model_id = :m ORDER BY chunk_index'
+                ),
+                {'m': str(model_id)},
+            )
+            .mappings()
+            .all()
+        )
+
+    assert rows, 'expected at least one drift row to be persisted'
+    seconds_present = {pd.Timestamp(r['chunk_start_date']).second for r in rows}
+    assert 30 in seconds_present, (
+        f'expected at least one chunk_start_date with seconds=30, got {seconds_present}'
+    )

e2e/test_minimal_retrain.py

@@ -0,0 +1,430 @@
+"""
+End-to-end tests for the MinimalRetrain workflow.
+
+The MLflow registry is fully stubbed because no real artifacts exist in a
+test container; we only validate that the workflow:
+
+- Loads training data via ``load_query_with_minio_offload``.
+- Calls ``retrain_model`` with a payload pointing at MinIO.
+- Calls ``update_production_model`` only when retrain succeeds.
+- Persists ``sientia_data.log_retrain`` rows with all required columns;
+  success rows carry the new ``version`` / ``mlflow_run_id`` /
+  ``mlflow_experiment_id`` while failure rows leave them ``NULL``.
+
+The production DDL drops the legacy ``id`` / ``created_at`` columns and
+moves ``mlflow_experiment_id`` to ``int8`` and ``model_id`` to ``text``.
+The stubs used here therefore emit ``experiment_id`` as an integer to fit
+the new column type.
+"""
+
+from contextlib import contextmanager
+from datetime import datetime, timedelta, timezone
+from unittest.mock import MagicMock, patch
+
+import pytest
+import pandas as pd
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import (
+    insert_target_data_for_drift,
+    load_scenario_input,
+    make_workflow_id,
+    start_and_await_workflow,
+)
+from laborious.activities.activities import Activities
+from laborious.workflows.minimal_retrain import MinimalRetrain
+from sientia_model.wrappers.sientia_model import SientiaModel
+
+# Columns defined by the production DDL for ``sientia_data.log_retrain``.
+# The legacy ``retrain_reports`` table had ``id`` and ``created_at``; the new
+# DDL drops both. ``mlflow_experiment_id`` is ``int8`` and ``model_id`` is
+# ``text``.
+EXPECTED_RETRAIN_REPORT_COLUMNS = [
+    'mlflow_experiment_id',
+    'mlflow_run_id',
+    'model_id',
+    'model_name',
+    'status',
+    'timestamp',
+    'version',
+]
+
+# Matches ``retrain_model`` return ``message`` when ``success`` is True (also written to ``log_retrain.status``).
+RETRAIN_ACTIVITY_SUCCESS_MESSAGE = 'Model retrained successfully.'
+
+
+class _FakeSientiaModelForMinimalRetrain(SientiaModel):
+    """
+    Fake SientiaModel that uses the real SientiaModel lifecycle to surface
+    index-alignment issues during ``retrain()``.
+
+    It intentionally performs strict alignment inside ``_retrain_model``:
+    ``y.loc[x.index]``.
+    """
+
+    def __init__(self, *, target: str = 'sensor_1'):
+        super().__init__(
+            model_type='FakeMinimalRetrain',
+            model_version='0.0.0',
+            model=object(),
+            transformer=object(),
+        )
+        self.target = target
+        self.model_is_fitted = True
+        self.force_retrain_error = False
+
+    def store_model(  # type: ignore[override]
+        self,
+        name: str,
+        signature=None,
+        pip_requirements=None,
+        code_path=None,
+    ) -> None:
+        # No-op: E2E tests validate workflow persistence, not real MLflow artifacts.
+        return None
+
+    def _predict(self, data: pd.DataFrame):
+        pred = pd.DataFrame({'prediction': [0.5] * len(data)}, index=data.index)
+        return pred, {}
+
+    def _transform(self, data: pd.DataFrame):
+        out = data.drop(columns=[self.target], errors='ignore').copy()
+        out.index = data.index
+        return out, {}
+
+    def _train_transformer(self, train_data: pd.DataFrame, val_data: pd.DataFrame) -> None:
+        return None
+
+    def _train_model(
+        self,
+        x: pd.DataFrame,
+        y: pd.DataFrame,
+        x_val: pd.DataFrame | None = None,
+        y_val: pd.DataFrame | None = None,
+    ) -> None:
+        return None
+
+    def _retrain_transformer(self, data: pd.DataFrame) -> None:
+        return None
+
+    def _retrain_model(self, x: pd.DataFrame, y: pd.DataFrame | None) -> None:
+        if self.force_retrain_error:
+            raise RuntimeError('training did not converge')
+        if y is None:
+            return
+        # Strict alignment on purpose to reproduce the production failure mode.
+        _ = y.loc[x.index]
+
+
+@pytest.fixture
+def mlflow_repository_stub():
+    """
+    Override the shared E2E fixture: return a real fake ``SientiaModel`` wrapper
+    instead of a MagicMock wrapper.
+    """
+    repo = MagicMock()
+    repo._client = MagicMock()
+
+    wrapper = _FakeSientiaModelForMinimalRetrain(target='sensor_1')
+    repo.get_cached_model = MagicMock(return_value=wrapper)
+    return repo
+
+
+def _retrain_input(model_id: int, **overrides) -> dict:
+    """Load and override the minimal-retrain base scenario."""
+    payload = load_scenario_input('minimal_retrain_base.json', model_id=model_id)
+    payload.update(overrides)
+    return payload
+
+
+def _seed_retrain_training_rows(postgres_engine, model_id: int) -> None:
+    """
+    Insert training rows in long format that pivot cleanly into
+    ``index=timestamp`` / ``columns={sensor_1, sensor_2}`` for ``retrain_model``.
+    """
+    target_timestamps = [
+        (datetime.now(timezone.utc).replace(second=0, microsecond=0) - timedelta(minutes=10 - i))
+        .strftime('%Y-%m-%d %H:%M:%S%z')
+        for i in range(5)
+    ]
+    insert_target_data_for_drift(
+        postgres_engine,
+        model_id=model_id,
+        timestamps=target_timestamps,
+        variables_values={
+            'sensor_1': [10.0, 11.0, 12.0, 13.0, 14.0],
+            'sensor_2': [20.0, 21.0, 22.0, 23.0, 24.0],
+        },
+    )
+
+
+def _configure_retrain_happy_path(mlflow_repository_stub) -> None:
+    """
+    Wire ``mlflow_repository_stub`` so retrain + update_production succeed.
+
+    Mocks (in order of consumption):
+
+    - ``_client.get_model_version_by_alias``: returns ``mv`` with a stable
+      ``run_id`` (used as ``source_run_id``).
+    - ``start_run``: returns a context manager yielding a ``run_info`` with
+      run/experiment ids.
+    - ``log_params``: inert.
+    - ``_client.search_model_versions``: returns one registry entry whose
+      ``version`` is promoted by ``update_production_model``.
+    - ``promote_to_alias``: inert success.
+    """
+    mv_src = MagicMock()
+    mv_src.run_id = 'source-run-id'
+
+    new_version = MagicMock()
+    new_version.version = '7'
+    new_version.run_id = 'retrain-run-id'
+
+    mlflow_repository_stub._client.get_model_version_by_alias.return_value = mv_src
+
+    @contextmanager
+    def fake_start_run(**kwargs):
+        run_info = MagicMock()
+        run_info.run_id = 'retrain-run-id'
+        # ``mlflow_experiment_id`` is ``int8`` in the new DDL, so we feed an
+        # integer-compatible id from the stubbed run info.
+        run_info.experiment_id = 4242
+        yield run_info
+
+    mlflow_repository_stub.start_run.side_effect = fake_start_run
+    mlflow_repository_stub.log_params = MagicMock(return_value=None)
+    mlflow_repository_stub._client.search_model_versions.return_value = [new_version]
+    mlflow_repository_stub.promote_to_alias = MagicMock(return_value=None)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_minimal_retrain_happy_path_writes_success_report(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_minimal_retrain: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario MR.1.1: Retrain succeeds. ``sientia_data.log_retrain`` must
+    contain a success row with version/mlflow_run_id/mlflow_experiment_id
+    populated and the registry must have been told to promote the new version
+    to the configured alias.
+    """
+    client = temporal_test_env.client
+    model_id = 711
+
+    _seed_retrain_training_rows(postgres_engine, model_id)
+    _configure_retrain_happy_path(mlflow_repository_stub)
+
+    input_data = _retrain_input(model_id)
+
+    with patch('laborious.activities.mlflow.mlflow.log_artifact') as log_artifact_mock:
+        await start_and_await_workflow(
+            client,
+            MinimalRetrain.run,
+            input_data,
+            make_workflow_id('test-retrain-happy'),
+        )
+
+    with postgres_engine.connect() as conn:
+        rows = (
+            conn.execute(
+                text(
+                    'SELECT * FROM sientia_data.log_retrain '
+                    'WHERE model_id = :m'
+                ),
+                {'m': str(model_id)},
+            )
+            .mappings()
+            .all()
+        )
+
+    assert len(rows) == 1
+    row = rows[0]
+    for column in EXPECTED_RETRAIN_REPORT_COLUMNS:
+        assert column in row, f'Missing log_retrain column: {column}'
+
+    assert row['status'] == RETRAIN_ACTIVITY_SUCCESS_MESSAGE, (
+        "Expected retrain_model to return success (experiment_response['success'] is True). "
+        'Persisted log_retrain.status is the activity message; when success is False the run '
+        'never reaches mlflow.log_artifact — diagnose the retrain failure from status below, '
+        'not from a skipped artifact upload. '
+        f"Got status={row['status']!r}, version={row.get('version')!r}, "
+        f"mlflow_run_id={row.get('mlflow_run_id')!r}."
+    )
+
+    assert log_artifact_mock.called, (
+        'After a successful retrain, retrain_model must call mlflow.log_artifact for the '
+        'input CSV inside start_run.'
+    )
+
+    # ``model_id`` is now ``text``; compare against the stringified id.
+    assert row['model_id'] == str(model_id)
+    assert row['model_name'] == 'test_model'
+    assert row['version'] == '7'
+    assert row['mlflow_run_id'] == 'retrain-run-id'
+    # ``mlflow_experiment_id`` is now ``int8``; assert the integer value
+    # provided by the stubbed run info.
+    assert row['mlflow_experiment_id'] == 4242
+    assert row['timestamp'] is not None
+
+    mlflow_repository_stub.promote_to_alias.assert_called_once()
+    promote_kwargs = mlflow_repository_stub.promote_to_alias.call_args.kwargs
+    assert promote_kwargs['model_name'] == 'test_model'
+    assert promote_kwargs['version'] == '7'
+    assert promote_kwargs['alias'] == 'production'
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_minimal_retrain_failure_writes_report_without_version_columns(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_minimal_retrain: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario MR.2.1: ``wrapper.retrain`` raises. The activity must catch the
+    error, return ``success=False`` so ``update_production_model`` is skipped,
+    and ``format_retrain_report`` must produce a row with the error message
+    and NULL version columns.
+    """
+    client = temporal_test_env.client
+    model_id = 721
+
+    _seed_retrain_training_rows(postgres_engine, model_id)
+    _configure_retrain_happy_path(mlflow_repository_stub)
+    mlflow_repository_stub.get_cached_model.return_value.force_retrain_error = True
+
+    input_data = _retrain_input(model_id)
+
+    with patch('laborious.activities.mlflow.mlflow.log_artifact'):
+        await start_and_await_workflow(
+            client,
+            MinimalRetrain.run,
+            input_data,
+            make_workflow_id('test-retrain-failure'),
+        )
+
+    with postgres_engine.connect() as conn:
+        rows = (
+            conn.execute(
+                text(
+                    'SELECT * FROM sientia_data.log_retrain '
+                    'WHERE model_id = :m'
+                ),
+                {'m': str(model_id)},
+            )
+            .mappings()
+            .all()
+        )
+
+    assert len(rows) == 1
+    row = rows[0]
+    assert row['model_id'] == str(model_id)
+    assert row['model_name'] == 'test_model'
+    assert 'training did not converge' in row['status']
+    assert row['version'] is None
+    assert row['mlflow_run_id'] is None
+    assert row['mlflow_experiment_id'] is None
+
+    mlflow_repository_stub.promote_to_alias.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_minimal_retrain_missing_target_writes_failure_report(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_minimal_retrain: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario MR.2.2: ``model_config`` does not declare ``target``. The retrain
+    activity must short-circuit before any MLflow call and the report row must
+    carry the explicit guard message.
+    """
+    client = temporal_test_env.client
+    model_id = 722
+
+    _seed_retrain_training_rows(postgres_engine, model_id)
+    _configure_retrain_happy_path(mlflow_repository_stub)
+
+    input_data = _retrain_input(model_id, model_config={})
+
+    with patch('laborious.activities.mlflow.mlflow.log_artifact'):
+        await start_and_await_workflow(
+            client,
+            MinimalRetrain.run,
+            input_data,
+            make_workflow_id('test-retrain-missing-target'),
+        )
+
+    with postgres_engine.connect() as conn:
+        row = (
+            conn.execute(
+                text(
+                    'SELECT * FROM sientia_data.log_retrain '
+                    'WHERE model_id = :m'
+                ),
+                {'m': str(model_id)},
+            )
+            .mappings()
+            .first()
+        )
+
+    assert row is not None
+    assert 'target' in row['status'].lower(), (
+        f"expected target-missing message, got status={row['status']!r}"
+    )
+    assert row['version'] is None
+    mlflow_repository_stub.get_cached_model.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_minimal_retrain_no_training_data_does_not_persist_report(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_minimal_retrain: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    Scenario MR.3.1: When the training query returns no rows the workflow must
+    not persist any report row. The workflow currently raises plain
+    ``ValueError`` which Temporal treats as a workflow-task failure (causing
+    indefinite retries until the test environment times out), so the assertion
+    here is constrained to the persistence side-effect. See ``e2e/CODE_ISSUES.md``
+    issue MR-1 for the recommended ``ApplicationError`` fix.
+    """
+    client = temporal_test_env.client
+    model_id = 731
+
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+
+    _configure_retrain_happy_path(mlflow_repository_stub)
+
+    input_data = _retrain_input(model_id)
+
+    with pytest.raises(Exception), patch('laborious.activities.mlflow.mlflow.log_artifact'):
+        await start_and_await_workflow(
+            client,
+            MinimalRetrain.run,
+            input_data,
+            make_workflow_id('test-retrain-no-data'),
+        )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.log_retrain WHERE model_id = :m'),
+            {'m': str(model_id)},
+        ).scalar()
+    assert count == 0

e2e/test_minio_offload.py

@@ -0,0 +1,108 @@
+"""
+E2E-style tests for MinIO offload using a real MinIO testcontainer.
+"""
+
+from unittest.mock import patch
+
+import pytest
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import (
+    insert_sample_data,
+    load_scenario_input,
+    make_workflow_id,
+    start_and_await_workflow,
+)
+from laborious.activities.activities import Activities
+from laborious.utils.models import minio_dataframe_payload as mdp
+from laborious.workflows.predictions_batch import PredictionsBatch
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_load_query_with_minio_offload_writes_object_to_bucket(
+    postgres_engine,
+    minio_container,
+    test_activities_real_minio: Activities,
+):
+    """
+    With a tiny offload threshold, query results are uploaded as Parquet to MinIO.
+
+    Uses real MinioRepository against testcontainers MinIO (no MinIO mock).
+    """
+    model_id = 501
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+    insert_sample_data(postgres_engine, model_id, [1.0, 2.0])
+
+    scenario_input = load_scenario_input('minio_offload_load_query.json', model_id=model_id)
+    metadata = {'metadata': scenario_input['metadata']}
+    with patch.object(mdp, 'OFFLOAD_THRESHOLD_BYTES', 1):
+        payload = test_activities_real_minio.load_query_with_minio_offload(scenario_input)
+    assert payload.object_key, 'offloaded payload must reference a MinIO object'
+    assert payload.data is None or payload.data == {}, 'large payloads should not inline tabular dict'
+
+    df = payload.retrieve(test_activities_real_minio.minio_repository, metadata['metadata'])
+    assert len(df) >= 1
+
+    client = minio_container.get_client()
+    listed = list(client.list_objects('test-bucket', recursive=True))
+    names = [getattr(o, 'object_name', None) or getattr(o, '_object_name', '') for o in listed]
+    assert any(n and 'prediction_datasets' in n for n in names), f'unexpected object listing: {names!r}'
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_predictions_batch_with_minio_offload_path(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_real_minio: Worker,
+    postgres_engine,
+    test_activities_real_minio: Activities,
+):
+    """
+    Full PredictionsBatch run with offload: load step stores Parquet in MinIO; pipeline completes.
+    """
+    model_id = 502
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+        conn.execute(text(f'DELETE FROM sientia_data.predictions WHERE model_id = {model_id}'))
+        conn.execute(text(f'DELETE FROM sientia_data.transformed_data WHERE model_id = {model_id}'))
+    insert_sample_data(postgres_engine, model_id, [10.0, 20.0, 30.0])
+
+    input_data = load_scenario_input('minio_offload_workflow.json', model_id=model_id)
+
+    with patch.object(mdp, 'OFFLOAD_THRESHOLD_BYTES', 1):
+        await start_and_await_workflow(
+            temporal_test_env.client,
+            PredictionsBatch.run,
+            input_data,
+            make_workflow_id('test-batch-minio-offload'),
+        )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text(f'SELECT COUNT(*) FROM sientia_data.predictions WHERE model_id = {model_id}')
+        ).scalar()
+    assert count == 1
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_load_query_with_inline_payload_when_below_threshold(
+    postgres_engine,
+    test_activities_real_minio: Activities,
+):
+    """Scenario 4.2.1: payload stays inline when threshold is high enough."""
+    model_id = 503
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+    insert_sample_data(postgres_engine, model_id, [1.0, 2.0])
+
+    scenario_input = load_scenario_input('minio_offload_load_query.json', model_id=model_id)
+    with patch.object(mdp, 'OFFLOAD_THRESHOLD_BYTES', 10**9):
+        payload = test_activities_real_minio.load_query_with_minio_offload(scenario_input)
+
+    assert payload.object_key is None
+    assert payload.data is not None

e2e/test_opc_real_server.py

@@ -0,0 +1,201 @@
+"""
+E2E tests for OPC export using an in-process asyncua server and real OpcRepository.
+
+Covers scenarios 3.1.2, 3.2.2, 3.2.4, and 3.2.5 from e2e/scenarios.md.
+Mock-based OPC tests remain in test_predictions_batch_format_export.py.
+"""
+
+import asyncio
+import threading
+import time
+
+import pytest
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import assert_prediction, insert_sample_data, make_workflow_id, start_and_await_workflow
+from e2e.opc_test_server import UNKNOWN_NODE_ID, OpcE2ETestServer, build_opc_output_config
+from e2e.test_predictions_batch_format_export import get_base_input_data
+from laborious.activities.activities import Activities
+from laborious.activities.opc import OPC_RECONNECT_IN_PROGRESS_COMMENT
+from laborious.utils.repository.opc_repository import OpcRepository
+from laborious.workflows.predictions_batch import PredictionsBatch
+
+
+def _slow_reconnect_under_lock(repo: OpcRepository, hold_seconds: float = 0.75) -> None:
+    """
+    Hold the connection lock briefly so concurrent writes see reconnect_in_progress.
+
+    Args:
+        repo (OpcRepository): Connected repository.
+        hold_seconds (float): Time to keep the lock before reconnecting.
+    """
+    with repo._connection_lock:
+        time.sleep(hold_seconds)
+        repo._reconnect_locked()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+@pytest.mark.opc
+async def test_scenario_3_1_2_export_with_opc_only_real_server(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_real_opc: Worker,
+    test_activities_real_opc: Activities,
+    opc_e2e_server: OpcE2ETestServer,
+    postgres_engine,
+):
+    """
+    Scenario 3.1.2 (real OPC): connect, write prediction and confidence, verify server values.
+    """
+    client = temporal_test_env.client
+    model_id = 412
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = build_opc_output_config(opc_e2e_server.node_ids)
+    input_data['pi_web_api_output_config'] = None
+
+    await start_and_await_workflow(
+        client,
+        PredictionsBatch.run,
+        input_data,
+        make_workflow_id('test-opc-real-happy'),
+    )
+
+    test_activities_real_opc.pi_web_api_client.write_value.assert_not_called()
+    assert await opc_e2e_server.read_prediction() == pytest.approx(0.5)
+    assert await opc_e2e_server.read_confidence() == pytest.approx(0.0)
+    assert_prediction(postgres_engine, model_id)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+@pytest.mark.opc
+async def test_scenario_3_2_2_opc_write_error_real_server(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_real_opc: Worker,
+    test_activities_real_opc: Activities,
+    opc_e2e_server: OpcE2ETestServer,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.2 (real OPC): unknown NodeId yields generic write failure (confidence 12).
+    """
+    client = temporal_test_env.client
+    model_id = 422
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    node_ids = opc_e2e_server.node_ids
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = build_opc_output_config(
+        node_ids,
+        prediction_tag=UNKNOWN_NODE_ID,
+        confidence_tag=UNKNOWN_NODE_ID,
+    )
+    input_data['pi_web_api_output_config'] = None
+
+    await start_and_await_workflow(
+        client,
+        PredictionsBatch.run,
+        input_data,
+        make_workflow_id('test-opc-real-bad-node'),
+    )
+
+    assert_prediction(
+        postgres_engine,
+        model_id,
+        prediction_confidence=12,
+        comments='Some data could not be written to OPC servers',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+@pytest.mark.opc
+async def test_scenario_3_2_4_opc_session_bad_real_server(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_real_opc: Worker,
+    test_activities_real_opc: Activities,
+    opc_e2e_server: OpcE2ETestServer,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.4 (real OPC): server PreWrite fault injects BadSessionIdInvalid (confidence 14).
+    """
+    client = temporal_test_env.client
+    model_id = 424
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    opc_e2e_server.set_session_bad_on_write(True)
+    try:
+        input_data = get_base_input_data(model_id)
+        input_data['opc_output_config'] = build_opc_output_config(
+            opc_e2e_server.node_ids,
+            prediction_only=True,
+        )
+        input_data['pi_web_api_output_config'] = None
+
+        await start_and_await_workflow(
+            client,
+            PredictionsBatch.run,
+            input_data,
+            make_workflow_id('test-opc-real-session-bad'),
+        )
+    finally:
+        opc_e2e_server.set_session_bad_on_write(False)
+
+    assert_prediction(
+        postgres_engine,
+        model_id,
+        prediction_confidence=14,
+        comments_contains='OPC UA session/channel error: BadSessionIdInvalid',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+@pytest.mark.opc
+async def test_scenario_3_2_5_opc_write_blocked_during_reconnect_real_server(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_real_opc: Worker,
+    test_activities_real_opc: Activities,
+    opc_e2e_server: OpcE2ETestServer,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.5 (real OPC): writes rejected while reconnect holds the connection lock.
+    """
+    client = temporal_test_env.client
+    model_id = 425
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    repo = test_activities_real_opc.opc_repository['1']
+    repo._session_ready.clear()
+    reconnect_thread = threading.Thread(
+        target=_slow_reconnect_under_lock,
+        args=(repo,),
+        daemon=True,
+    )
+    reconnect_thread.start()
+
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = build_opc_output_config(opc_e2e_server.node_ids)
+    input_data['pi_web_api_output_config'] = None
+
+    try:
+        await start_and_await_workflow(
+            client,
+            PredictionsBatch.run,
+            input_data,
+            make_workflow_id('test-opc-real-reconnect-block'),
+        )
+    finally:
+        reconnect_thread.join(timeout=5.0)
+
+    assert_prediction(
+        postgres_engine,
+        model_id,
+        prediction_confidence=14,
+        comments_contains=OPC_RECONNECT_IN_PROGRESS_COMMENT,
+    )

e2e/test_predictions_batch_format_export.py

@@ -0,0 +1,836 @@
+"""
+End-to-end tests for PredictionsBatch workflow - Format and Export scenarios.
+"""
+
+from decimal import Decimal
+from typing import Any, cast
+from unittest.mock import call
+
+import pytest
+from sientia_do.notifications.models import NotificationLevel
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import (
+    assert_prediction,
+    insert_sample_data,
+    load_scenario_input,
+    make_workflow_id,
+    start_and_await_workflow,
+)
+from laborious.activities.activities import Activities
+from laborious.workflows.predictions_batch import PredictionsBatch
+
+def get_base_input_data(model_id):
+    return load_scenario_input('format_export_base.json', model_id=model_id)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_1_1_default_prediction_export(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.1.1: Default prediction export (non-None path_flag).
+
+    Triggers input_gate CONTINUE via SPECIFIC_VARIABLES_NULL_VALUES so
+    PredictionProcess calls FormatAndExportPrediction with path_flag set.
+    That workflow uses format_default_prediction (not format_prediction) and
+    skips format_transformed_data / transform Postgres export.
+
+    Optional PI Web API and OPC outputs still run when configured.
+    """
+    client = temporal_test_env.client
+
+    model_id = 311
+
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f"DELETE FROM sientia_data.predictions WHERE model_id = {model_id}"))
+        conn.execute(text(f"DELETE FROM sientia_data.transformed_data WHERE model_id = {model_id}"))
+    insert_sample_data(postgres_engine, model_id, ['NULL', 78.2])
+
+    input_data = get_base_input_data(model_id)
+    input_data['input_filters'] = {
+        'SPECIFIC_VARIABLES_NULL_VALUES': {
+            'POLICY': 'CONTINUE',
+            'CONFIG': {'variables': ['sensor_1']},
+        },
+    }
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {
+                'addr_1': {
+                    'data_type': 'float',
+                }
+            },
+            'confidence_tags': {
+                'addr_2': {
+                    'data_type': 'float',
+                }
+            },
+        }
+    }
+
+    wid = make_workflow_id('test-default-prediction')
+
+    await start_and_await_workflow(client, PredictionsBatch.run, input_data, wid)
+
+    test_activities.pi_web_api_client.write_value.assert_has_calls(
+        [
+            call(
+                web_ids=['web_id_1'],
+                value={
+                    'Timestamp': '2024-01-01 12:00:00+0000',
+                    'Value': 0,
+                },
+                metadata={
+                    'model_id': 311,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+            call(
+                web_ids=['web_id_2'],
+                value={
+                    'Timestamp': '2024-01-01 12:00:00+0000',
+                    'Value': 2,
+                },
+                metadata={
+                    'model_id': 311,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+        ],
+        any_order=True,
+    )
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.assert_has_calls(
+        [
+            call(
+                'addr_1',
+                0,
+                'float',
+                {
+                    'model_id': 311,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+            call(
+                'addr_2',
+                2,
+                'float',
+                {
+                    'model_id': 311,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+        ]
+    )
+
+    with postgres_engine.connect() as conn:
+        tf_count = conn.execute(
+            text(f"SELECT COUNT(*) FROM sientia_data.transformed_data WHERE model_id = {model_id}")
+        ).scalar()
+        assert tf_count == 0, 'transform export must be skipped when path_flag is set'
+
+    assert_prediction(
+        postgres_engine,
+        model_id,
+        prediction=0,
+        prediction_confidence=Decimal(2),
+        prediction_status='Bad',
+        comments='Input data with bad quality',
+    )
+
+
+
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_1_2_export_with_opc_only(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.1.2: Export with OPC only
+    
+    Description:
+    Export to PostgreSQL and OPC server only (no PI Web API).
+    
+    Expected Behavior:
+    - Normal formatting
+    - PostgreSQL export executed
+    - OPC export executed
+    - PI Web API activity skipped
+    - Metrics written with OPC metrics
+    
+    Assertions:
+    - PI Web API activity NOT called
+    - OPC activity called
+    - PostgreSQL export called
+    - Metrics written with opc_metrics populated
+    """
+    client = temporal_test_env.client
+
+    model_id = 312
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {
+                'addr_1': {
+                    'data_type': 'float',
+                }
+            },
+            'confidence_tags': {
+                'addr_2': {
+                    'data_type': 'float',
+                }
+            },
+        }
+    }
+    input_data['pi_web_api_output_config'] = None  # No PI Web API config
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-opc-only')
+    )
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.assert_has_calls(
+        [
+            call(
+                'addr_1',
+                0.5,
+                'float',
+                {
+                    'model_id': 312,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+            call(
+                'addr_2',
+                0,
+                'float',
+                {
+                    'model_id': 312,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+        ]
+    )
+
+    test_activities.pi_web_api_client.write_value.assert_not_called()
+
+    assert_prediction(postgres_engine, model_id)
+
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_1_3_export_with_pi_web_api_only(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.1.3: Export with PI Web API only
+    
+    Description:
+    Export to PostgreSQL and PI Web API only (no OPC).
+    
+    Expected Behavior:
+    - Normal formatting
+    - PostgreSQL export executed
+    - PI Web API export executed
+    - OPC activity skipped
+    - Metrics written without OPC metrics
+    
+    Assertions:
+    - OPC activity NOT called
+    - PI Web API activity called
+    - PostgreSQL export called
+    - Metrics written with empty opc_metrics
+    """
+    client = temporal_test_env.client
+
+    model_id = 313
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    input_data = get_base_input_data(model_id)
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+    input_data['opc_output_config'] = None  # No OPC config
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-pi-api-only')
+    )
+
+    test_activities.pi_web_api_client.write_value.assert_has_calls(
+        [
+            call(
+                web_ids=['web_id_1'],
+                value={
+                    'Timestamp': '2024-01-01 12:00:00+0000',
+                    'Value': 0.5,
+                },
+                metadata={
+                    'model_id': 313,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+            call(
+                web_ids=['web_id_2'],
+                value={
+                    'Timestamp': '2024-01-01 12:00:00+0000',
+                    'Value': 0,
+                },
+                metadata={
+                    'model_id': 313,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+        ],
+        any_order=True,
+    )
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.assert_not_called()
+
+    assert_prediction(postgres_engine, model_id)
+
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_1_4_export_without_optional_outputs(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.1.4: Export Without Optional Outputs
+    
+    Description:
+    Export only to PostgreSQL (no OPC or PI Web API).
+    
+    Expected Behavior:
+    - Normal formatting
+    - Only PostgreSQL export executed
+    - OPC and PI Web API activities skipped
+    - Metrics written without OPC metrics
+    
+    Assertions:
+    - PI Web API activity NOT called
+    - OPC activity NOT called
+    - PostgreSQL export called
+    - Metrics written with empty opc_metrics
+    """
+    client = temporal_test_env.client
+
+    model_id = 314
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = None  # No OPC config
+    input_data['pi_web_api_output_config'] = None  # No PI Web API config
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-no-optional-outputs')
+    )
+
+    test_activities.pi_web_api_client.write_value.assert_not_called()
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.assert_not_called()
+
+    assert_prediction(postgres_engine, model_id)
+
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_1_5_export_without_transformed_data(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.1.5: Export Without Transformed Data
+    
+    Description:
+    Only prediction exported, no transform table.
+    
+    Expected Behavior:
+    - Only prediction formatted and exported
+    - Transform export skipped
+    - Single PostgreSQL write
+    
+    Assertions:
+    - format_transformed_data NOT called
+    - One PostgreSQL export
+    - Transform table remains empty
+    """
+    client = temporal_test_env.client
+
+    model_id = 315
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f"DELETE FROM sientia_data.transformed_data WHERE model_id = {model_id}"))
+
+    input_data = get_base_input_data(model_id)
+    input_data['save_transform'] = False  # Don't save transformed data
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {
+                'addr_1': {
+                    'data_type': 'float',
+                }
+            },
+            'confidence_tags': {
+                'addr_2': {
+                    'data_type': 'float',
+                }
+            },
+        }
+    }
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-no-transform-export')
+    )
+
+    test_activities.pi_web_api_client.write_value.assert_has_calls(
+        [
+            call(
+                web_ids=['web_id_1'],
+                value={
+                    'Timestamp': '2024-01-01 12:00:00+0000',
+                    'Value': 0.5,
+                },
+                metadata={
+                    'model_id': 315,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+            call(
+                web_ids=['web_id_2'],
+                value={
+                    'Timestamp': '2024-01-01 12:00:00+0000',
+                    'Value': 0,
+                },
+                metadata={
+                    'model_id': 315,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+        ],
+        any_order=True,
+    )
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.assert_has_calls(
+        [
+            call(
+                'addr_1',
+                0.5,
+                'float',
+                {
+                    'model_id': 315,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+            call(
+                'addr_2',
+                0,
+                'float',
+                {
+                    'model_id': 315,
+                    'model_name': 'test_model',
+                    'schedule_name': 'test-schedule',
+                    'workflow_name': 'predictions_batch',
+                },
+            ),
+        ]
+    )
+
+    with postgres_engine.connect() as conn:
+        result_query = conn.execute(
+            text(f"SELECT COUNT(*) FROM sientia_data.transformed_data WHERE model_id = {model_id}")
+        )
+        count = result_query.scalar()
+        assert count == 0, f"Expected transform table to be empty, but found {count} records"
+
+    assert_prediction(postgres_engine, model_id)
+
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_2_1_pi_web_api_write_error(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    notification_inserts,
+):
+    """
+    Scenario 3.2.1: PI Web API Write Error
+
+    Export failure is handled inside the activity; there is no retry loop. The
+    workflow completes and PostgreSQL stores prediction_confidence 13 and the
+    error message in comments.
+    """
+    client = temporal_test_env.client
+
+    model_id = 321
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    test_activities.pi_web_api_client.write_value.side_effect = Exception(
+        "PI Web API service unavailable")
+
+    input_data = get_base_input_data(model_id)
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {
+                'addr_1': {
+                    'data_type': 'float',
+                }
+            },
+            'confidence_tags': {
+                'addr_2': {
+                    'data_type': 'float',
+                }
+            },
+        }
+    }
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-pi-api-error')
+    )
+
+    assert_prediction(
+        postgres_engine, model_id,
+        prediction_confidence=13,
+        comments='PI Web API service unavailable',
+    )
+    assert notification_inserts.call_count >= 1
+
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_2_2_opc_write_error(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.2: OPC Write Error
+
+    OPC failure is reported without failing the workflow; there is no retry
+    loop. PostgreSQL stores prediction_confidence 12 and OPC error comments.
+    """
+    client = temporal_test_env.client
+
+    model_id = 322
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.return_value = (False, {
+        'notification_id': 'OPC_WRITE_DATA_ERROR_1',
+        'message': 'OPC server unavailable',
+        'block': 'opc_repository',
+        'level': NotificationLevel.ERROR,
+        'attachment_content': 'OPC server unavailable',
+    })
+    
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {
+                'addr_1': {
+                    'data_type': 'float',
+                }
+            },
+            'confidence_tags': {
+                'addr_2': {
+                    'data_type': 'float',
+                }
+            },
+        }
+    }
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-opc-error')
+    )
+
+    assert_prediction(
+        postgres_engine, model_id,
+        prediction_confidence=12,
+        comments='Some data could not be written to OPC servers',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_2_4_opc_session_bad_mock(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.4 (mock): Tier-1 session error maps to confidence 14.
+    """
+    client = temporal_test_env.client
+    model_id = 324
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.return_value = (
+        False,
+        {
+            'opc_error_kind': 'session_bad',
+            'opc_status': 'BadSessionIdInvalid',
+            'notification_id': 'OPC_WRITE_DATA_ERROR_1',
+            'message': 'OPC session invalid',
+            'block': 'opc_repository',
+            'level': NotificationLevel.ERROR,
+            'attachment_content': 'BadSessionIdInvalid',
+        },
+    )
+
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {'addr_1': {'data_type': 'float'}},
+            'confidence_tags': {'addr_2': {'data_type': 'float'}},
+        }
+    }
+    input_data['pi_web_api_output_config'] = None
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-opc-session-bad-mock')
+    )
+
+    assert_prediction(
+        postgres_engine,
+        model_id,
+        prediction_confidence=14,
+        comments_contains='OPC UA session/channel error: BadSessionIdInvalid',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_2_5_opc_reconnect_in_progress_mock(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.5 (mock): reconnect_in_progress maps to confidence 14.
+    """
+    client = temporal_test_env.client
+    model_id = 325
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    opc_write_data.return_value = (
+        False,
+        {
+            'opc_error_kind': 'reconnect_in_progress',
+            'notification_id': 'OPC_WRITE_DATA_ERROR_1',
+            'message': 'OPC reconnect in progress',
+            'block': 'opc_repository',
+            'level': NotificationLevel.ERROR,
+            'attachment_content': 'reconnect',
+        },
+    )
+
+    input_data = get_base_input_data(model_id)
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {'addr_1': {'data_type': 'float'}},
+            'confidence_tags': {'addr_2': {'data_type': 'float'}},
+        }
+    }
+    input_data['pi_web_api_output_config'] = None
+
+    await start_and_await_workflow(
+        client,
+        PredictionsBatch.run,
+        input_data,
+        make_workflow_id('test-opc-reconnect-mock'),
+    )
+
+    assert_prediction(
+        postgres_engine,
+        model_id,
+        prediction_confidence=14,
+        comments_contains='OPC UA reconnect in progress',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_2_3_pi_web_api_partial_write_error(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.2.3: PI Web API Partial Write Error
+
+    Partial PI write: confidence 13, descriptive comments, workflow completes
+    without an activity retry loop.
+    """
+    client = temporal_test_env.client
+
+    model_id = 323
+
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    test_activities.pi_web_api_client.set_side_effect(
+        [
+            # Prediction batch: two web_ids requested, only one acknowledged.
+            [{'WebId': 'web_id_1', 'Errors': []}],
+            # Confidence write succeeds.
+            [{'WebId': 'web_id_2', 'Errors': []}],
+        ]
+    )
+
+    input_data = get_base_input_data(model_id)
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1', 'tag_3': 'web_id_3'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {
+                'addr_1': {
+                    'data_type': 'float',
+                }
+            },
+            'confidence_tags': {
+                'addr_2': {
+                    'data_type': 'float',
+                }
+            },
+        }
+    }
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-pi-api-partial-error')
+    )
+
+    assert_prediction(
+        postgres_engine, model_id,
+        prediction_confidence=13,
+        comments="The number of written tags does not match the number of tag names: Expected ['tag_1', 'tag_3'] tags, but ['tag_1'] tags were written.",
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_3_3_1_combined_pi_and_opc_outputs(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario 3.3.1: PI and OPC enabled together.
+    """
+    client = temporal_test_env.client
+    model_id = 333
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+
+    input_data = get_base_input_data(model_id)
+    input_data['pi_web_api_output_config'] = {
+        'endpoint': 'test_endpoint',
+        'prediction_tags': {'tag_1': 'web_id_1'},
+        'confidence_tags': {'tag_2': 'web_id_2'},
+    }
+    input_data['opc_output_config'] = {
+        '1': {
+            'prediction_tags': {'addr_1': {'data_type': 'float'}},
+            'confidence_tags': {'addr_2': {'data_type': 'float'}},
+        }
+    }
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-pi-opc-combined')
+    )
+
+    assert test_activities.pi_web_api_client.write_value.call_count == 2
+    opc_write_data = cast(Any, test_activities.opc_repository['1'].write_data)
+    assert opc_write_data.call_count == 2
+    assert_prediction(postgres_engine, model_id)
+    

e2e/test_predictions_batch_main_workflow.py

@@ -0,0 +1,181 @@
+"""
+End-to-end tests for PredictionsBatch workflow - Main workflow scenarios.
+"""
+
+import asyncio
+
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+import pytest
+
+from e2e.helpers import load_scenario_input, make_workflow_id, start_and_await_workflow
+from laborious.activities.activities import Activities
+from laborious.workflows.predictions_batch import PredictionsBatch
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_1_1_1_happy_path_complete_success(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """Scenario 1.1.1: Happy path with SQL load, MLflow mocks, Postgres predictions and transforms."""
+    client = temporal_test_env.client
+
+    with postgres_engine.begin() as conn:
+        conn.execute(text('DELETE FROM sientia_data.laborious_data WHERE model_id = 123'))
+        insert_sql = """
+            INSERT INTO sientia_data.laborious_data (model_id, variable, value, timestamp, created_at)
+            VALUES
+                (123, 'sensor_1', 23.5, '2024-01-01 12:00:00+00:00', '2024-01-01 12:00:00+00:00'),
+                (123, 'sensor_2', 78.2, '2024-01-01 12:00:00+00:00', '2024-01-01 12:00:00+00:00'),
+                (123, 'sensor_3', 120.8, '2024-01-01 12:00:00+00:00', '2024-01-01 12:00:00+00:00')
+        """
+        conn.execute(text(insert_sql))
+
+    input_data = load_scenario_input('main_happy_path.json', model_id=123)
+
+    await start_and_await_workflow(
+        client,
+        PredictionsBatch.run,
+        input_data,
+        make_workflow_id('test-predictions-batch'),
+    )
+
+    schema_name = 'sientia_data'
+    with postgres_engine.connect() as conn:
+        result_query = conn.execute(
+            text(
+                f'SELECT model_id, prediction, prediction_confidence, response_time, prediction_status, comments '
+                f'FROM {schema_name}.predictions WHERE model_id = 123'
+            )
+        )
+        prediction_rows = result_query.fetchall()
+        assert len(prediction_rows) == 1
+        row = prediction_rows[0]
+        assert row[0] == 123
+        assert row[1] == 0.5
+        assert row[2] == 0, f'Expected prediction_confidence=0, got {row[2]}'
+        assert row[3] is not None
+        assert row[4] == 'Good'
+        assert row[5] == ''
+
+        result_query = conn.execute(
+            text(
+                f'SELECT model_id, variable, value FROM {schema_name}.transformed_data WHERE model_id = 123'
+            )
+        )
+        transformed_rows = result_query.fetchall()
+        assert len(transformed_rows) == 2
+        assert transformed_rows[0][0] == 123
+        assert transformed_rows[0][1] == 'feature_1'
+        assert float(transformed_rows[0][2]) == 0.234
+        assert transformed_rows[1][0] == 123
+        assert transformed_rows[1][1] == 'feature_2'
+        assert float(transformed_rows[1][2]) == 0.783
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_1_2_1_sql_query_execution_error(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """Invalid SQL: workflow may complete with early exit; no prediction rows."""
+    client = temporal_test_env.client
+
+    input_data = load_scenario_input('main_sql_error.json', model_id=128)
+
+    await start_and_await_workflow(
+        client,
+        PredictionsBatch.run,
+        input_data,
+        make_workflow_id('test-sql-error'),
+    )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.predictions WHERE model_id = 128')
+        ).scalar()
+    assert count == 0
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_1_2_2_missing_required_parameters(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """Missing query: workflow does not produce predictions and is terminated explicitly."""
+    client = temporal_test_env.client
+
+    input_data = load_scenario_input('main_missing_required.json', model_id=129)
+
+    handle = await client.start_workflow(
+        PredictionsBatch.run,
+        input_data,
+        id=make_workflow_id('test-missing-param'),
+        task_queue='test-queue',
+    )
+
+    # Let Temporal process a few workflow tasks; for this case, result() can hang.
+    await asyncio.sleep(2.0)
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.predictions WHERE model_id = 129')
+        ).scalar()
+    assert count == 0
+
+    await handle.terminate('expected failure path in e2e test (missing required parameters)')
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_1_2_3_invalid_datetime_column_specification(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """Invalid datetime column: no predictions persisted; workflow terminated after validation."""
+    client = temporal_test_env.client
+
+    with postgres_engine.begin() as conn:
+        conn.execute(text('DELETE FROM sientia_data.laborious_data WHERE model_id = 130'))
+        conn.execute(
+            text(
+                """
+            INSERT INTO sientia_data.laborious_data (model_id, variable, value, timestamp, created_at)
+            VALUES (130, 'sensor_1', 23.5, '2024-01-01 12:00:00+00:00', '2024-01-01 12:00:00+00:00')
+        """
+            )
+        )
+
+    input_data = load_scenario_input('main_invalid_datetime.json', model_id=130)
+
+    handle = await client.start_workflow(
+        PredictionsBatch.run,
+        input_data,
+        id=make_workflow_id('test-invalid-datetime-col'),
+        task_queue='test-queue',
+    )
+
+    # Let Temporal process and surface the failure path internally.
+    await asyncio.sleep(2.0)
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text('SELECT COUNT(*) FROM sientia_data.predictions WHERE model_id = 130')
+        ).scalar()
+    assert count == 0
+
+    await handle.terminate('expected failure path in e2e test (invalid datetime column)')

e2e/test_predictions_batch_prediction_process.py

@@ -0,0 +1,488 @@
+"""
+End-to-end tests for PredictionsBatch workflow - Prediction Process scenarios.
+"""
+
+from decimal import Decimal
+from unittest.mock import MagicMock, patch
+
+import numpy as np
+import pandas as pd
+import pytest
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import (
+    assert_continue,
+    assert_postgres_unique_violation_in_chain,
+    assert_prediction,
+    assert_prediction_row_count,
+    assert_repeat,
+    assert_stop,
+    insert_sample_data,
+    insert_sample_prediction,
+    load_scenario_input,
+    make_workflow_id,
+    start_and_await_workflow,
+)
+from laborious.activities.activities import Activities
+from laborious.utils.models import minio_dataframe_payload as minio_payload_module
+from laborious.workflows.predictions_batch import PredictionsBatch
+
+DISTINCT_BATCH_TIMESTAMP = '2024-01-01 13:00:00+00:00'
+HISTORY_TIMESTAMP = '2024-01-01 12:00:00+00:00'
+
+
+def get_base_input_data(model_id):
+    return load_scenario_input('prediction_process_base.json', model_id=model_id)
+
+
+@pytest.fixture
+def bad_data_model(mlflow_repository_stub):
+    mlflow_repository_stub.stub_wrapper.transform = MagicMock(
+        side_effect=Exception('Bad data model')
+    )
+    return mlflow_repository_stub.stub_wrapper
+
+
+@pytest.fixture
+def bad_predict_model(mlflow_repository_stub):
+    wrapper = mlflow_repository_stub.stub_wrapper
+
+    def _good_transform(data):
+        result = pd.DataFrame(
+            {
+                'feature_1': [0.234] * len(data),
+                'feature_2': [0.783] * len(data),
+            }
+        )
+        result.index = data.index
+        return result, {}
+
+    wrapper.transform.side_effect = _good_transform
+    wrapper.predict = MagicMock(side_effect=Exception('Bad predict model'))
+    return wrapper
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_1_1_input_gate_triggers_continue(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """Input gate CONTINUE: export default prediction; MLflow transform/predict not used."""
+    client = temporal_test_env.client
+    model_id = 211
+    insert_sample_data(postgres_engine, model_id, ['NULL', 78.2])
+    input_data = get_base_input_data(model_id)
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-continue-policy')
+    )
+    assert_continue(postgres_engine, model_id)
+    mlflow_repository_stub.stub_wrapper.transform.assert_not_called()
+    mlflow_repository_stub.stub_wrapper.predict.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_1_2_input_gate_triggers_stop(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """Input gate STOP: no export, no MLflow."""
+    client = temporal_test_env.client
+    model_id = 212
+    insert_sample_data(postgres_engine, model_id, ['NULL', 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['input_filters']['SPECIFIC_VARIABLES_NULL_VALUES']['POLICY'] = 'STOP'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-input-stop')
+    )
+    assert_stop(postgres_engine, model_id)
+    mlflow_repository_stub.stub_wrapper.transform.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_1_3_input_gate_repeat_batch_timestamp_equals_history_fails(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    REPEAT uses ``last_timestamp`` from the batch payload as the new row's ``timestamp``.
+    When it equals the only historical prediction row, Postgres rejects the duplicate key.
+    """
+    client = temporal_test_env.client
+    model_id = 213
+    insert_sample_data(postgres_engine, model_id, ['NULL', 78.2], data_timestamp=HISTORY_TIMESTAMP)
+    insert_sample_prediction(postgres_engine, model_id, prediction_timestamp=HISTORY_TIMESTAMP)
+    input_data = get_base_input_data(model_id)
+    input_data['input_filters']['SPECIFIC_VARIABLES_NULL_VALUES']['POLICY'] = 'REPEAT'
+    with pytest.raises(Exception) as excinfo:
+        await start_and_await_workflow(
+            client, PredictionsBatch.run, input_data, make_workflow_id('test-input-repeat-collision')
+        )
+    assert_postgres_unique_violation_in_chain(excinfo.value)
+    assert_prediction_row_count(postgres_engine, model_id, 1)
+    mlflow_repository_stub.stub_wrapper.transform.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_1_3_input_gate_repeat_distinct_batch_timestamp_inserts_second_row(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """REPEAT succeeds when batch ``last_timestamp`` differs from the historical prediction row."""
+    client = temporal_test_env.client
+    model_id = 2131
+    insert_sample_data(
+        postgres_engine, model_id, ['NULL', 78.2], data_timestamp=DISTINCT_BATCH_TIMESTAMP
+    )
+    data = insert_sample_prediction(postgres_engine, model_id, prediction_timestamp=HISTORY_TIMESTAMP)
+    input_data = get_base_input_data(model_id)
+    input_data['input_filters']['SPECIFIC_VARIABLES_NULL_VALUES']['POLICY'] = 'REPEAT'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-input-repeat-ok')
+    )
+    assert_repeat(postgres_engine, model_id, data)
+    mlflow_repository_stub.stub_wrapper.transform.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_1_4_input_gate_repeat_without_prior_prediction(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """REPEAT when no prior row in predictions: repeat_last_prediction runs; still no new duplicate export path."""
+    client = temporal_test_env.client
+    model_id = 214
+    insert_sample_data(postgres_engine, model_id, ['NULL', 78.2])
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.predictions WHERE model_id = {model_id}'))
+    input_data = get_base_input_data(model_id)
+    input_data['input_filters']['SPECIFIC_VARIABLES_NULL_VALUES']['POLICY'] = 'REPEAT'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-input-repeat-no-history')
+    )
+    assert_stop(postgres_engine, model_id)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_2_1_transform_gate_triggers_continue(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_data_model,
+):
+    client = temporal_test_env.client
+    model_id = 221
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_transform_filters']['API_ERROR']['POLICY'] = 'CONTINUE'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-transform-continue')
+    )
+    assert_continue(
+        postgres_engine=postgres_engine,
+        model_id=model_id,
+        prediction_confidence=Decimal(10),
+        comments='Unknown MLFlow API error',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_2_2_transform_gate_triggers_stop(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_data_model,
+    mlflow_repository_stub,
+):
+    client = temporal_test_env.client
+    model_id = 222
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_transform_filters']['API_ERROR']['POLICY'] = 'STOP'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-transform-stop')
+    )
+    assert_stop(postgres_engine, model_id)
+    mlflow_repository_stub.stub_wrapper.predict.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_2_3_transform_gate_repeat_batch_timestamp_equals_history_fails(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_data_model,
+):
+    client = temporal_test_env.client
+    model_id = 223
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2], data_timestamp=HISTORY_TIMESTAMP)
+    insert_sample_prediction(postgres_engine, model_id, prediction_timestamp=HISTORY_TIMESTAMP)
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_transform_filters']['API_ERROR']['POLICY'] = 'REPEAT'
+    with pytest.raises(Exception) as excinfo:
+        await start_and_await_workflow(
+            client, PredictionsBatch.run, input_data, make_workflow_id('test-transform-repeat-collision')
+        )
+    assert_postgres_unique_violation_in_chain(excinfo.value)
+    assert_prediction_row_count(postgres_engine, model_id, 1)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_2_3_transform_gate_repeat_distinct_batch_timestamp_inserts_second_row(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_data_model,
+):
+    client = temporal_test_env.client
+    model_id = 2231
+    insert_sample_data(
+        postgres_engine, model_id, [60.0, 78.2], data_timestamp=DISTINCT_BATCH_TIMESTAMP
+    )
+    data = insert_sample_prediction(postgres_engine, model_id, prediction_timestamp=HISTORY_TIMESTAMP)
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_transform_filters']['API_ERROR']['POLICY'] = 'REPEAT'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-transform-repeat-ok')
+    )
+    assert_repeat(postgres_engine, model_id, data)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_2_4_transform_content_gate_nan_values_stop(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """mlflow_content_gate triggers STOP when transform output is all NaN (NAN_VALUES filter)."""
+    client = temporal_test_env.client
+    model_id = 224
+
+    def all_nan_transform(data):
+        result = pd.DataFrame(
+            {'feature_1': [np.nan] * len(data), 'feature_2': [np.nan] * len(data)}
+        )
+        result.index = data.index
+        return result, {}
+
+    mlflow_repository_stub.stub_wrapper.transform = MagicMock(side_effect=all_nan_transform)
+
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_transform_filters'] = {
+        'API_ERROR': {'POLICY': 'STOP', 'CONFIG': {}},
+        'NAN_VALUES': {'POLICY': 'STOP', 'CONFIG': {}},
+    }
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-transform-content-stop')
+    )
+    assert_stop(postgres_engine, model_id)
+    mlflow_repository_stub.stub_wrapper.predict.assert_not_called()
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_3_1_predict_gate_triggers_continue(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_predict_model,
+):
+    client = temporal_test_env.client
+    model_id = 231
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_predict_filters']['API_ERROR']['POLICY'] = 'CONTINUE'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-predict-continue')
+    )
+    assert_continue(
+        postgres_engine=postgres_engine,
+        model_id=model_id,
+        prediction_confidence=Decimal(10),
+        comments='Unknown MLFlow API error',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_3_2_predict_gate_triggers_stop(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_predict_model,
+):
+    client = temporal_test_env.client
+    model_id = 232
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_predict_filters']['API_ERROR']['POLICY'] = 'STOP'
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-predict-stop')
+    )
+    assert_stop(postgres_engine, model_id)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_3_3_predict_gate_repeat_batch_timestamp_equals_history_fails(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_predict_model,
+):
+    client = temporal_test_env.client
+    model_id = 233
+    insert_sample_data(postgres_engine, model_id, [23.5, 78.2], data_timestamp=HISTORY_TIMESTAMP)
+    insert_sample_prediction(postgres_engine, model_id, prediction_timestamp=HISTORY_TIMESTAMP)
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_predict_filters']['API_ERROR']['POLICY'] = 'REPEAT'
+    input_data['path_priority'] = ['REPEAT', 'STOP', 'CONTINUE']
+    with pytest.raises(Exception) as excinfo:
+        await start_and_await_workflow(
+            client, PredictionsBatch.run, input_data, make_workflow_id('test-predict-repeat-collision')
+        )
+    assert_postgres_unique_violation_in_chain(excinfo.value)
+    assert_prediction_row_count(postgres_engine, model_id, 1)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_3_3_predict_gate_repeat_distinct_batch_timestamp_inserts_second_row(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_predict_model,
+):
+    client = temporal_test_env.client
+    model_id = 2331
+    insert_sample_data(
+        postgres_engine, model_id, [23.5, 78.2], data_timestamp=DISTINCT_BATCH_TIMESTAMP
+    )
+    data = insert_sample_prediction(postgres_engine, model_id, prediction_timestamp=HISTORY_TIMESTAMP)
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_predict_filters']['API_ERROR']['POLICY'] = 'REPEAT'
+    input_data['path_priority'] = ['REPEAT', 'STOP', 'CONTINUE']
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-predict-repeat-ok')
+    )
+    assert_repeat(postgres_engine, model_id, data)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_4_1_input_empty_data_stop(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """EMPTY_DATA filter with STOP when query returns no rows (offload payload empty)."""
+    client = temporal_test_env.client
+    model_id = 241
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+    input_data = get_base_input_data(model_id)
+    input_data['input_filters'] = {'EMPTY_DATA': {'POLICY': 'STOP', 'CONFIG': {}}}
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-empty-data-stop')
+    )
+    assert_stop(postgres_engine, model_id)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_scenario_2_4_1_priority_conflict_resolution(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    bad_data_model,
+):
+    """Conflicting filter outputs must honor configured path_priority order."""
+    client = temporal_test_env.client
+    model_id = 242
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2])
+    input_data = get_base_input_data(model_id)
+    input_data['mlflow_transform_filters'] = {'API_ERROR': {'POLICY': 'CONTINUE', 'CONFIG': {}}}
+    input_data['mlflow_predict_filters'] = {'API_ERROR': {'POLICY': 'STOP', 'CONFIG': {}}}
+    input_data['path_priority'] = ['STOP', 'CONTINUE', 'REPEAT']
+
+    await start_and_await_workflow(
+        client, PredictionsBatch.run, input_data, make_workflow_id('test-priority-conflict')
+    )
+    assert_continue(
+        postgres_engine=postgres_engine,
+        model_id=model_id,
+        prediction_confidence=Decimal(10),
+        comments='Unknown MLFlow API error',
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_e2e_request_predict_inline_minio_payload_with_datetimeindex(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker: Worker,
+    test_activities: Activities,
+    postgres_engine,
+    mlflow_repository_stub,
+):
+    """
+    High offload threshold forces inline tabular dicts; ``DatetimeIndex`` must serialize as JSON
+    (string index keys via ``MinioDataFramePayload.from_dataframe``) so ``request_predict`` completes.
+    """
+    client = temporal_test_env.client
+    model_id = 252
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+        conn.execute(text(f'DELETE FROM sientia_data.predictions WHERE model_id = {model_id}'))
+        conn.execute(text(f'DELETE FROM sientia_data.transformed_data WHERE model_id = {model_id}'))
+    insert_sample_data(postgres_engine, model_id, [60.0, 78.2])
+    input_data = get_base_input_data(model_id)
+
+    with patch.object(minio_payload_module, 'OFFLOAD_THRESHOLD_BYTES', 10**9):
+        await start_and_await_workflow(
+            client,
+            PredictionsBatch.run,
+            input_data,
+            make_workflow_id('test-predict-inline-json-datetimeindex'),
+        )
+
+    assert_prediction(postgres_engine, model_id, prediction=0.5, prediction_confidence=0)
+    mlflow_repository_stub.stub_wrapper.predict.assert_called()

e2e/test_simple_metrics.py

@@ -0,0 +1,333 @@
+"""
+End-to-end tests for the SimpleMetrics workflow.
+
+Coverage focus:
+
+- Happy path computes rmse/mse/mae/r2 from predictions joined against ``laborious_data``
+  and persists rows to ``sientia_data.simple_metrics`` with all required columns.
+- Subset metric selection (only rmse) writes exactly the requested rows.
+- Zero-variance target produces ``r2=0`` per division-by-zero guard.
+- Empty join (no overlapping data) short-circuits without persisting anything.
+"""
+
+from datetime import datetime, timedelta, timezone
+from decimal import Decimal
+
+import math
+import pytest
+from sqlalchemy import text
+from temporalio.testing import WorkflowEnvironment
+from temporalio.worker import Worker
+
+from e2e.helpers import load_scenario_input, make_workflow_id, start_and_await_workflow
+from laborious.activities.activities import Activities
+from laborious.workflows.simple_metrics import SimpleMetrics
+
+# Columns defined by the production DDL for ``sientia_data.simple_metrics``.
+EXPECTED_SIMPLE_METRICS_COLUMNS = [
+    'id',
+    'model_id',
+    'metric',
+    'value',
+    'timestamp',
+    'data_size',
+    'interval_minutes',
+    'created_at',
+]
+
+# ``timestamp`` is now nullable per the new DDL (production code may write it
+# null when the upstream data has no usable instant); skip the non-null check
+# for it while still validating presence.
+NULLABLE_SIMPLE_METRICS_COLUMNS = {'timestamp'}
+
+
+def _simple_metrics_input(model_id: int, **overrides) -> dict:
+    """Load and override the simple-metrics base scenario."""
+    payload = load_scenario_input('simple_metrics_base.json', model_id=model_id)
+    payload.update(overrides)
+    return payload
+
+
+def _seed_predictions_and_targets(
+    postgres_engine,
+    model_id: int,
+    pairs: list[tuple[float, float]],
+    target_name: str = 'sensor_target',
+    offset_minutes: int = 6,
+) -> list[str]:
+    """
+    Insert matching prediction/target rows used by the SimpleMetrics SQL JOIN.
+
+    For each ``(prediction, target)`` pair we write a row in ``predictions`` and
+    a matching row in ``laborious_data`` with ``variable=target_name`` so the
+    inner join in the workflow query yields one row per pair.
+
+    Args:
+     - postgres_engine: SQLAlchemy engine bound to the test container.
+     - model_id: Model id stamped on every row.
+     - pairs: ``(prediction, target)`` pairs, one per minute.
+     - target_name: Variable name in ``laborious_data`` representing the target.
+     - offset_minutes: Earliest row sits this many minutes ago so timestamps fall
+       inside the workflow's recent-data window.
+
+    Return:
+     List of timestamp strings written for the inserted rows.
+    """
+    base = datetime.now(timezone.utc).replace(second=0, microsecond=0) - timedelta(
+        minutes=offset_minutes
+    )
+    timestamps = [
+        (base + timedelta(minutes=i)).strftime('%Y-%m-%d %H:%M:%S%z')
+        for i in range(len(pairs))
+    ]
+
+    prediction_rows = []
+    target_rows = []
+    for index, (prediction, target_value) in enumerate(pairs):
+        ts = timestamps[index]
+        prediction_rows.append(
+            f"({model_id}, {prediction}, 0, 0, 'Good', '{ts}', '{ts}')"
+        )
+        target_rows.append(
+            f"({model_id}, '{target_name}', {target_value}, '{ts}', '{ts}')"
+        )
+
+    with postgres_engine.begin() as conn:
+        conn.execute(text(f'DELETE FROM sientia_data.predictions WHERE model_id = {model_id}'))
+        conn.execute(text(f'DELETE FROM sientia_data.laborious_data WHERE model_id = {model_id}'))
+        # The SimpleMetrics SQL JOIN only filters ``predictions.model_id``; it does
+        # NOT filter ``laborious_data.model_id`` (see ``e2e/CODE_ISSUES.md`` issue
+        # SM-1). Without this cross-model cleanup, a previous test's target rows
+        # under the same variable name would join into this test's predictions
+        # whenever timestamps happened to overlap.
+        conn.execute(
+            text(
+                "DELETE FROM sientia_data.laborious_data "
+                "WHERE variable IN (:sensor_default, :target_name) "
+                "AND timestamp >= NOW() - INTERVAL '120 minutes'"
+            ),
+            {'sensor_default': 'sensor_target', 'target_name': target_name},
+        )
+        if prediction_rows:
+            conn.execute(
+                text(
+                    'INSERT INTO sientia_data.predictions '
+                    '(model_id, prediction, prediction_confidence, response_time, '
+                    'prediction_status, "timestamp", created_at) VALUES '
+                    + ', '.join(prediction_rows)
+                )
+            )
+            conn.execute(
+                text(
+                    'INSERT INTO sientia_data.laborious_data '
+                    '(model_id, variable, value, "timestamp", created_at) VALUES '
+                    + ', '.join(target_rows)
+                )
+            )
+    return timestamps
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_simple_metrics_happy_path_persists_all_metrics_and_columns(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_simple_metrics: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario S.1.1: rmse/mse/mae/r2 are calculated from a deterministic
+    prediction/target pair set and written one row per metric. Every column
+    expected by ``sientia_data.simple_metrics`` must be populated (except the
+    nullable ``timestamp`` column) and the numerical values must match
+    closed-form expectations.
+    """
+    client = temporal_test_env.client
+    model_id = 511
+
+    pairs = [
+        (1.0, 2.0),
+        (2.0, 4.0),
+        (3.0, 5.0),
+        (4.0, 9.0),
+        (5.0, 12.0),
+    ]
+    diffs = [target - prediction for prediction, target in pairs]
+    n = len(diffs)
+    expected_rmse = math.sqrt(sum(d * d for d in diffs) / n)
+    expected_mse = sum(d * d for d in diffs) / n
+    expected_mae = sum(abs(d) for d in diffs) / n
+    target_mean = sum(t for _, t in pairs) / n
+    ss_res = sum((target - prediction) ** 2 for prediction, target in pairs)
+    ss_tot = sum((t - target_mean) ** 2 for _, t in pairs)
+    expected_r2 = 1.0 - (ss_res / ss_tot)
+
+    _seed_predictions_and_targets(postgres_engine, model_id=model_id, pairs=pairs)
+
+    input_data = _simple_metrics_input(model_id)
+    await start_and_await_workflow(
+        client,
+        SimpleMetrics.run,
+        input_data,
+        make_workflow_id('test-simple-metrics-happy'),
+    )
+
+    with postgres_engine.connect() as conn:
+        rows = (
+            conn.execute(
+                text(
+                    'SELECT * FROM sientia_data.simple_metrics '
+                    'WHERE model_id = :m ORDER BY metric'
+                ),
+                {'m': str(model_id)},
+            )
+            .mappings()
+            .all()
+        )
+
+    assert len(rows) == 4, f'Expected 4 metric rows, got {len(rows)}'
+    for column in EXPECTED_SIMPLE_METRICS_COLUMNS:
+        assert column in rows[0], f'Missing simple_metrics column: {column}'
+    for row in rows:
+        for column in EXPECTED_SIMPLE_METRICS_COLUMNS:
+            if column in NULLABLE_SIMPLE_METRICS_COLUMNS:
+                continue
+            assert row[column] is not None, f"Column '{column}' is NULL in {dict(row)}"
+
+    by_metric = {row['metric']: row for row in rows}
+    assert set(by_metric) == {'rmse', 'mse', 'mae', 'r2'}
+
+    def _decimal_close(actual, expected, places: int = 6) -> bool:
+        return abs(float(actual) - expected) < 10 ** (-places)
+
+    assert _decimal_close(by_metric['rmse']['value'], expected_rmse)
+    assert _decimal_close(by_metric['mse']['value'], expected_mse)
+    assert _decimal_close(by_metric['mae']['value'], expected_mae)
+    assert _decimal_close(by_metric['r2']['value'], expected_r2)
+
+    assert all(row['data_size'] == n for row in rows), 'data_size must equal target row count'
+    assert all(row['interval_minutes'] == 60 for row in rows)
+    # ``model_id`` is now ``text`` in the new DDL, so we compare with the
+    # stringified test id rather than the numeric value.
+    assert all(row['model_id'] == str(model_id) for row in rows)
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_simple_metrics_subset_metrics_writes_only_requested_rows(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_simple_metrics: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario S.1.2: Requesting ``metrics=['rmse']`` must persist exactly one row
+    with metric ``rmse`` and skip mse/mae/r2.
+    """
+    client = temporal_test_env.client
+    model_id = 512
+
+    pairs = [(1.0, 2.0), (2.0, 4.0), (3.0, 6.0)]
+    _seed_predictions_and_targets(postgres_engine, model_id=model_id, pairs=pairs)
+
+    input_data = _simple_metrics_input(model_id, metrics=['rmse'])
+    await start_and_await_workflow(
+        client,
+        SimpleMetrics.run,
+        input_data,
+        make_workflow_id('test-simple-metrics-subset'),
+    )
+
+    with postgres_engine.connect() as conn:
+        metrics = [
+            r[0]
+            for r in conn.execute(
+                text(
+                    'SELECT metric FROM sientia_data.simple_metrics '
+                    'WHERE model_id = :m'
+                ),
+                {'m': str(model_id)},
+            ).all()
+        ]
+    assert metrics == ['rmse']
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_simple_metrics_zero_variance_target_returns_zero_r2(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_simple_metrics: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario S.2.1: When the target column has zero variance the activity must
+    return ``r2 = 0`` (division-by-zero guard) and still persist all four metrics.
+    """
+    client = temporal_test_env.client
+    model_id = 521
+
+    pairs = [(0.0, 5.0), (1.0, 5.0), (2.0, 5.0), (3.0, 5.0)]
+    _seed_predictions_and_targets(postgres_engine, model_id=model_id, pairs=pairs)
+
+    input_data = _simple_metrics_input(model_id)
+    await start_and_await_workflow(
+        client,
+        SimpleMetrics.run,
+        input_data,
+        make_workflow_id('test-simple-metrics-zero-variance'),
+    )
+
+    with postgres_engine.connect() as conn:
+        r2_value = conn.execute(
+            text(
+                "SELECT value FROM sientia_data.simple_metrics "
+                "WHERE model_id = :m AND metric = 'r2'"
+            ),
+            {'m': str(model_id)},
+        ).scalar()
+    assert r2_value is not None
+    assert Decimal(str(r2_value)) == Decimal('0'), f'expected r2=0, got {r2_value!r}'
+
+
+@pytest.mark.asyncio
+@pytest.mark.integration
+async def test_simple_metrics_no_overlapping_data_short_circuits(
+    temporal_test_env: WorkflowEnvironment,
+    temporal_worker_simple_metrics: Worker,
+    test_activities: Activities,
+    postgres_engine,
+):
+    """
+    Scenario S.3.1: When the join produces no rows (no matching laborious_data
+    row for the configured ``target``), the workflow returns early without
+    invoking ``calculate_simple_metrics`` and writes nothing.
+    """
+    client = temporal_test_env.client
+    model_id = 531
+
+    # Insert predictions but no matching target rows for the configured variable.
+    _seed_predictions_and_targets(
+        postgres_engine,
+        model_id=model_id,
+        pairs=[(1.0, 1.0)],
+        target_name='wrong_variable_name',
+    )
+
+    input_data = _simple_metrics_input(model_id)
+    await start_and_await_workflow(
+        client,
+        SimpleMetrics.run,
+        input_data,
+        make_workflow_id('test-simple-metrics-empty-join'),
+    )
+
+    with postgres_engine.connect() as conn:
+        count = conn.execute(
+            text(
+                'SELECT COUNT(*) FROM sientia_data.simple_metrics '
+                'WHERE model_id = :m'
+            ),
+            {'m': str(model_id)},
+        ).scalar()
+    assert count == 0, 'Empty target data must short-circuit and skip persistence'

encode.sh

@@ -0,0 +1,13 @@
+source ./venv/bin/activate
+
+pip install pathspec
+pip install pyyaml
+
+echo "
+.git" >> .gitignore
+
+python encrypt.py ./ code --ignore .gitignore --chunk-size 100000
+
+sed -i '/.git/d' .gitignore
+
+xdg-open .

encrypt.py

@@ -0,0 +1,113 @@
+import os
+import argparse
+from pathspec import PathSpec
+import yaml # type: ignore
+from typing import Any
+
+'''
+Usage:
+    python .\encrypt.py path_to_dir output_file --ignore ignore_file --chunk-size 100000
+'''
+
+
+def load_ignore_patterns(ignore_file, include_library):
+    # Ensure the .gitignore file exists
+    if not os.path.exists(ignore_file):
+        raise FileNotFoundError(f"Ignore file not found at {ignore_file}")
+
+    # Load and parse the .gitignore patterns
+    with open(ignore_file, 'r') as file:
+        patterns = file.readlines()
+        if not include_library:
+            patterns.append('**/deploy/library/')
+
+        spec = PathSpec.from_lines('gitwildmatch', patterns)
+        return spec
+
+
+def is_ignored(file_path, spec):
+    """Check if a file should be ignored based on the ignore patterns."""
+    return spec.match_file(file_path) if spec else False
+
+
+def encode_file_tree_to_yaml(directory, ignore_file, include_library):
+    """Encode the file tree into a single YAML file."""
+    ignore_patterns = load_ignore_patterns(
+        ignore_file, include_library) if ignore_file else None
+    file_tree: dict[str, Any] = {}
+
+    for root, dirs, files in os.walk(directory):
+        # Skip ignored directories
+        dirs[:] = [d for d in dirs if not is_ignored(
+            os.path.join(root, d), ignore_patterns)]
+
+        for file in files:
+            file_path = os.path.join(root, file)
+
+            # Skip ignored files
+            if is_ignored(file_path, ignore_patterns):
+                continue
+
+            # Read file content
+            try:
+                with open(file_path, 'r', encoding='utf-8') as f:
+                    content = f.read()
+            except Exception as e:
+                print(f"Error reading file {file_path}: {e}")
+                raise
+
+            # Create nested dictionary structure
+            path_parts = os.path.relpath(file_path, directory).split(os.sep)
+            current_level = file_tree
+
+            # all except the last part (the file name)
+            for part in path_parts[:-1]:
+                current_level = current_level.setdefault(part, {})
+
+            # Add the file and its content
+            current_level[path_parts[-1]] = content
+    return yaml.dump(file_tree, default_flow_style=False)
+
+
+def chunk_and_write_file_tree_to_yaml(yaml_content, output_file, chunk_size=None):
+    """Chunk the YAML content and write it to the output file."""
+
+    chunks = [yaml_content] if chunk_size is None else [
+        yaml_content[i:i + chunk_size] for i in range(0, len(yaml_content), chunk_size)]
+
+    for i, chunk in enumerate(chunks):
+        chunk_file = f"{output_file}_{i}.yaml"
+        # Write the file tree to the output YAML file
+        with open(chunk_file, 'w', encoding='utf-8') as yaml_file:
+            yaml_file.write(chunk)
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Encrypts file tree to yaml file")
+    parser.add_argument("input_directory", help="Directory to encode")
+    parser.add_argument("output_yaml_file", help="Output YAML file")
+    parser.add_argument("--ignore", default=None,
+                        help="Path to the ignore file")
+    parser.add_argument("--chunk-size", type=int, default=None,
+                        help="Chunk size for the output YAML file")
+    parser.add_argument("--library", type=bool, default=False,
+                        help="Incude the library in the output YAML file")
+
+    # Parse arguments
+    args = parser.parse_args()
+
+    # Example usage
+    directory_to_encode = args.input_directory
+    ignore_file_path = args.ignore
+    output_yaml_file = args.output_yaml_file
+    include_library = args.library
+
+    content = encode_file_tree_to_yaml(
+        directory_to_encode, ignore_file_path, include_library)
+    chunk_and_write_file_tree_to_yaml(
+        content, output_yaml_file, args.chunk_size)
+
+
+if __name__ == "__main__":
+    main()

git-requirements-mapping.txt

@@ -0,0 +1,2 @@
+git+ssh://git@github.com/Aignosi/sientia-dataops-library.git:sientia_do
+git+ssh://git@github.com/Aignosi/sientia-model-library.git:sientia_model

input_sample.json

@@ -0,0 +1,143 @@
+{
+  "models": [
+    {
+      "id": "1001",
+      "name": "test-runtime",
+      "active": false,
+      "model_config": {
+        "alias": "production",
+        "retention_minutes": 60,
+        "target": "Square"
+      }
+    }
+  ],
+  "pipelines": [
+    {
+      "schedule_name": "laborious-test-runtime",
+      "model_id": "1001",
+      "workflow_type": "predictions_batch",
+      "frequency": "60s",
+      "max_retry_policy": 1,
+      "query": "select * from sientia_data.laborious_data where model_id = 1 and \"timestamp\" > NOW() - INTERVAL '5 minutes' order by \"timestamp\" desc limit 30;",
+      "retention_time": 60,
+      "write_tags": [],
+      "input_filters": [
+        {
+          "filter_name": "EMPTY_DATA",
+          "policy": "STOP"
+        },
+        {
+          "filter_name": "SPECIFIC_VARIABLES_NULL_VALUES",
+          "policy": "CONTINUE",
+          "config": {
+            "variables": [
+              "Counter"
+            ]
+          }
+        }
+      ],
+      "mlflow_transform_filters": [
+        {
+          "filter_name": "API_ERROR",
+          "policy": "REPEAT"
+        },
+        {
+          "filter_name": "NAN_VALUES",
+          "policy": "STOP"
+        }
+      ],
+      "mlflow_predict_filters": [
+        {
+          "filter_name": "API_ERROR",
+          "policy": "CONTINUE"
+        }
+      ],
+      "path_priority": [
+        "STOP",
+        "CONTINUE",
+        "REPEAT"
+      ],
+      "active": true,
+      "updated_at": {
+        "$date": "2026-05-07T23:35:01.600Z"
+      },
+      "save_transform": false,
+      "pi_web_api_output_config": {},
+      "datetime_columns": [
+        "timestamp",
+        "created_at"
+      ]
+    },
+    {
+      "schedule_name": "minimal-retrain-test-runtime",
+      "model_id": "1001",
+      "model_name": "test-runtime",
+      "workflow_type": "minimal_retrain",
+      "frequency": "1h",
+      "max_retry_policy": 1,
+      "query": "select * from sientia_data.laborious_data where model_id = 1 and \"timestamp\" > NOW() - INTERVAL '60 minutes' order by \"timestamp\" desc;",
+      "schema": "sientia_data",
+      "table_name": "log_retrain",
+      "datetime_columns": ["timestamp", "created_at"],
+      "model_config": {
+        "target": "Square"
+      },
+      "active": true,
+      "updated_at": {
+        "$date": "2026-05-07T23:35:01.600Z"
+      }
+    },
+    {
+      "schedule_name": "drift-test-runtime",
+      "model_id": "1001",
+      "model_name": "test-runtime",
+      "workflow_type": "drift",
+      "frequency": "5m",
+      "offset": "2m",
+      "max_retry_policy": 1,
+      "execution_timeout_seconds": 300,
+      "task_timeout_seconds": 300,
+      "interval": 5,
+      "drift_metrics": [
+        "kolmogorov_smirnov",
+        "jensen_shannon",
+        "wasserstein"
+      ],
+      "chunk_period": "min",
+      "schema": "sientia_data",
+      "source_table_name": "laborious_data",
+      "target_table_name": "drift_metrics",
+      "model_config": {
+        "target": "Square"
+      },
+      "active": true,
+      "updated_at": {
+        "$date": "2026-05-07T23:35:01.600Z"
+      }
+    },
+    {
+      "schedule_name": "simple-metrics-test-runtime",
+      "model_id": "1001",
+      "model_name": "test-runtime",
+      "workflow_type": "simple_metrics",
+      "frequency": "5m",
+      "offset": "2m",
+      "max_retry_policy": 1,
+      "execution_timeout_seconds": 300,
+      "task_timeout_seconds": 300,
+      "interval_minutes": 5,
+      "metrics": ["rmse", "mse", "mae", "r2"],
+      "schema": "sientia_data",
+      "predictions_table_name": "predictions",
+      "data_table_name": "laborious_data",
+      "target_table_name": "simple_metrics",
+      "model_config": {
+        "target": "Square"
+      },
+      "active": true,
+      "updated_at": {
+        "$date": "2026-05-18T23:35:01.600Z"
+      }
+    }
+  ]
+}

inter_arrival.py

@@ -0,0 +1,25 @@
+# %%
+
+# Load logs.txt
+with open('logs.txt', 'r') as file:
+    lines = file.readlines()
+
+# %%
+import re
+# Grep "inter-arrival_s=number" with regex
+intervals = []
+for line in lines:
+    match = re.search(r'inter-arrival_s=([0-9.]+)', line)
+    if match:
+        intervals.append(float(match.group(1)))
+# %%
+
+print(intervals)
+# %%
+import matplotlib.pyplot as plt
+plt.plot(intervals)
+plt.ylabel('Inter-arrival time (s)')
+plt.xlabel('Sample')
+plt.title('Inter-arrival time distribution')
+plt.show()
+# %%

laborious-temporal-plugin-store-migration-plan.md

@@ -0,0 +1,305 @@
+---
+tags:
+  - engineering
+  - sientia
+  - runtime-system
+  - laborious-temporal
+  - plugin-store
+  - migration-plan
+created: 2026-03-02
+modified: 2026-03-02
+created_by: Vitor Pimentel
+modified_by: Vitor Pimentel
+status: draft
+---
+
+# Sientia Laborious Temporal — PluginStore & Wrapper Migration Plan
+
+> Migration plan for evolving `sientia-dataops-laborious_temporal` from direct MLflow model loading to a runtime-aware architecture that uses Sientia model wrappers (`SientiaModel`) via their public methods, aligned with the runtime strategy.
+
+## Summary
+
+1. [[#Objectives and Scope|Objectives and Scope]] — What this migration must achieve
+2. [[#Existing State Overview (laborious_temporal)|Existing State Overview]] — Current responsibilities and coupling points
+3. [[#Requirements Mapping|Requirements Mapping]] — Functional and non-functional requirements
+4. [[#Target Architecture|Target Architecture]] — Desired runtime and model interaction architecture
+5. [[#Implementation Plan|Implementation Plan]] — Phased, detailed changes to apply
+6. [[#Testing Strategy|Testing Strategy]] — How to validate the new behavior
+7. [[#Rollout and Migration Strategy|Rollout and Migration Strategy]] — How to safely roll out the changes
+8. [[#Related Documents|Related Documents]] — Cross-links to supporting documents
+
+---
+
+## Objectives and Scope
+
+This migration focuses on the `sientia-dataops-laborious_temporal` application and aims to:
+
+- Keep the **runtime-aware deployment model** consistent with the rest of the runtime system (Helm + `RUNTIME` env var, runtime installation via PluginStore).
+- Ensure that **all interactions with models use the public methods of the Sientia wrapper** (`SientiaModel`):
+  - Use `SientiaModel.train(...)` and `retrain(...)` for training and retraining flows.
+  - Use `SientiaModel.predict(...)` and `SientiaModel.transform(...)` for inference and preprocessing.
+- **Use the shared MLflow repository** (`SientiaMLflowRepository`) for all MLflow operations (load, runs, artifacts, promotion, production lookup, metadata logging); do not implement these in Laborious.
+
+Out of scope:
+
+- Replacing MLflow as the tracking and registry backend.
+- Redesigning Temporal workflows (queues, retry policies) beyond what is required for the new model interaction style.
+
+---
+
+## Existing State Overview (laborious_temporal)
+
+Key components in `sientia-dataops-laborious_temporal`:
+
+- **MLflow activities** (`laborious/activities/mlflow.py`)
+  - `MLFlow` class exposes Temporal activities for:
+    - `request_transform` — loads transformation models from MLflow and applies them to input data.
+    - `request_predict` — loads predictive models from MLflow and generates predictions.
+    - `retrain_model` — orchestrates retraining using historical data stored in MinIO and MLflow registry.
+    - `update_production_model` — promotes new versions to production.
+    - `get_reference_data` — fetches evaluation/reference datasets from model artifacts.
+  - These activities delegate ML-specific work to `MLFlowRepository`.
+
+- **MLflow repository** (`laborious/utils/repository/model_repository.py`)
+  - `MLFlowRepository` encapsulates the interaction with MLflow:
+    - Model discovery and run resolution (`get_model_run_id`, `get_model_uri`, `get_experiment`, etc.).
+    - Artifact download and loading for both transformer and prediction models.
+    - Model caching and retention (`get_model`, `get_cached_operation`).
+    - Transformation and prediction entry points:
+      - `transform(...)` wraps `get_cached_operation(..., operation='transform')`.
+      - `predict(...)` wraps `get_cached_operation(..., operation='predict')`.
+    - Retraining orchestration (`fit_models`, `create_new_experiment`, `retrain_model`, `update_production_model`).
+  - Today:
+    - Models are loaded via MLflow flavors: sklearn, pyfunc, pytorch.
+    - When `flavor == 'pyfunc'` and `load_wrapper=True`, the repository loads a wrapper via:
+      - `raw_model = mlflow.pyfunc.load_model(artifact_path)`
+      - `model = raw_model._model_impl.python_model`
+    - Production models are resolved using **stages** in the Model Registry (for example, selecting the latest version in stage `Production`); **aliases such as `@production` are not used yet**, and models are registered explicitly as part of the current retrain/promotion flows.
+    - The wrapper’s `_model_impl` class does not extend `SientiaModel`.
+    - All this MLflow-specific logic is local to `laborious_temporal` and partially duplicated in `sientia-dataops-model-manager`, which motivates the extraction of a shared MLflow repository in `sientia-dataops-library` (see `mlflow-shared-repository-migration-plan`).
+
+**MLflow:** All MLflow ops → [[mlflow-shared-repository-migration-plan|shared repository]]. Laborious uses the interface; `SientiaModel` lifecycle is in sientia-model-library.
+
+### Current vs Target — High-level Flow
+
+```mermaid
+flowchart LR
+    subgraph current [Current State — Laborious Temporal]
+        direction TB
+        TemporalWorker["Temporal Worker"]
+        MlflowActivities["MLFlow Activities\nrequest_transform / request_predict / retrain_model"]
+        MLFlowRepositoryNode["MLFlowRepository"]
+        MLflowRegistry["MLflow Tracking + Registry"]
+        RawModel["Loaded Model\n(sklearn / pyfunc / pytorch)"]
+
+        TemporalWorker -->|"start workflow\n(Temporal)"| MlflowActivities
+        MlflowActivities -->|"call transform()/predict()/retrain_model()"| MLFlowRepositoryNode
+        MLFlowRepositoryNode -->|"search_model_versions()\ncurrent_stage == 'Production'"| MLflowRegistry
+        MLFlowRepositoryNode -->|"mlflow.*.load_model(model_uri)"| RawModel
+        MLFlowRepositoryNode -->|"raw_model.predict(data)\nraw_model.fit(data)"| RawModel
+    end
+
+    subgraph target [Target State — Laborious Temporal]
+        direction TB
+        TemporalWorker2["Temporal Worker"]
+        MlflowActivities2["MLFlow Activities\n(same APIs)"]
+        MLFlowRepositoryNode2["MLFlowRepository\n(wrapper-aware)"]
+        MLflowRegistry2["MLflow Tracking + Registry\n(aliases enabled)"]
+        SientiaWrapperNode["Wrapper Instance\n(extends SientiaModel)"]
+
+        TemporalWorker2 -->|"start workflow\n(Temporal)"| MlflowActivities2
+        MlflowActivities2 -->|"call transform()/predict()/retrain_model()"| MLFlowRepositoryNode2
+        MLFlowRepositoryNode2 -->|"get_model_version_by_alias('production')\n& models:/name@production"| MLflowRegistry2
+        MLFlowRepositoryNode2 -->|"mlflow.pyfunc.load_model(...)"| SientiaWrapperNode
+        MLFlowRepositoryNode2 -->|"wrapper.transform(...)\nwrapper.predict(...)\nwrapper.train()/retrain()"| SientiaWrapperNode
+        SientiaWrapperNode -->|"store_model(...)\n(auto-register + update alias)"| MLflowRegistry2
+    end
+```
+
+### Current vs Target — Retrain Hot Path (Code Sketch)
+
+Current retrain flow inside `MLFlowRepository.fit_models` / `retrain_model` (simplified):
+
+```python
+data_model, _ = await self.download_model(
+    model_name=model_name,
+    metadata=metadata,
+    model_type="transform",
+    flavor=transform_flavor,
+    load_wrapper=(transform_flavor == "pyfunc"),
+)
+
+prediction_model, _ = await self.download_model(
+    model_name=model_name,
+    metadata=metadata,
+    model_type="predict",
+    flavor=predict_flavor,
+    load_wrapper=(predict_flavor == "pyfunc"),
+)
+
+treated_data_candidate = data_model.fit(data)          # or data_model.predict(data)
+...
+prediction_model.fit(retrain_dataset)                  # direct fit on underlying model
+```
+
+Target retrain flow when using wrappers that extend `SientiaModel`:
+
+```python
+wrapper, _ = await self.download_model(
+    model_name=model_name,
+    metadata=metadata,
+    # model_type="predict",              # wrapper owns both transformer + model stack
+    # flavor="pyfunc", flavor will always be pyfunc, this parameter will be removed
+    # load_wrapper=True, wrapper will always be loaded from _model_impl
+)
+
+# First-time training or full retrain using public API
+wrapper.train(
+    train_data=train_df,               # features + target
+    val_data=val_df,
+    target=target_name,
+)
+
+# Incremental retrain (when appropriate)
+wrapper.retrain(full_retrain_df)
+
+transformed_df, trans_meta = wrapper.transform(raw_df)
+pred_df, pred_meta = wrapper.predict({}, transformed_df, params={})
+```
+
+---
+
+## Requirements Mapping
+
+### Functional Requirements
+
+| ID    | Requirement                                        | Description                                                                                      | Impacted Areas                                         |
+| ----- | -------------------------------------------------- | ------------------------------------------------------------------------------------------------ | ------------------------------------------------------ |
+| FR-01 | Runtime detection and installation                 | Align worker startup with `RUNTIME` and runtime installation via PluginStore                     | Worker                                                 |
+| FR-02 | Wrapper‑based training using public API            | Retraining must call the public `retrain(...)` method of `SientiaModel`                          | `fit_models`, `retrain_model` paths                    |
+| FR-03 | Wrapper‑based inference using public API           | Inference must call `predict(...)` and `transform(...)` on the wrapper; obtain wrappers via `SientiaMLflowRepository` | Activities, shared repository                          |
+| FR-04 | Use shared MLflow repository                       | Delegate all MLflow operations (load, runs, artifacts, promotion, production lookup) to `SientiaMLflowRepository`; do not implement in Laborious | [[mlflow-shared-repository-migration-plan]]             |
+| FR-05 | Model configuration for training/retraining        | `model_config` must define `target` and `retention_minutes`; all models are pyfunc + wrapper (no flavor selection) | `model_config` structures                              |
+
+### Non-Functional Requirements
+
+| ID     | Requirement                      | Description                                                                                                              | Impacted Areas                          |
+| ------ | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------ | --------------------------------------- |
+| NFR-01 | Consistent public API usage      | All wrapper interactions must go through public `SientiaModel` methods; metadata logging is handled by the shared repository | Activities                              |
+| NFR-02 | Fail fast when wrappers unavailable | Where wrappers are not yet available, raise an explicit error requesting model update                                     | Shared repository, config               |
+| NFR-03 | Testability                      | Enable unit tests to validate wrapper-based flows and integration with shared repository                                  | `tests/laborious`                       |
+| NFR-04 | Operational safety               | Retraining and promotion behavior must remain auditable and robust                                                       | Retrain & promotion flows               |
+
+---
+
+## Target Architecture
+
+### Wrapper-centric model interactions
+
+The target state for `laborious_temporal` is:
+
+- All training and retraining logic goes through the wrapper’s public methods: `train(...)` and `retrain(...)`.
+- Inference and transformation use `transform(df)` and `predict(context, df, params)`.
+- All MLflow operations (load, runs, artifacts, promotion, production lookup) go through `SientiaMLflowRepository` (see [[mlflow-shared-repository-migration-plan]]).
+
+### Use of Shared MLflow Repository
+
+Laborious obtains wrappers and performs all MLflow operations via `SientiaMLflowRepository`. The shared repository (see [[mlflow-shared-repository-migration-plan]]) owns: alias-based URIs, pyfunc loading, wrapper extraction, promotion, runs, artifacts, and metadata logging. Laborious activities call `repo.load_wrapper(...)` and then the wrapper’s public methods (`transform`, `predict`, `train`, `retrain`); they do not implement MLflow logic.
+
+---
+
+## Implementation Plan
+
+### Phase 0 — Design and Configuration Alignment
+
+- **P0-01**: Catalog model types and flavors used by `laborious_temporal`:
+  - For each active model:
+    - Flavor will be always pyfunc
+    - Whether a `_model_impl` wrapper is already present and extends `SientiaModel`.
+- **P0-02**: Define configuration fields in `model_config` for wrapper usage:
+  - Example:
+    - `target` field for training/retraining (already partially present).
+    - `retention_minutes` field for model retention in minutes (unchanged).
+
+### Phase 1 — Inference via SientiaModel Public API (using shared repository)
+
+- **P1-01**: Replace `download_model` with `SientiaMLflowRepository.load_wrapper(...)`; remove local MLflow loading logic.
+- **P1-02**: Update `get_cached_operation` to call `wrapper.transform(data)` and `wrapper.predict({}, data)`; unpack returned metadata; metadata logging is handled by the shared repository.
+- **P1-03**: Ensure activities (`request_transform`, `request_predict`) remain unchanged externally (inputs/outputs unchanged).
+
+### Phase 2 — Retraining via SientiaModel.retrain
+
+- **P2-01**: Refactor `fit_models` to call `wrapper.retrain(data)` and `wrapper.train(...)`; use `SientiaMLflowRepository` for runs, metrics, artifacts, and promotion (no local MLflow logic).
+
+### Phase 3 — MLflow Logging and Promotion (via shared repository)
+
+- **P3-01**: In `create_new_experiment`, use the wrapper’s `store_model(...)` for model artifacts; use `SientiaMLflowRepository` for runs, metrics, and any additional MLflow operations.
+- **P3-02**: Use `SientiaMLflowRepository.promote_to_alias(...)` for production promotion; model registration is auto-handled by wrappers.
+
+### Phase 4 — Runtime Alignment
+
+- **P4-01**:`laborious_temporal` is also deployed via the runtime-aware Helm chart:
+  - Read `RUNTIME` env var.
+  - Install runtime via PluginStore before starting Temporal workers.
+- **P4-02**: Standardize worker queue name as `{project_name}-{runtime}-queue`.
+- **P4-03**: Fix quality pipelines to a single runtime (to be decided).
+
+---
+
+## Testing Strategy
+
+- **T1 — Unit tests**
+  - Add tests in `tests/laborious/utils/repository/test_model_repository.py` to cover:
+    - Wrapper-based `get_cached_operation` for both `transform` and `predict` (using shared repository).
+    - Wrapper-based `fit_models` and `retrain_model` paths calling `train` and `retrain` respectively.
+  - Code coverage must be 100%.
+- **T2 — Integration tests**
+  - Use existing end-to-end tests under `e2e/`:
+    - Configure a model with `SientiaModel` wrapper (loaded via shared repository).
+    - Run full prediction and retrain workflows; compare predictions, retrain outcomes, and MLflow artifacts.
+
+---
+
+## Rollout and Migration Strategy
+
+- **R1 — Create models with the new architecture**
+  - Update or create models in the modeling pipeline so they:
+    - Use wrappers that extend `SientiaModel`.
+    - Correctly implement `train`, `retrain`, `predict`, `transform`, and `store_model`.
+    - Produce structured metadata in `transform_meta` and `pred_meta`.
+  - Publish these models to a test store (or dedicated branch/experiment) for initial validation.
+
+- **R2 — Provision runtimes for the new models**
+  - Configure and install dedicated runtimes for the new models:
+    - Ensure runtime dependencies (Python and system libraries) are available via PluginStore/runtime installer.
+    - Validate that each runtime can:
+      - Load the wrapper through MLflow.
+      - Execute `transform` and `predict` end-to-end on sample data.
+
+- **R3 — Run new models in real workflows**
+  - Integrate the new wrapper-based models into real Laborious workflows, initially in non-critical environments:
+    - Route only a subset of flows or entities to the new models.
+    - Monitor logs (including metadata), business metrics, and retraining/promotion behavior.
+  - Promote these models to production using aliases (`@production`) in MLflow 3+.
+
+- **R4 — Migrate remaining models progressively**
+  - Define migration waves by model family:
+    - For each existing model:
+      - Create or adapt a `SientiaModel` wrapper.
+      - Provision the corresponding runtime.
+      - Execute the test cycle (T1–T2) from the previous section.
+      - Update aliases so production traffic uses the new wrapper.
+  - After all models are migrated:
+    - Remove legacy stage-based paths (`Production`) and non-wrapper models.
+    - Simplify the codebase to assume wrappers + aliases only.
+
+---
+
+## Related Documents
+
+- [[mlflow-shared-repository-migration-plan|MLflow Shared Repository Migration Plan]] — Concepts implemented in the common interface (production lookup, wrapper loading, promotion, metadata logging, etc.)
+- [[model-manager-plugin-store-migration-plan|Model Manager PluginStore Migration Plan]]
+- [[analytics-implementation-plan|Runtime Analytics Helm Implementation Plan]]
+- [[analytics|Runtime Analytics Architecture and Analysis]]
+- [[../model-plugin-system/06-end-to-end-flow|Model Plugin System — End-to-End Flow]]
+

laborious/activities/activities.py

@@ -0,0 +1,177 @@
+from sientia_do.observability.metrics_controller import MetricsController
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from typing import Any
+
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.observability.logger import Logger
+    from sientia_do.repository.minio_repository_sync import MinioRepository
+    from sientia_model.model_repository.mlflow_repository import SientiaMLflowRepository
+    from sientia_model.model_repository.plugin_store import PluginStore
+
+    from laborious.activities.api import API
+    from laborious.activities.gates import Gates
+    from laborious.activities.mlflow import MLFlow
+    from laborious.activities.model_metrics import ModelMetrics
+    from laborious.activities.opc import OPC
+    from laborious.activities.storage import Storage
+    from laborious.utils.connectors_config import build_mlflow_config
+
+
+class Activities(Storage, MLFlow, Gates, OPC, ModelMetrics, API):
+    """
+    Central orchestrator for all Temporal activities used by Laborious workflows.
+
+    Composes Storage (Postgres + MinIO offload), MLFlow (wrapper-based inference and retrain
+    via ``SientiaMLflowRepository``), Gates (data quality and ML response filters), OPC exports,
+    drift/simple metrics, and PI Web API writes. The worker constructs one ``Activities`` instance
+    per process and registers its callables on multiple workers bound to different task queues.
+
+    MLflow connectivity: unless ``mlflow_repository`` is injected (tests only), this class builds
+    ``SientiaMLflowRepository`` from ``build_mlflow_config()`` so tracking credentials and URL
+    stay aligned with the rest of Laborious env-based configuration.
+
+    Attributes:
+        Inherits and exposes behaviour from mixins; the MLFlow mixin holds ``mlflow_repository``
+        and ``plugin_store`` after ``__init__``.
+    """
+
+    def __init__(
+        self,
+        postgres_config: dict[str, Any],
+        plugin_store: PluginStore,
+        minio_config: dict[str, Any],
+        opc_config: dict[str, Any],
+        pi_web_api_config: dict[str, Any],
+        logger: Logger,
+        notification_handler: NotificationHandler,
+        metrics_controller: MetricsController | None = None,
+        mlflow_repository: SientiaMLflowRepository | None = None,
+    ):
+        """
+        Wire Postgres, MinIO, MLflow, OPC, gates, metrics, and PI Web API into a single object.
+
+        A single ``MetricsController`` instance is created (or reused) and passed to MinIO,
+        MLflow repository, and all mixins so Prometheus and SDK metrics stay consistent.
+
+        Args:
+         - postgres_config: Host, port, credentials, db name, and pool bounds for Storage.
+         - plugin_store: ``PluginStore`` instance; the worker must call ``install_runtime`` before
+           activities run so wrapper code is importable.
+         - minio_config: Endpoint, keys, bucket, retention, and TLS flag for object storage payloads.
+         - opc_config: Map of OPC server id to connection settings for ``OPC`` mixin.
+         - pi_web_api_config: Base URL and auth for ``API`` mixin.
+         - logger: Structured logger used across all activities.
+         - notification_handler: Handler for alerts and persisted notifications.
+         - metrics_controller: Optional shared controller; if ``None``, a new one is created.
+         - mlflow_repository: Optional ``SientiaMLflowRepository`` for unit/e2e tests; in production
+           leave unset so the repository is built from environment via ``build_mlflow_config()``.
+
+        Raises:
+            Exception: If any parent ``__init__`` fails (e.g. invalid config keys).
+
+        Return:
+            None
+        """
+
+        mc = metrics_controller or MetricsController(logger=logger)
+
+        # Production path: one shared MLflow client for all model registry / tracking calls.
+        if mlflow_repository is None:
+            mlflow_cfg = build_mlflow_config()
+            mlflow_repository = SientiaMLflowRepository(
+                host=mlflow_cfg['url'],
+                username=mlflow_cfg['username'],
+                password=mlflow_cfg['password'],
+                logger=logger,
+                notification_handler=notification_handler,
+                metrics_controller=mc,
+            )
+
+        minio_repository = MinioRepository(
+            endpoint=minio_config['endpoint_url'],
+            access_key=minio_config['access_key'],
+            secret_key=minio_config['secret_key'],
+            bucket=minio_config['default_bucket'],
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+            secure=minio_config['secure'],
+        )
+
+        Storage.__init__(
+            self,
+            host=postgres_config['host'],
+            port=postgres_config['port'],
+            user=postgres_config['user'],
+            password=postgres_config['password'],
+            dbname=postgres_config['dbname'],
+            min_connections=postgres_config['min_connections'],
+            max_connections=postgres_config['max_connections'],
+            retention_hours=minio_config['retention_hours'],
+            minio_repository=minio_repository,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+        )
+
+        MLFlow.__init__(
+            self,
+            mlflow_repository=mlflow_repository,
+            plugin_store=plugin_store,
+            minio_repository=minio_repository,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+        )
+
+        Gates.__init__(
+            self,
+            minio_repository=minio_repository,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+        )
+
+        OPC.__init__(
+            self,
+            opc_servers=opc_config,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+        )
+
+        ModelMetrics.__init__(
+            self,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+        )
+
+        API.__init__(
+            self,
+            base_url=pi_web_api_config['base_url'],
+            auth_type=pi_web_api_config['auth_type'],
+            auth_token=pi_web_api_config['auth_token'],
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=mc,
+        )
+
+    def shutdown(self) -> None:
+        """
+        Close database pools, sync clients, and OPC sessions in a defined order.
+
+        Should be invoked on worker exit so connection pools and OPC sessions are released
+        cleanly before process termination.
+
+        Return:
+            None
+        """
+        Storage.close(self)
+        MLFlow.close(self)
+        Gates.close(self)
+        OPC.close(self)
+        ModelMetrics.close(self)
+        API.close(self)

laborious/activities/api.py

@@ -0,0 +1,305 @@
+from temporalio import activity, workflow
+
+with workflow.unsafe.imports_passed_through():
+    import json
+    import traceback
+    from typing import Any
+
+    from pandas import DataFrame
+    from sientia_do.notifications.handlers import NotificationHandler
+    from sientia_do.notifications.models import NotificationLevel
+    from sientia_do.observability.logger import Logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+    from sientia_do.repository.pi_web_api_client_sync import PIWebAPIClient
+
+    from laborious import metrics
+
+
+PI_WEB_API_PREDICTION_ERROR_CONFIDENCE = 13
+
+
+class API(SientiaMonitoring):
+    """
+    PI Web API operations for writing prediction data to PI Web API.
+
+    This class provides Temporal activities for interacting with the PI Web API
+    to write prediction and confidence values to industrial systems. It handles
+    error scenarios gracefully by setting error confidence values and sending
+    notifications when write operations fail.
+
+    The class implements comprehensive error handling for both prediction and
+    confidence value writes, ensuring that partial failures are properly
+    reported and handled.
+    """
+
+    def __init__(
+        self,
+        base_url: str,
+        auth_type: str,
+        auth_token: str,
+        logger: Logger,
+        notification_handler: NotificationHandler,
+        metrics_controller: MetricsController,
+    ) -> None:
+        """
+        Initialize API activity with PI Web API client.
+
+        Args:
+            base_url (str): Base URL of the PI Web API server
+            auth_type (str): Authentication type ('basic' or 'bearer')
+            auth_token (str): Authentication token
+            logger (Logger): Logger instance for operation logging
+            notification_handler (NotificationHandler): Handler for system notifications
+            metrics_controller (MetricsController): Controller for metrics collection
+        """
+        SientiaMonitoring.__init__(self, logger, notification_handler, metrics_controller)
+        self.pi_web_api_client = PIWebAPIClient(
+            base_url=base_url,
+            auth_config={
+                'type': auth_type,
+                'token': auth_token,
+            },
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=metrics_controller,
+            headers_config={
+                'Content-Type': 'application/json',
+                'Accept': 'application/json',
+                'x-requested-with': 'piwebapistreams',
+                'User-Agent': 'Aig-Laborious-Agent/1.0',
+            },
+        )
+
+    def get_pi_web_api_core_labels(
+        self,
+        metadata: dict[str, Any],
+        operation_type: str = 'write_pi_web_api_data',
+    ) -> dict[str, Any]:
+        """
+        Generate core labels for PI Web API metrics.
+
+        PI Web API metrics in laborious use the shared ``CORE_LABELS`` from
+        ``sientia_do``, which includes ``operation_type``. For this reason,
+        operation_type must always be present in emitted labels.
+
+        Args:
+         - metadata (dict[str, Any]): Workflow execution metadata used to derive labels.
+         - operation_type (str): Operation type label for metric cardinality.
+
+        Return:
+         dict[str, Any]: Core labels dictionary including operation_type.
+        """
+        return super().get_core_labels(
+            metadata=metadata,
+            operation_type=operation_type,
+        )
+
+    def close(self) -> None:
+        """
+        Close the PI Web API client and shutdown monitoring services.
+
+        This method properly closes all connections and resources associated
+        with the PI Web API client and monitoring services.
+        """
+        self.pi_web_api_client.close()
+        SientiaMonitoring.shutdown(self)
+
+    def process_pi_web_api_response(
+        self,
+        response_data: list[dict[str, Any]],
+        tags: dict[str, str],
+        core_labels: dict[str, str],
+        metadata: dict[str, Any],
+    ) -> tuple[int, str]:
+        """
+        Process the response data from PI Web API write operation.
+
+        Validates that all tags were successfully written, emits metrics for each tag
+        (success or error), and returns the appropriate prediction confidence value.
+        Sets error confidence if any tag write fails or if the number of written tags
+        doesn't match the expected count.
+
+        Args:
+            - response_data (dict[str, Any]): The response data from the PI Web API write operation.
+            - tags (dict[str, str]): The tags that were written to the PI Web API.
+            - core_labels (dict[str, str]): The core labels of the workflow execution.
+            - metadata (dict[str, Any]): The metadata of the workflow execution.
+        Returns:
+            int: Prediction confidence value (0 for success, 13 for errors)
+        """
+
+        # Convert tags from name:webid to webid:name
+        tags = {w: t for t, w in tags.items()}
+
+        tag_names = list[str](tags.values())
+
+        confidence = 0
+
+        message = ''
+
+        # Evaluate response for each tag
+        written_tags = []
+        for item in response_data:
+            web_id = item.get('WebId')
+            if not web_id:
+                self.error('The response did not contain some WebIds', metadata)
+                continue
+            errors = item.get('Errors', [])
+            tag_name = tags.get(web_id)
+            if not tag_name:
+                self.error(
+                    f'The response did not contain the tag name for WebId {web_id}', metadata
+                )
+                continue
+            if errors:
+                self.error(
+                    f'Error writing tag {tag_name}:{web_id} to PI Web API: {errors}', metadata
+                )
+                self.emit_metric_sync(
+                    metric_object=metrics.PI_WEB_API_PREDICTION_WRITTEN_ERROR_COUNT,
+                    tags={
+                        **core_labels,
+                        'tag_name': tag_name,
+                    },
+                )
+                confidence = PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+            else:
+                self.emit_metric_sync(
+                    metric_object=metrics.PI_WEB_API_PREDICTION_WRITTEN_COUNT,
+                    tags={
+                        **core_labels,
+                        'tag_name': tag_name,
+                    },
+                )
+                written_tags.append(tag_name)
+
+        if len(written_tags) != len(tag_names):
+            message = f'The number of written tags does not match the number of tag names: Expected {tag_names} tags, but {written_tags} tags were written.'
+
+            self.error(
+                f'{message}\nResponse:\n {json.dumps(response_data, indent=4)}\nTags:\n {json.dumps(tags, indent=4)}',
+                metadata,
+            )
+
+            self.send_notification(
+                metadata=metadata,
+                notification_id='WRITE_PI_WEB_API_PREDICTION_ERROR',
+                message=f'The number of written tags does not match the number of tag names: Expected {tag_names} tags, but {written_tags} tags were written.\nResponse:\n {json.dumps(response_data, indent=4)}\nTags:\n {json.dumps(tags, indent=4)}',
+                block='write_pi_web_api_data',
+                level=NotificationLevel.ERROR,
+            )
+            confidence = PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+
+        return confidence, message
+
+    @activity.defn(name='write_pi_web_api_data')
+    def write_pi_web_api_data(self, input_data: dict[str, Any]) -> dict[Any, Any]:
+        """
+        Write prediction and confidence data to PI Web API.
+
+        Writes prediction values and confidence scores to PI Web API using configured
+        web IDs. Processes responses to validate writes and emit metrics. Handles errors
+        gracefully by setting error confidence values when writes fail and sending
+        notifications for both prediction and confidence write errors.
+
+        Args:
+            input_data (dict[str, Any]): The input data containing:
+                - metadata (dict[str, Any]): Workflow execution metadata
+                - pi_web_api_output_config (dict[str, Any]): PI Web API configuration with:
+                    - endpoint (str): PI Web API endpoint URL
+                    - prediction_tags (dict[str, str]): Mapping of tag names to web IDs for predictions
+                    - confidence_tags (dict[str, str]): Mapping of tag names to web IDs for confidence
+                - data (dict[str, Any]): Prediction data, its a dataframe converted to dict.
+        Returns:
+            dict[Any, Any]: Data dictionary with potentially modified confidence values
+                If prediction write fails, prediction_confidence is set to error value (13)
+        """
+        metadata = input_data['metadata']
+        data = DataFrame(input_data['data'])
+        pi_web_api_output_config = input_data['pi_web_api_output_config']
+
+        self.info(f'Writing data to PI Web API... config: {pi_web_api_output_config}', metadata)
+
+        raw_prediction_tags = pi_web_api_output_config['prediction_tags']
+        raw_confidence_tags = pi_web_api_output_config['confidence_tags']
+        prediction_tags = list[str](raw_prediction_tags.values())
+        confidence_tags = list(raw_confidence_tags.values())
+
+        core_labels = self.get_pi_web_api_core_labels(metadata)
+
+        prediction_value = data.head(1)['prediction'].values[0]
+        confidence_value = data.head(1)['prediction_confidence'].values[0]
+
+        try:
+            prediction_response = self.pi_web_api_client.write_value(
+                web_ids=prediction_tags,
+                value={
+                    'Timestamp': data.head(1)['timestamp'].values[0],
+                    'Value': prediction_value,
+                },
+                metadata=metadata,
+            )
+
+            confidence, message = self.process_pi_web_api_response(
+                response_data=prediction_response,
+                tags=raw_prediction_tags,
+                core_labels=core_labels,
+                metadata=metadata,
+            )
+
+            # Preserve incoming confidence/comments on successful PI writes.
+            # Only downgrade confidence or override comments when PI response
+            # explicitly reports a problem (e.g. partial write mismatch).
+            if confidence != 0:
+                data['prediction_confidence'] = confidence
+            if message:
+                data['comments'] = message
+
+        except Exception as e:
+            trace = traceback.format_exc()
+            self.send_notification(
+                metadata=metadata,
+                notification_id='WRITE_PI_WEB_API_PREDICTION_ERROR',
+                message=f'Error writing prediction data to PI Web API: {e}\n Tags: {raw_prediction_tags}',
+                block='write_pi_web_api_data',
+                level=NotificationLevel.ERROR,
+                attachment_content=trace,
+            )
+
+            self.error(trace, metadata)
+
+            data['prediction_confidence'] = PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+            data['comments'] = str(e)
+
+            return data.to_dict()
+
+        try:
+            confidence_response = self.pi_web_api_client.write_value(
+                web_ids=confidence_tags,
+                value={
+                    'Timestamp': data.head(1)['timestamp'].values[0],
+                    'Value': float(confidence_value),
+                },
+                metadata=metadata,
+            )
+
+            self.process_pi_web_api_response(
+                response_data=confidence_response,
+                tags=raw_confidence_tags,
+                core_labels=core_labels,
+                metadata=metadata,
+            )
+
+        except Exception as e:
+            trace = traceback.format_exc()
+            self.send_notification(
+                metadata=metadata,
+                notification_id='WRITE_PI_WEB_API_CONFIDENCE_ERROR',
+                message=f'Error writing confidence data to PI Web API: {e}\n Tags: {raw_confidence_tags}',
+                block='write_pi_web_api_data',
+                level=NotificationLevel.ERROR,
+                attachment_content=trace,
+            )
+
+        return data.to_dict()

laborious/activities/gates.py

@@ -0,0 +1,812 @@
+from temporalio import activity, workflow
+
+with workflow.unsafe.imports_passed_through():
+    import traceback
+    from collections.abc import Callable, Mapping
+    from typing import Any
+
+    from pandas import DataFrame
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.notifications.models import NotificationLevel
+    from sientia_do.observability.logger import Logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+    from sientia_do.repository.minio_repository_sync import MinioRepository
+    from sientia_do.utils.formatters import create_sample_dict
+
+    from laborious import metrics
+    from laborious.utils.dataframe_debug import build_dataframe_debug_message
+    from laborious.utils.filters.conditional_filters import (
+        filter_empty_data,
+        filter_specific_variables_null_values,
+    )
+    from laborious.utils.filters.mlflow_filters import api_error_filter, nan_values_filter
+    from laborious.utils.models.minio_dataframe_payload import MinioDataFramePayload
+
+# Strongly-typed filter function signatures
+InputFilterFunc = Callable[[DataFrame, dict[str, Any]], bool]
+ResponseFilterFunc = Callable[[dict[str, Any], dict[str, Any]], bool]
+ContentFilterFunc = Callable[[DataFrame, dict[str, Any]], bool]
+
+# Input filter function mappings
+input_filter_functions: dict[str, InputFilterFunc] = {
+    'SPECIFIC_VARIABLES_NULL_VALUES': filter_specific_variables_null_values,
+    'EMPTY_DATA': filter_empty_data,
+}
+
+# Confidence mappings kept separate from function maps to avoid Union types
+input_path_confidence: Mapping[str, int] = {
+    'STOP': -1,
+    'CONTINUE': 2,
+    'REPEAT': -1,
+}
+
+# MLFlow response filter function mappings
+mlflow_response_filter_functions: dict[str, ResponseFilterFunc] = {
+    'API_ERROR': api_error_filter,
+}
+
+mlflow_response_path_confidence: Mapping[str, int] = {
+    'STOP': -1,
+    'CONTINUE': 10,
+    'REPEAT': -1,
+}
+
+# MLFlow content filter function mappings
+mlflow_content_filter_functions: dict[str, ContentFilterFunc] = {
+    'NAN_VALUES': nan_values_filter,
+    'EMPTY_DATA': filter_empty_data,
+}
+
+mlflow_content_path_confidence: Mapping[str, int] = {
+    'STOP': -1,
+    'CONTINUE': 18,
+    'REPEAT': -1,
+}
+
+
+class Gates(SientiaMonitoring):
+    """
+    Data quality gates and filtering activities for the Laborious system.
+
+    This class implements comprehensive data quality validation and filtering
+    mechanisms that can be applied at different stages of the prediction pipeline.
+    It provides configurable filters with policy-based decision making to ensure
+    data integrity and quality throughout the ML workflow.
+
+    The class supports multiple filter types and implements a flexible policy
+    system that can be configured for different validation requirements. Each
+    filter returns a path decision (STOP, CONTINUE, REPEAT) along with confidence
+    scores and detailed comments for monitoring and debugging.
+
+    Attributes:
+        input_filter_functions (dict): Mapping of input filter names to functions
+        mlflow_response_filter_functions (dict): Mapping of MLFlow response filter names to functions
+        mlflow_content_filter_functions (dict): Mapping of MLFlow content filter names to functions
+    """
+
+    minio_repository: MinioRepository | None = None
+    _MAX_DEBUG_DATAFRAME_ROWS = 100
+
+    def __init__(
+        self,
+        minio_repository: MinioRepository | None = None,
+        logger: Logger | None = None,
+        notification_handler: NotificationHandler | None = None,
+        metrics_controller: MetricsController | None = None,
+    ):
+        """
+        Initialize data quality gates with logging and notification capabilities.
+
+        Args:
+            logger: Logger instance for observability and debugging
+            notification_handler: Notification handler for alerts and monitoring
+
+        Raises:
+            Exception: If BaseActivity initialization fails
+        """
+        self.minio_repository = minio_repository
+        SientiaMonitoring.__init__(
+            self,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=metrics_controller,
+        )
+
+    def close(self) -> None:
+        """
+        Close the gates activity and clean up resources.
+        """
+
+        if self.minio_repository is not None:
+            try:
+                self.minio_repository.close()
+            finally:
+                self.minio_repository = None
+        SientiaMonitoring.shutdown(self)
+
+    def __del__(self):
+        self.close()
+
+    def _debug_dataframe(self, message: str, data: Any, metadata: dict[str, Any]) -> None:
+        """
+        Log dataframe content only when row count is below the configured threshold
+
+        Args:
+         - message (str): Base log message to identify the dataframe in logs
+         - data (Any): Dataframe-like payload to be logged
+         - metadata (dict[str, Any]): Workflow metadata for contextual logging
+        """
+        self.debug(
+            build_dataframe_debug_message(
+                message=message,
+                data=data,
+                max_rows=self._MAX_DEBUG_DATAFRAME_ROWS,
+            ),
+            metadata,
+        )
+
+    @staticmethod
+    def _read_filter_entry(config: dict[str, Any]) -> tuple[str, dict[str, Any]]:
+        """
+        Read filter policy/config keys in a case-insensitive way.
+
+        Args:
+            config (dict[str, Any]): Filter configuration dictionary.
+
+        Return:
+            tuple[str, dict[str, Any]]: Parsed policy and config payload.
+        """
+        normalized = {str(key).upper(): value for key, value in config.items()}
+        policy = normalized['POLICY']
+        filter_config = normalized.get('CONFIG', {})
+        return policy, filter_config
+
+    @activity.defn(name='input_gate')
+    def input_gate(self, input_data: dict[str, Any]) -> tuple[str | None, int, str]:
+        """
+        Apply input data quality filters and validation.
+
+        This activity validates input data quality using configurable filters
+        before proceeding with ML operations. It applies multiple filter types
+        and returns a path decision based on the filter results and configured
+        policies.
+
+        The method implements a comprehensive filtering system that:
+        1. Applies configured filters to input data
+        2. Evaluates filter results against policy configurations
+        3. Determines appropriate path decisions (STOP, CONTINUE, REPEAT)
+        4. Provides confidence scores and detailed comments
+        5. Handles errors gracefully with notification integration
+
+        Args:
+            input_data: Configuration and data for input validation
+                Required keys:
+                    - metadata (dict): Workflow execution metadata
+                    - filters (dict): Filter configuration and policies
+                    - data (dict): Input data to validate
+                    - path_priority (list[str]): Priority order for path decisions
+
+        Returns:
+            tuple: (path_flag, confidence, comment)
+                - path_flag (str | None): Decision path (STOP, CONTINUE, REPEAT, or None)
+                - confidence (int): Confidence score for the decision
+                - comment (str): Detailed explanation of the decision
+
+        Raises:
+            Exception: If filter execution fails or configuration is invalid
+        """
+        metadata = input_data['metadata']
+
+        self.info('Performing input gate...', metadata)
+
+        filters = input_data['filters']
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+        path_priority = input_data['path_priority']
+
+        filter_output = []
+
+        self._debug_dataframe('Input data:', data, metadata)
+        self.debug(f'Filters: {filters}', metadata)
+
+        # Apply each configured filter
+        for fil, config in filters.items():
+            if fil not in input_filter_functions:
+                self.error(f'Filter {fil} not found', metadata)
+                continue
+            policy, filter_config = self._read_filter_entry(config)
+            try:
+                if input_filter_functions[fil](data, filter_config):
+                    self.debug(f'Data not passed the input filter {fil}:{config}', metadata)
+                    filter_output.append(policy)
+            except Exception as e:
+                trace = traceback.format_exc()
+                self.send_notification(
+                    metadata=metadata,
+                    notification_id=f'INTPUT_GATE_ERROR__{fil}',
+                    message=f'Error in filter {fil}:{config}: \n {e}',
+                    block='input_gate',
+                    level=NotificationLevel.ERROR,
+                    attachment_content=trace,
+                )
+
+        for path_flag in path_priority:
+            if path_flag in filter_output:
+                self.info(f'Input gate result: {path_flag}', metadata)
+                return path_flag, input_path_confidence[path_flag], 'Input data with bad quality'
+
+        self.info('Nothing was filtered by the input gate', metadata)
+
+        del data
+
+        return None, 0, ''
+
+    @activity.defn(name='mlflow_response_gate')
+    def mlflow_response_gate(self, input_data: dict[str, Any]) -> tuple[str | None, int, str]:
+        """
+        Validate MLFlow API response quality and integrity.
+
+        This activity validates MLFlow API responses to ensure they meet quality
+        standards before proceeding with further processing. It applies response-specific
+        filters and determines appropriate path decisions based on response quality.
+
+        The method implements response validation that:
+        1. Applies MLFlow response-specific filters
+        2. Evaluates API response quality and integrity
+        3. Determines path decisions based on response validation results
+        4. Provides confidence scores and detailed validation comments
+        5. Handles API errors and response validation failures
+
+        Args:
+            input_data: Configuration and data for response validation
+                Required keys:
+                    - metadata (dict): Workflow execution metadata
+                    - filters (dict): Response filter configuration and policies
+                    - data (dict): MLFlow API response data to validate
+                    - type (str): Type of MLFlow operation (transform, predict)
+                    - path_priority (list[str]): Priority order for path decisions
+
+        Returns:
+            tuple: (path_flag, confidence, comment)
+                - path_flag (str | None): Decision path (STOP, CONTINUE, REPEAT, or None)
+                - confidence (int): Confidence score for the decision
+                - comment (str): Detailed explanation of the decision
+
+        Raises:
+            Exception: If response validation fails or configuration is invalid
+        """
+        metadata = input_data['metadata']
+        self.info('Performing mlflow response gate...', metadata)
+        raw_data = input_data['data']
+        filters = input_data['filters']
+
+        self.debug(
+            f'Input data: \n {create_sample_dict(raw_data, max_items=5, max_depth=5)}', metadata
+        )
+        self.debug(f'Filters: {filters}', metadata)
+
+        payload = MinioDataFramePayload.from_dict(raw_data)
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        gate_type = input_data['type']
+        path_priority = input_data['path_priority']
+
+        filter_output = []
+
+        comments = []
+
+        status = payload.status or {}
+
+        for fil, config in filters.items():
+            if fil not in mlflow_response_filter_functions:
+                continue
+            policy, filter_config = self._read_filter_entry(config)
+            try:
+                if mlflow_response_filter_functions[fil](status, filter_config):
+                    filter_output.append(policy)
+                    comments.append(status.get('message', 'Unknown MLFlow API error'))
+                    self.send_notification(
+                        metadata=metadata,
+                        notification_id=f'{gate_type.upper()}_GATE_RESPONSE_FILTER__{fil}',
+                        message=status.get('message', 'Unknown MLFlow API error'),
+                        block='mlflow_gate',
+                        level=NotificationLevel.ERROR,
+                        attachment_content=status.get('traceback'),
+                    )
+            except Exception as e:
+                trace = traceback.format_exc()
+                self.send_notification(
+                    metadata=metadata,
+                    notification_id=f'MLFLOW_GATE_RESPONSE_FILTER__{fil}',
+                    message=f'Error in filter {fil}:{config}: \n {e}',
+                    block='mlflow_gate',
+                    level=NotificationLevel.ERROR,
+                    attachment_content=trace,
+                )
+
+        for path_flag in path_priority:
+            if path_flag in filter_output:
+                self.info(f'Mlflow response gate result: {path_flag}', metadata)
+                return path_flag, mlflow_response_path_confidence[path_flag], ', '.join(comments)
+
+        self.info('Nothing was filtered by the mlflow response gate', metadata)
+
+        del data
+
+        return None, 0, ''
+
+    @activity.defn(name='mlflow_content_gate')
+    def mlflow_content_gate(self, input_data: dict[str, Any]) -> tuple[str | None, int, str]:
+        """
+        Validate MLFlow prediction content quality and integrity.
+
+        This activity validates the content of MLFlow predictions to ensure they
+        meet quality standards before export and persistence. It applies content-specific
+        filters and determines appropriate path decisions based on content quality.
+
+        The method implements content validation that:
+        1. Applies MLFlow content-specific filters
+        2. Evaluates prediction content quality and integrity
+        3. Determines path decisions based on content validation results
+        4. Provides confidence scores and detailed validation comments
+        5. Handles content validation failures and quality issues
+
+        Args:
+            input_data: Configuration and data for content validation
+                Required keys:
+                    - metadata (dict): Workflow execution metadata
+                    - filters (dict): Content filter configuration and policies
+                    - data (dict): MLFlow prediction content to validate
+                    - type (str): Type of MLFlow operation (transform, predict)
+                    - path_priority (list[str]): Priority order for path decisions
+
+        Returns:
+            tuple: (path_flag, confidence, comment)
+                - path_flag (str | None): Decision path (STOP, CONTINUE, REPEAT, or None)
+                - confidence (int): Confidence score for the decision
+                - comment (str): Detailed explanation of the decision
+
+        Raises:
+            Exception: If content validation fails or configuration is invalid
+        """
+        metadata = input_data['metadata']
+        self.info('Performing mlflow content gate...', metadata)
+
+        filters = input_data['filters']
+
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        gate_type = input_data['type']
+        path_priority = input_data['path_priority']
+
+        filter_output = []
+
+        self._debug_dataframe('Input data:', data, metadata)
+        self.debug(f'Filters: \n {filters}', metadata)
+
+        for fil, config in filters.items():
+            if fil not in mlflow_content_filter_functions:
+                continue
+            policy, filter_config = self._read_filter_entry(config)
+            try:
+                if mlflow_content_filter_functions[fil](data, filter_config):
+                    filter_output.append(policy)
+                    self.send_notification(
+                        metadata=metadata,
+                        notification_id=f'{gate_type.upper()}_GATE_CONTENT_FILTER__{fil}',
+                        message=f'Data not passed the content filter {fil}:{config}',
+                        block='mlflow_gate',
+                        level=NotificationLevel.WARNING,
+                        attachment_content=data.to_string(),
+                    )
+            except Exception as e:
+                trace = traceback.format_exc()
+                self.send_notification(
+                    metadata=metadata,
+                    notification_id=f'MLFLOW_GATE_CONTENT_FILTER__{fil}',
+                    message=f'Error in filter {fil}:{config}: \n {e}',
+                    block='mlflow_gate',
+                    level=NotificationLevel.ERROR,
+                    attachment_content=trace,
+                )
+
+        for path_flag in path_priority:
+            if path_flag in filter_output:
+                self.info(f'Mlflow content gate result: {path_flag}', metadata)
+                return (
+                    path_flag,
+                    mlflow_content_path_confidence[path_flag],
+                    'Transformed data not passed the content filter',
+                )
+
+        self.info('Nothing was filtered by the mlflow content gate', metadata)
+
+        del data
+
+        return None, 0, ''
+
+    def get_prediction_store_policy(
+        self, prediction_store_policy: str, metadata: dict[str, Any]
+    ) -> tuple[str, int]:
+        """
+        Parse and validate prediction store policy configuration.
+
+        This method parses prediction store policy strings in the format 'type:value'
+        and validates them against allowed policy types and values. It provides
+        sensible defaults for invalid configurations and logs policy validation
+        failures for operational monitoring.
+
+        Supported Policy Types:
+        - 'lts': Latest timestamp - sorts data by timestamp descending
+        - 'erl': Earliest timestamp - sorts data by timestamp ascending
+
+        Args:
+            prediction_store_policy (str): Policy string in format 'type:value'
+            metadata (dict[str, Any]): Context metadata for logging and notifications
+
+        Returns:
+            tuple[str, int]: (policy_type, policy_value)
+                - policy_type (str): Validated policy type ('lts' or 'erl')
+                - policy_value (int): Number of rows to retain
+        """
+        policy_elements = prediction_store_policy.split(':')
+
+        if len(policy_elements) < 2:
+            self.error(
+                f'Invalid prediction store policy: {prediction_store_policy}, using default policy',
+                metadata,
+            )
+            return 'lts', 1
+
+        policy_type = policy_elements[0]
+        policy_value = policy_elements[1]
+
+        # If the policy_type is not lts or erl, we use the default policy
+        # If the policty_value is not a number or 0, we use the default policy
+        if (
+            policy_type not in ['lts', 'erl']
+            or not policy_value.isdigit()
+            or int(policy_value) == 0
+        ):
+            self.error(
+                f'Invalid prediction store policy: {prediction_store_policy}, using default policy',
+                metadata,
+            )
+            return 'lts', 1
+
+        return policy_type, int(policy_value)
+
+    @activity.defn(name='format_transformed_data')
+    def format_transformed_data(self, input_data: dict[str, Any]) -> MinioDataFramePayload:
+        """
+        Format transformed data for storage and export operations.
+
+        This method formats transformed data from MLFlow model transformations
+        into a standardized format suitable for database storage. It converts
+        wide-format data (columns as variables) into long-format (melted)
+        with proper timestamp handling and model identification.
+
+        The formatting process includes:
+        1. Converting input data dictionary to DataFrame
+        2. Extracting timestamps from DataFrame index
+        3. Resetting index to create sequential row numbers
+        4. Melting data from wide format to long format (variable-value pairs)
+        5. Adding model_id for data lineage tracking
+
+        Args:
+            input_data (dict): Input data containing:
+                - metadata (dict): Workflow execution metadata
+                - data (dict[str, Any]): Transformed data to format (DataFrame-compatible dict)
+                - model_id (str): Unique identifier for the ML model
+
+        Returns:
+            dict: Formatted data dictionary with keys:
+                - timestamp (dict): Timestamp values indexed by row number
+                - variable (dict): Variable names indexed by row number
+                - value (dict): Variable values indexed by row number
+                - model_id (dict): Model identifiers indexed by row number
+        """
+        metadata = input_data['metadata']
+
+        model_id = input_data['model_id']
+
+        self.info('Formatting transformed data...', metadata)
+
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        data['timestamp'] = data.index
+        data = data.reset_index(drop=True)
+
+        data = data.melt(id_vars='timestamp', var_name='variable', value_name='value')
+        data['model_id'] = model_id
+
+        return MinioDataFramePayload.from_dataframe(
+            dataframe=data,
+            minio_repo=self.minio_repository,
+            model_name=input_data['model_name'],
+            operation='transform',
+            workflow_metadata=metadata,
+            last_timestamp=payload.last_timestamp,
+            logger=self.logger,
+        )
+
+    @activity.defn(name='format_prediction')
+    def format_prediction(self, input_data: dict[str, Any]) -> dict:
+        """
+        Format prediction data according to configured storage policies.
+
+        This method formats prediction data for storage and export operations.
+        It applies timestamp-based sorting policies, adds metadata fields,
+        and ensures data consistency before persistence. The method supports
+        multiple storage policies for flexible data retention strategies.
+
+        If only one row is present, we use the last timestamp as the timestamp
+
+        Storage Policies:
+        - 'lts:N': Latest timestamp - retains N most recent predictions
+        - 'erl:N': Earliest timestamp - retains N oldest predictions
+
+        Args:
+            input_data (dict): Input data containing:
+                - data (dict[str, Any]): Raw prediction data to format
+                - timestamp (str): Timestamp of the data
+                - model_id (str): Unique identifier for the ML model
+                - prediction_confidence (float): Confidence score for the prediction
+                - prediction_store_policy (str): Storage policy in format 'type:value'
+
+        Returns:
+            dict: Formatted prediction data ready for storage and export
+        """
+        metadata = input_data['metadata']
+        last_timestamp = input_data['timestamp']
+        prediction_store_policy = input_data['prediction_store_policy']
+        self.info('Formatting prediction...', metadata)
+
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        # Create timestamp column from index and reset index
+        data['timestamp'] = data.index
+        data = data.reset_index(drop=True)
+
+        self.debug(f'Prediction store policy: {prediction_store_policy}', metadata)
+        self._debug_dataframe('Prediction data:', data, metadata)
+
+        policy_type, policy_value = self.get_prediction_store_policy(
+            prediction_store_policy, metadata
+        )
+
+        # If data has no timestamp, we use the default timestamp and not sort the data
+        self.info(
+            f'Sorting data by timestamp and applying policy: {policy_type}:{policy_value}', metadata
+        )
+
+        # If policy_type is lts, we need to sort the data by timestamp descending and take the first policy_value rows
+        if policy_type == 'lts':
+            self.debug('Sorting data by timestamp descending', metadata)
+            data = data.sort_values(by='timestamp', ascending=False)
+        # If policy_type is erl, we need to sort the data by timestamp ascending and take the first policy_value rows
+        elif policy_type == 'erl':
+            self.debug('Sorting data by timestamp ascending', metadata)
+            data = data.sort_values(by='timestamp', ascending=True)
+        else:
+            self.error(f'Invalid policy type: {policy_type}, using default policy', metadata)
+            raise ValueError(f'Invalid policy type: {policy_type}')
+
+        int_policy_value = int(policy_value)
+
+        data = data.head(int_policy_value)
+        if int_policy_value == 1:
+            data['timestamp'] = last_timestamp
+
+        data['model_id'] = input_data['model_id']
+        data['prediction_confidence'] = input_data['prediction_confidence']
+        data['prediction_status'] = 'Good'
+        data['comments'] = ''
+        data = data.sort_values(by='timestamp', ascending=False)
+        data = data.reset_index(drop=True)
+
+        self.info(f'Prediction formatted: {len(data)} rows', metadata)
+        self._debug_dataframe('Prediction data:', data, metadata)
+
+        return data.to_dict()
+
+    @activity.defn(name='format_default_prediction')
+    def format_default_prediction(self, input_data: dict[str, Any]) -> dict:
+        """
+        Create and format default prediction data for error conditions.
+
+        This method generates default prediction data when the main prediction
+        pipeline encounters errors or quality issues. It creates a standardized
+        data structure with zero values for predictions and useful metadata
+        for operational monitoring and debugging.
+
+        The default prediction serves as a fallback mechanism to:
+        1. Maintain data pipeline continuity during failures
+        2. Provide operational visibility into prediction quality issues
+        3. Enable downstream systems to handle error conditions gracefully
+        4. Support debugging and troubleshooting efforts
+
+        Args:
+            input_data (dict): Input data containing:
+                - timestamp (str): Timestamp for the default prediction
+                - model_id (str): Unique identifier for the ML model
+                - prediction_confidence (float): Confidence score (typically low for errors)
+                - comment (str): Error description or operational comment
+
+        Returns:
+            dict: Formatted default prediction data with error indicators
+        """
+
+        metadata = input_data['metadata']
+        self.debug('Formatting default prediction...', metadata)
+
+        data = DataFrame(
+            {
+                'prediction': [0],
+                'response_time': [0],
+                'timestamp': [input_data['timestamp']],
+                'model_id': [input_data['model_id']],
+                'prediction_confidence': [input_data['prediction_confidence']],
+                'prediction_status': ['Bad'],
+                'comments': [input_data['comment']],
+            }
+        )
+
+        self.info(f'Default prediction formatted: {data.size} rows', metadata)
+        return data.to_dict()
+
+    @activity.defn(name='format_retrain_report')
+    def format_retrain_report(self, input_data: dict[str, Any]) -> dict:
+        """
+        Format retrain report data for storage and audit trail maintenance.
+
+        This method formats model retraining operation results into a standardized
+        report format suitable for database storage and operational monitoring.
+        It captures retraining status, timestamps, and model version information
+        for comprehensive audit trails and operational visibility.
+
+        The formatting process includes:
+        1. Extracting retraining experiment response data
+        2. Capturing model update report information (version, MLflow IDs)
+        3. Formatting timestamps and status information
+        4. Conditionally including version information for successful retrains
+
+        Args:
+            input_data (dict): Input data containing:
+                - metadata (dict): Workflow execution metadata
+                - experiment_response (dict): Retraining experiment response containing:
+                    - success (bool): Retraining operation success status
+                    - timestamp (str): Timestamp of the retraining operation
+                    - message (str): Status message or error description
+                - update_report (dict): Model update report containing:
+                    - version (str): New model version identifier
+                    - mlflow_run_id (str): MLflow run identifier
+                    - mlflow_experiment_id (str): MLflow experiment identifier
+                - model_id (str): Unique identifier for the ML model
+                - model_name (str): Name of the ML model
+
+        Returns:
+            dict: Formatted retrain report dictionary with keys:
+                - model_id (dict): Model identifiers indexed by row number
+                - model_name (dict): Model names indexed by row number
+                - timestamp (dict): Retraining timestamps indexed by row number
+                - status (dict): Retraining status messages indexed by row number
+                - version (dict, optional): Model versions indexed by row number
+                    Only included if experiment_response['success'] is True
+                - mlflow_run_id (dict, optional): MLflow run IDs indexed by row number
+                    Only included if experiment_response['success'] is True
+                - mlflow_experiment_id (dict, optional): MLflow experiment IDs indexed by row number
+                    Only included if experiment_response['success'] is True
+        """
+        metadata = input_data['metadata']
+        self.info('Formatting retrain report...', metadata)
+
+        experiment_response = input_data['experiment_response']
+        update_report = input_data['update_report']
+        model_id = input_data['model_id']
+        model_name = input_data['model_name']
+
+        report = DataFrame(
+            {
+                'model_id': [model_id],
+                'model_name': [model_name],
+                'timestamp': [experiment_response['timestamp']],
+                'status': [experiment_response['message']],
+            }
+        )
+
+        if experiment_response['success']:
+            # Retrain was successfull
+            report['version'] = update_report['version']
+            report['mlflow_run_id'] = update_report['mlflow_run_id']
+            report['mlflow_experiment_id'] = update_report['mlflow_experiment_id']
+
+        self._debug_dataframe('Retrain report:', report, metadata)
+
+        return report.to_dict()
+
+    @activity.defn(name='write_metrics')
+    def write_metrics(self, input_data: dict[str, Any]):
+        """
+        Write prediction performance metrics to Prometheus monitoring system.
+
+        This method records comprehensive metrics for prediction operations,
+        enabling operational monitoring, performance analysis, and alerting.
+        It tracks prediction counts, confidence levels, and response times
+        for each model and pipeline combination.
+
+        Metrics Recorded:
+        1. Prediction Count: Incremental counter for successful predictions
+        2. Confidence Monitor: Current confidence level for predictions
+        3. Response Time Monitor: Histogram of prediction response times
+
+        Args:
+            input_data (dict): Input data containing:
+                - metadata (dict[str, Any]): Workflow execution metadata
+                - prediction (dict[str, Any]): Prediction data with metrics
+
+        Raises:
+            Exception: If metrics writing fails or configuration is invalid
+        """
+        metadata = input_data['metadata']
+        prediction = DataFrame(input_data['prediction'])
+        prediction_confidence = prediction['prediction_confidence'].values[0]
+        response_time = prediction['response_time'].values[0]
+        opc_metrics = input_data['opc_metrics']
+
+        self.info(f'Writing metrics for model {metadata["model_name"]}', metadata)
+
+        core_tags = {
+            'pod_id': self.pod_id,
+            'runtime': self.runtime,
+            'operation_type': 'predict',
+            'model_name': metadata['model_name'],
+            'workflow_name': metadata['workflow_name'],
+        }
+        self.emit_metric_sync(
+            metric_object=metrics.PREDICTIONS_WRITTEN_COUNT,
+            tags=core_tags,
+        )
+
+        self.emit_metric_sync(
+            metric_object=metrics.PREDICTION_CONFIDENCE_MONITOR,
+            method='set',
+            tags=core_tags,
+            value=prediction_confidence,
+        )
+
+        self.emit_metric_sync(
+            metric_object=metrics.PREDICTION_RESPONSE_TIME_MONITOR,
+            method='observe',
+            tags=core_tags,
+            value=response_time,
+        )
+
+        for server_id, tags in opc_metrics.items():
+            for tag, response_time in tags.items():
+                if response_time is not None:
+                    self.emit_metric_sync(
+                        metric_object=metrics.PREDICTION_OPC_WRITING_RESPONSE_TIME_MONITOR,
+                        method='observe',
+                        tags={
+                            **core_tags,
+                            'opc_server_id': server_id,
+                            'tag': tag,
+                        },
+                        value=response_time,
+                    )
+
+                    self.emit_metric_sync(
+                        metric_object=metrics.PREDICTION_OPC_WRITING_COUNT,
+                        tags={
+                            **core_tags,
+                            'opc_server_id': server_id,
+                            'tag': tag,
+                        },
+                    )
+
+        self.info(f'Metrics written for model {metadata["model_name"]}', metadata)

laborious/activities/mlflow.py

@@ -0,0 +1,755 @@
+from temporalio import activity, workflow
+
+with workflow.unsafe.imports_passed_through():
+    import tempfile
+    import traceback
+    from datetime import datetime
+    from pathlib import Path
+    from shutil import rmtree
+    from typing import Any
+
+    import mlflow
+    import numpy as np
+    import pandas as pd
+    from pandas import DataFrame, to_datetime
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.notifications.models import NotificationLevel
+    from sientia_do.observability.logger import Logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+    from sientia_do.repository.minio_repository_sync import MinioRepository
+    from sientia_do.temporal.constants import (
+        DATETIME_FORMAT,
+        DATETIME_FORMAT_MS_WITH_TZ,
+        DATETIME_FORMAT_WITH_TZ,
+        now,
+    )
+    from sientia_do.utils.formatters import create_sample_dict
+    from sientia_model.model_repository.mlflow_repository import SientiaMLflowRepository
+    from sientia_model.model_repository.plugin_store import PluginStore
+
+    from laborious.utils.dataframe_debug import build_dataframe_debug_message
+    from laborious.utils.models.minio_dataframe_payload import MinioDataFramePayload
+
+
+class MLFlow(SientiaMonitoring):
+    """
+    Temporal activities that talk to MLflow through ``SientiaMLflowRepository`` and ``SientiaModel`` wrappers.
+
+    Models are resolved by registered name and the ``production`` alias (not by legacy stages or
+    separate transform/predict flavors). ``get_cached_model`` loads or reuses a wrapper; inference
+    uses ``wrapper.transform`` / ``wrapper.predict``; retrain uses ``wrapper.retrain`` or
+    ``wrapper.train`` plus ``store_model`` and registry promotion via ``promote_to_alias``.
+
+    Large inputs and outputs flow through ``MinioDataFramePayload`` when workflows offload parquet
+    to MinIO. On failure, transform/predict still return a payload with ``success: False`` and
+    error details for downstream gates.
+
+    Attributes:
+        mlflow_repository: Client for tracking, registry, artifact download, and run lifecycle.
+        plugin_store: Reference to the store (runtime is installed on the worker; reserved for
+            future store-backed helpers).
+    """
+
+    _MAX_DEBUG_DATAFRAME_ROWS = 100
+    _DEFAULT_MODEL_ALIAS = 'production'
+    _REFERENCE_ARTIFACT_CANDIDATES = ('evaluation_data.csv', 'test_data.csv')
+
+    def __init__(
+        self,
+        mlflow_repository: SientiaMLflowRepository,
+        plugin_store: PluginStore,
+        minio_repository: MinioRepository | None = None,
+        logger: Logger | None = None,
+        notification_handler: NotificationHandler | None = None,
+        metrics_controller: MetricsController | None = None,
+    ):
+        """
+        Attach shared MLflow and MinIO clients used by all ML activities in this mixin.
+
+        Args:
+         - mlflow_repository: Repository built by ``Activities`` (or injected in tests).
+         - plugin_store: Plugin store instance from worker bootstrap.
+         - minio_repository: MinIO client for ``MinioDataFramePayload`` upload/download.
+         - logger: Structured logger.
+         - notification_handler: Notifications on hard failures where applicable.
+         - metrics_controller: Shared metrics controller.
+
+        Return:
+            None
+        """
+
+        self.minio_repository = minio_repository
+        SientiaMonitoring.__init__(
+            self,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=metrics_controller,
+        )
+        self.mlflow_repository = mlflow_repository
+        self.plugin_store = plugin_store
+
+    def close(self) -> None:
+        """
+        Release MinIO manager resources held by the mixin.
+
+        Return:
+            None
+        """
+        if self.minio_repository is not None:
+            try:
+                self.minio_repository.close()
+            finally:
+                self.minio_repository = None
+        SientiaMonitoring.shutdown(self)
+
+    def __del__(self):
+        self.close()
+
+    def _debug_dataframe(self, message: str, data: Any, metadata: dict[str, Any]) -> None:
+        """
+        Log dataframe content only when row count is below the configured threshold
+
+        Args:
+         - message (str): Base log message to identify the dataframe in logs
+         - data (Any): Dataframe-like object expected to expose shape and to_csv
+         - metadata (dict[str, Any]): Workflow metadata for contextual logging
+        """
+        self.debug(
+            build_dataframe_debug_message(
+                message=message,
+                data=data,
+                max_rows=self._MAX_DEBUG_DATAFRAME_ROWS,
+            ),
+            metadata,
+        )
+
+    def _detect_and_parse_datetime_index(self, data: pd.DataFrame, metadata: dict) -> pd.DataFrame:
+        """
+        Ensure the transform output index is homogeneous and encoded as ``DATETIME_FORMAT_WITH_TZ`` strings.
+
+        Accepts an all-string index (validated against the format), or all-``datetime`` /
+        ``Timestamp`` (naive timestamps are localized to UTC before formatting). Mixed element types
+        or unsupported types raise ``ValueError`` with a message logged at info level.
+
+        Args:
+         - data: DataFrame whose index carries the time dimension after transform.
+         - metadata: Workflow metadata for log correlation.
+
+        Return:
+            ``pd.DataFrame``: Same frame with a normalized string index; empty frames are returned as-is.
+        """
+
+        if data.empty:
+            self.info('Data is empty, skipping datetime index detection and parsing', metadata)
+            return data
+
+        index = data.index
+        index_type = type(index[0])
+
+        self.info(f'Index type: {index_type}', metadata)
+
+        message = (
+            f'Index must be all timestamp like column. Valid formats are: pandas Timestamp, datetime, '
+            f'string in format {DATETIME_FORMAT_WITH_TZ}.'
+        )
+
+        if not all(isinstance(i, index_type) for i in index):
+            types = map(str, map(type, index))
+            raise ValueError(f'{message}. Elements are {",".join(types)}')
+
+        if index_type is str:
+            try:
+                pd.to_datetime(data.index, format=DATETIME_FORMAT_WITH_TZ)
+            except ValueError as e:
+                raise ValueError(f'{message}. Unable to parse given date format: {e}') from e
+
+        elif index_type is datetime or index_type is pd.Timestamp:
+            idx = data.index
+            if hasattr(idx, 'tz') and idx.tz is None:
+                data.index = idx.tz_localize('UTC')  # type: ignore[attr-defined]
+
+            data.index = data.index.strftime(DATETIME_FORMAT_WITH_TZ)  # type: ignore[attr-defined]
+        else:
+            raise ValueError(f'{message}. Got {index_type}.')
+
+        return data
+
+    def _resolve_model_version_for_run(self, run_id: str) -> str:
+        """
+        Map an MLflow ``run_id`` to the latest registered model version that produced that run.
+
+        ``search_model_versions`` may return multiple versions if the model was registered more than
+        once for the same run; the highest numeric ``version`` wins so promotion targets the newest
+        artifact set.
+
+        Args:
+         - run_id: Run UUID from ``retrain_model`` / experiment payload.
+
+        Return:
+            str: Registry version string acceptable by ``promote_to_alias``.
+
+        Raises:
+            ValueError: If the filter returns no versions (model not registered for this run).
+        """
+
+        versions = self.mlflow_repository._client.search_model_versions(
+            filter_string=f"run_id='{run_id}'"
+        )
+        if not versions:
+            raise ValueError(f'No registered model version found for run_id={run_id}')
+        latest = max(versions, key=lambda v: int(v.version))
+        return str(latest.version)
+
+    def _resolve_model_alias(self, model_config: dict[str, Any] | None = None) -> str:
+        """
+        Resolve which MLflow alias should be used for model lookup/promotion.
+
+        Args:
+         - model_config: Optional model configuration that may include ``alias``.
+
+        Return:
+            str: Alias name trimmed and normalized; defaults to ``production``.
+        """
+        if not model_config:
+            return self._DEFAULT_MODEL_ALIAS
+        alias = str(model_config.get('alias', self._DEFAULT_MODEL_ALIAS)).strip()
+        return alias or self._DEFAULT_MODEL_ALIAS
+
+    @activity.defn(name='request_transform')
+    def request_transform(self, input_data: dict[str, Any]) -> MinioDataFramePayload:
+        """
+        Pivot long-format sensor rows, load the production wrapper, and run ``wrapper.transform``.
+
+        Expected tabular shape after load: columns including ``variable``, ``timestamp``, ``value``,
+        and ``created_at`` for deduplication. Data are sorted by ``created_at``, de-duplicated per
+        ``(variable, timestamp)``, pivoted wide, then passed to the model. ``model_config`` may
+        include ``retention_minutes`` for wrapper cache TTL.
+
+        Args:
+         - input_data: Dict with ``metadata``, ``model_name``, ``data`` (``MinioDataFramePayload``
+           dict or inline dataframe dict), and optional ``model_config``.
+
+        Return:
+            ``MinioDataFramePayload`` with transformed frame and ``success: True``, or a payload
+            with ``success: False`` and exception details in ``status`` if transform fails.
+        """
+        metadata = input_data['metadata']
+        self.info('Transforming data...', metadata)
+
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        model_name = input_data['model_name']
+        model_config = input_data.get('model_config', {})
+        model_alias = self._resolve_model_alias(model_config)
+
+        self._debug_dataframe('Raw input data:', data, metadata)
+
+        # Long → wide: keep newest row per (variable, timestamp), then pivot for the wrapper API.
+        data = data.sort_values('created_at', ascending=False).drop_duplicates(
+            subset=['variable', 'timestamp'], keep='first'
+        )
+
+        data = data.pivot(index='timestamp', columns='variable', values='value')
+        data.fillna(np.nan, inplace=True)
+
+        data.columns.name = None
+        data.index.name = None
+
+        data['timestamp'] = data.index
+
+        self._debug_dataframe('Processed input data:', data, metadata)
+
+        try:
+            wrapper = self.mlflow_repository.get_cached_model(
+                model_name=model_name,
+                alias=model_alias,
+                retention_minutes=model_config.get('retention_minutes', 0),
+                metadata=metadata,
+            )
+            transformed_df, transform_meta = wrapper.transform(data)
+            if transform_meta:
+                self.info(f'Wrapper transform metadata: {transform_meta}', metadata)
+
+            transformed_df = self._detect_and_parse_datetime_index(transformed_df, metadata)
+
+            response_data: dict[str, Any] = {'success': True, 'content': transformed_df}
+        except Exception as e:
+            response_data = {
+                'success': False,
+                'content': {'message': str(e), 'traceback': traceback.format_exc()},
+            }
+
+        self.debug(
+            f'Transform raw response data: \n {create_sample_dict(response_data, max_items=5, max_depth=5)}',
+            metadata,
+        )
+
+        self.debug(
+            f'Transform response data: \n {create_sample_dict(response_data, max_items=5, max_depth=5)}',
+            metadata,
+        )
+
+        self.info('Data transformed successfully', metadata)
+
+        if not response_data.get('success', False):
+            return MinioDataFramePayload.from_dataframe(
+                dataframe=None,
+                minio_repo=self.minio_repository,
+                model_name=model_name,
+                operation='transform',
+                status=response_data,
+                workflow_metadata=metadata,
+                last_timestamp=payload.last_timestamp,
+                logger=self.logger,
+            )
+
+        return MinioDataFramePayload.from_dataframe(
+            dataframe=response_data['content'],
+            minio_repo=self.minio_repository,
+            model_name=model_name,
+            operation='transform',
+            workflow_metadata=metadata,
+            status={
+                'success': True,
+            },
+            last_timestamp=payload.last_timestamp,
+            logger=self.logger,
+        )
+
+    @activity.defn(name='request_predict')
+    def request_predict(self, input_data: dict[str, Any]) -> MinioDataFramePayload:
+        """
+        Load the production wrapper and call ``wrapper.predict`` on the prepared feature frame.
+
+        The activity normalizes ``NaN`` to ``None`` for JSON-friendly columns, sets the row index
+        the same way as ``retrain_model`` (UTC ``DatetimeIndex`` from ``DATETIME_FORMAT_WITH_TZ``),
+        restores that index on the prediction frame, normalizes the prediction index to
+        ``DATETIME_FORMAT_WITH_TZ`` strings like ``request_transform``, and records ``response_time``.
+        Non-DataFrame predictions are coerced to a single ``prediction`` column.
+
+        Args:
+         - input_data: Same envelope as ``request_transform`` (``metadata``, ``model_name``,
+           ``data``, optional ``model_config`` with ``retention_minutes``).
+
+        Return:
+            ``MinioDataFramePayload`` with predictions or error status mirroring transform behaviour.
+        """
+        metadata = input_data['metadata']
+        self.info('Predicting data...', metadata)
+
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        model_name = input_data['model_name']
+        model_config = input_data.get('model_config', {})
+        model_alias = self._resolve_model_alias(model_config)
+
+        self._debug_dataframe('Input data for prediction:', data, metadata)
+
+        data.replace(np.nan, None, inplace=True)
+
+        data.index = pd.DatetimeIndex(
+            to_datetime(data.index, format=DATETIME_FORMAT_WITH_TZ, utc=True)
+        )
+        input_index = data.index
+
+        try:
+            wrapper = self.mlflow_repository.get_cached_model(
+                model_name=model_name,
+                alias=model_alias,
+                retention_minutes=model_config.get('retention_minutes', 0),
+                metadata=metadata,
+            )
+            start_time = datetime.now()
+            predict_data, pred_meta = wrapper.predict({}, data)
+            end_time = datetime.now()
+
+            if pred_meta:
+                self.info(f'Wrapper predict metadata: {pred_meta}', metadata)
+
+            if isinstance(predict_data, DataFrame):
+                self._debug_dataframe(
+                    'Data received from model prediction:', predict_data, metadata
+                )
+                predict_data.columns = pd.Index(['prediction'])
+            else:
+                self.debug(
+                    f'Data received from model prediction (not a DataFrame): {predict_data}',
+                    metadata,
+                )
+                predict_data = pd.DataFrame(predict_data, columns=['prediction'])
+
+            predict_data.index = input_index
+            predict_data['response_time'] = (end_time - start_time).total_seconds()
+            predict_data = self._detect_and_parse_datetime_index(predict_data, metadata)
+
+            response_data: dict[str, Any] = {'success': True, 'content': predict_data}
+        except Exception as e:
+            response_data = {
+                'success': False,
+                'content': {'message': str(e), 'traceback': traceback.format_exc()},
+            }
+
+        self.debug(
+            f'Prediction response data: \n {create_sample_dict(response_data, max_items=5, max_depth=5)}',
+            metadata,
+        )
+
+        self.info('Data predicted successfully', metadata)
+
+        if not response_data.get('success', False):
+            return MinioDataFramePayload.from_dataframe(
+                dataframe=None,
+                minio_repo=self.minio_repository,
+                model_name=model_name,
+                operation='predict',
+                status=response_data,
+                workflow_metadata=metadata,
+                last_timestamp=payload.last_timestamp,
+                logger=self.logger,
+            )
+
+        return MinioDataFramePayload.from_dataframe(
+            dataframe=response_data['content'],
+            minio_repo=self.minio_repository,
+            model_name=model_name,
+            operation='predict',
+            workflow_metadata=metadata,
+            status={
+                'success': True,
+            },
+            last_timestamp=payload.last_timestamp,
+            logger=self.logger,
+        )
+
+    @activity.defn(name='retrain_model')
+    def retrain_model(self, input_data: dict[str, Any]) -> dict[str, Any]:
+        """
+        Fit an updated wrapper from historical data, then log and register in MLflow.
+
+        Flow: load long-format data from MinIO → dedupe/pivot like inference prep → require
+        ``model_config['target']`` → read current ``production`` version for ``source_run_id`` tag →
+        run ``wrapper.retrain`` outside run timing → ``start_run`` with retrain tags → log input
+        CSV artifact → ``store_model`` and ``log_params``. Does not promote; the workflow calls
+        ``update_production_model`` after validation.
+
+        Args:
+         - input_data: Must include ``metadata``, ``model_name``, ``data`` (payload), and
+           ``model_config`` with at least ``target``.
+
+        Return:
+            On success: ``success``, ``experiment`` (``run_id``, ``experiment_id``, ``experiment_name``),
+            ``message``, ``timestamp``. On failure: ``success: False``, error fields, and optional trace.
+        """
+
+        if self.minio_repository is None:
+            raise ValueError('Minio repository not initialized')
+
+        metadata = input_data['metadata']
+
+        try:
+            payload = MinioDataFramePayload.from_dict(input_data['data'])
+            data = payload.retrieve(self.minio_repository, metadata)
+
+        except Exception as e:
+            trace = traceback.format_exc()
+            self.send_notification(
+                metadata=metadata,
+                notification_id='ERROR_LOADING_RETRAIN_DATA',
+                message=f'Error loading retrain data: {e}',
+                block='retrain_model',
+                level=NotificationLevel.ERROR,
+                attachment_content=trace,
+            )
+            self.error(trace, metadata)
+            return {
+                'success': False,
+                'message': f'Error loading retrain data: {e}',
+                'traceback': trace,
+                'timestamp': now().strftime(DATETIME_FORMAT_MS_WITH_TZ),
+            }
+
+        self.debug(f'Retrain data loaded successfully: shape {data.shape}', metadata)
+
+        model_name = input_data['model_name']
+        model_config = input_data.get('model_config', {})
+
+        self.info(f'Retraining model {model_name}...', metadata)
+
+        timestamp = data['timestamp'].max()
+        self.debug(f'Timestamp: {timestamp}', metadata)
+
+        if 'created_at' in data.columns:
+            data = data.sort_values('created_at', ascending=False).drop_duplicates(
+                subset=['variable', 'timestamp'], keep='first'
+            )
+        else:
+            data = data.drop_duplicates(subset=['variable', 'timestamp'], keep='first')
+
+        data.drop(columns=['model_id'], inplace=True, errors='ignore')
+        data.drop(columns=['created_at'], inplace=True, errors='ignore')
+
+        data = data.pivot(index='timestamp', columns='variable', values='value')
+        data.fillna(np.nan, inplace=True)
+        data.columns.name = None
+        data.index.name = None
+
+        data.index = pd.DatetimeIndex(
+            to_datetime(data.index, format=DATETIME_FORMAT_WITH_TZ, utc=True)
+        )
+
+        target = model_config.get('target')
+        if target is None:
+            msg = 'model_config must include "target" for retraining'
+            self.info(msg, metadata)
+            return {
+                'success': False,
+                'experiment': None,
+                'message': msg,
+                'traceback': '',
+                'timestamp': str(timestamp),
+            }
+
+        try:
+            model_alias = self._resolve_model_alias(model_config)
+            mv_src = self.mlflow_repository._client.get_model_version_by_alias(
+                name=model_name,
+                alias=model_alias,
+            )
+            source_run_id = mv_src.run_id
+
+            wrapper = self.mlflow_repository.get_cached_model(
+                model_name=model_name,
+                alias=model_alias,
+                retention_minutes=0,
+                metadata=metadata,
+            )
+
+            # Keep heavy model fitting outside MLflow run timing.
+            prediction_data = wrapper.retrain(data)
+            prediction_data.rename(columns={target: 'prediction'}, inplace=True)
+
+            # Merge prediction data with retrain data
+            evaluation_data = pd.merge(
+                data, prediction_data, left_index=True, right_index=True, how='left'
+            )
+
+            # Rename target column to "target"
+            evaluation_data.rename(columns={target: 'target'}, inplace=True)
+
+            # Reset index and put as column "timestamp"
+            evaluation_data['timestamp'] = evaluation_data.index
+            evaluation_data.reset_index(drop=True, inplace=True)
+            evaluation_data.sort_values(by='timestamp', inplace=True, ascending=True)
+
+            run_name = f'{model_name}-retrain-{datetime.now().strftime("%Y%m%d%H%M%S")}'
+
+            with self.mlflow_repository.start_run(
+                model_name=model_name,
+                run_name=run_name,
+                experiment_name=model_name,
+                tags={'retrain': 'true', 'source_run_id': source_run_id},
+                metadata=metadata,
+            ) as run_info:
+                tmp_dir = tempfile.mkdtemp(prefix='laborious_retrain_')
+                try:
+                    raw_csv = Path(tmp_dir) / 'retrain_input.csv'
+                    evaluation_csv = Path(tmp_dir) / 'evaluation_data.csv'
+                    data.to_csv(raw_csv, index=False)
+                    evaluation_data.to_csv(evaluation_csv, index=False)
+
+                    mlflow.log_artifact(str(raw_csv))
+                    mlflow.log_artifact(str(evaluation_csv))
+                finally:
+                    rmtree(tmp_dir, ignore_errors=True)
+
+                wrapper.store_model(name=model_name)
+
+                self.mlflow_repository.log_params(
+                    {
+                        'retrain': 'true',
+                        'retrain_date': datetime.now().isoformat(),
+                        'source_run_id': source_run_id,
+                        'retrain_samples': str(data.shape),
+                    }
+                )
+
+                experiment_payload = {
+                    'run_id': run_info.run_id,
+                    'experiment_id': run_info.experiment_id,
+                    'experiment_name': model_name,
+                }
+
+            return {
+                'success': True,
+                'experiment': experiment_payload,
+                'message': 'Model retrained successfully.',
+                'timestamp': str(timestamp),
+            }
+
+        except Exception as e:
+            error_msg = f'Error retraining model {model_name}: {e}'
+            self.info(error_msg, metadata)
+            return {
+                'success': False,
+                'experiment': None,
+                'message': error_msg,
+                'traceback': traceback.format_exc(),
+                'timestamp': str(timestamp),
+            }
+
+    @activity.defn(name='update_production_model')
+    def update_production_model(self, input_data: dict[str, Any]) -> dict[Any, Any]:
+        """
+        Point the ``production`` alias at the model version registered for the retrain run.
+
+        Resolves the highest numeric registry version whose ``run_id`` matches
+        ``experiment['run_id']``, then calls ``promote_to_alias``. On failure, sends a notification
+        and re-raises so the workflow can surface the error.
+
+        Args:
+         - input_data: ``metadata``, ``model_name``, and ``experiment`` with ``run_id`` and
+           ``experiment_id`` (as returned from ``retrain_model``).
+
+        Return:
+            Dict with ``model_name``, promoted ``version``, ``mlflow_run_id``, ``mlflow_experiment_id``.
+        """
+        metadata = input_data['metadata']
+        model_name = input_data['model_name']
+        experiment = input_data['experiment']
+        self.info(
+            f'Updating production model {model_name} from experiment {experiment}...', metadata
+        )
+
+        try:
+            run_id = experiment['run_id']
+            experiment_id = experiment['experiment_id']
+
+            version = self._resolve_model_version_for_run(run_id)
+
+            promote_alias = self._resolve_model_alias(input_data.get('model_config'))
+            self.mlflow_repository.promote_to_alias(
+                model_name=model_name,
+                version=version,
+                alias=promote_alias,
+                metadata=metadata,
+            )
+
+            self.info(f'Production model {model_name} updated successfully', metadata)
+            return {
+                'model_name': model_name,
+                'version': version,
+                'mlflow_run_id': run_id,
+                'mlflow_experiment_id': experiment_id,
+            }
+
+        except Exception as e:
+            trace = traceback.format_exc()
+            self.send_notification(
+                metadata=metadata,
+                notification_id='UPDATE_PRODUCTION_MODEL_ERROR',
+                message=f'Error updating production model {model_name}: {e}',
+                block='update_production_model',
+                level=NotificationLevel.ERROR,
+                attachment_content=trace,
+            )
+            self.error(trace, metadata=metadata)
+            raise e
+
+    def _resolve_reference_artifact_name(self, run_id: str) -> str | None:
+        """
+        Pick the first available reference CSV artifact path from the MLflow run.
+
+        Candidates are checked in priority order: ``retrain_input.csv``, then ``train_data.csv``.
+        A path matches when it equals the candidate or ends with ``/<candidate>`` for nested layouts.
+
+        Args:
+         - run_id: MLflow run UUID linked to the production model version.
+
+        Return:
+            Artifact path string for ``download_artifacts``, or ``None`` if no candidate exists.
+        """
+        listed = self.mlflow_repository._client.list_artifacts(run_id)
+        paths = [file_info.path for file_info in listed]
+        for candidate in self._REFERENCE_ARTIFACT_CANDIDATES:
+            for path in paths:
+                if path == candidate or path.endswith(f'/{candidate}'):
+                    return path
+        return None
+
+    def _find_downloaded_csv(self, tmpdir: str, artifact_name: str) -> Path | None:
+        """
+        Locate a downloaded reference CSV in the temp directory.
+
+        Args:
+         - tmpdir: Directory where ``download_artifacts`` wrote files.
+         - artifact_name: Basename of the resolved artifact (e.g. ``retrain_input.csv``).
+
+        Return:
+            ``Path`` to the CSV file if found, else ``None``.
+        """
+        direct = Path(tmpdir) / artifact_name
+        if direct.exists():
+            return direct
+        matches = list(Path(tmpdir).rglob(artifact_name))
+        return matches[0] if matches else None
+
+    @activity.defn(name='get_reference_data')
+    def get_reference_data(self, input_data: dict[str, Any]) -> list[dict] | None:
+        """
+        Download reference training CSV from the MLflow run linked to the production alias.
+
+        Resolves ``retrain_input.csv`` or ``train_data.csv`` via artifact listing before download.
+        ``retrain_input.csv`` is preferred when both exist (most recent retrain snapshot). Used by
+        drift workflows to compare live data against the reference distribution logged with the model.
+        Timestamps are normalized to ``DATETIME_FORMAT`` string columns before returning records.
+
+        Args:
+         - input_data: ``metadata``, ``model_name``, and optional ``model_config`` with ``alias``.
+
+        Return:
+            List of row dicts with normalized timestamps, or ``None`` if resolution or load fails.
+        """
+        metadata = input_data['metadata']
+        model_name = input_data['model_name']
+
+        try:
+            model_alias = self._resolve_model_alias(input_data.get('model_config'))
+            mv = self.mlflow_repository._client.get_model_version_by_alias(
+                name=model_name,
+                alias=model_alias,
+            )
+            run_id = mv.run_id
+
+            artifact_path = self._resolve_reference_artifact_name(run_id)
+            if artifact_path is None:
+                self.warning(f'Reference data not found for model {model_name}', metadata)
+                return None
+
+            artifact_name = Path(artifact_path).name
+            tmpdir = tempfile.mkdtemp(prefix='laborious_eval_')
+            try:
+                self.mlflow_repository.download_artifacts(
+                    run_id=run_id,
+                    artifact_path=artifact_path,
+                    dst_path=tmpdir,
+                    metadata=metadata,
+                )
+                csv_path = self._find_downloaded_csv(tmpdir, artifact_name)
+                if csv_path is None:
+                    self.warning(f'Reference data not found for model {model_name}', metadata)
+                    return None
+                reference_data = pd.read_csv(csv_path)
+            finally:
+                rmtree(tmpdir, ignore_errors=True)
+
+            reference_data['timestamp'] = to_datetime(reference_data['timestamp'])
+            reference_data['timestamp'] = reference_data['timestamp'].dt.strftime(DATETIME_FORMAT)
+
+            return reference_data.to_dict(orient='records')
+
+        except Exception as e:
+            self.warning(f'Reference data not found for model {model_name}: {e}', metadata)
+            return None

laborious/activities/model_metrics.py

@@ -0,0 +1,432 @@
+from temporalio import activity, workflow
+
+with workflow.unsafe.imports_passed_through():
+    import time
+    import traceback
+    import warnings
+    from typing import Any
+
+    import numpy as np
+    import pandas as pd
+    from pandas import DataFrame, Index, Series, to_datetime
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.notifications.models import NotificationLevel
+    from sientia_do.observability.logger import Logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+    from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+    from sientia_model.analytics.drift_analysis import DriftAnalysis, DriftInsufficientDataError
+
+    from laborious import metrics
+    from laborious.utils.dataframe_debug import build_dataframe_debug_message
+
+warnings.filterwarnings('ignore', category=RuntimeWarning, message='Degrees of freedom <= 0')
+warnings.filterwarnings(
+    'ignore', category=RuntimeWarning, message='invalid value encountered in scalar divide'
+)
+
+
+class ModelMetrics(SientiaMonitoring):
+    """
+    Metrics and statistical analysis activities for the Laborious pipeline.
+
+    This class centralizes drift/statistical computations and model-quality
+    aggregates used by scheduled workflows. Besides producing tabular outputs
+    for persistence, it also emits operational metrics (count, lag, error)
+    through ``SientiaMonitoring`` so execution health is observable in runtime.
+    """
+
+    _MAX_DEBUG_DATAFRAME_ROWS = 100
+
+    def __init__(
+        self,
+        logger: Logger,
+        notification_handler: NotificationHandler,
+        metrics_controller: MetricsController,
+    ):
+        SientiaMonitoring.__init__(self, logger, notification_handler, metrics_controller)
+
+    def close(self) -> None:
+        """
+        Shutdown monitoring resources associated with model metrics activities.
+
+        This is invoked during worker teardown to flush/close metric controller
+        internals and prevent dangling telemetry tasks.
+        """
+        SientiaMonitoring.shutdown(self)
+
+    def __del__(self):
+        self.close()
+
+    def _debug_dataframe(self, message: str, data: Any, metadata: dict[str, Any]) -> None:
+        """
+        Log dataframe content only when row count is below the configured threshold
+
+        Args:
+         - message (str): Base log message to identify the dataframe in logs
+         - data (Any): Dataframe-like payload to be logged
+         - metadata (dict[str, Any]): Workflow metadata for contextual logging
+        """
+        self.debug(
+            build_dataframe_debug_message(
+                message=message,
+                data=data,
+                max_rows=self._MAX_DEBUG_DATAFRAME_ROWS,
+            ),
+            metadata,
+        )
+
+    def _drift_analyze_stage_error(
+        self,
+        exc: Exception,
+        context: str,
+        metadata: dict[str, Any],
+        core_labels: dict[str, Any],
+    ) -> None:
+        """
+        Log analyzer failure for a drift stage and increment the analyze error metric.
+
+        Args:
+         - exc (Exception): Failure raised by ``sientia_model``.
+         - context (str): Short label for the log line (e.g. univariate detection).
+         - metadata (dict[str, Any]): Workflow metadata for logging.
+         - core_labels (dict[str, Any]): Tags from ``get_core_labels`` for metrics.
+        """
+        self.error(f'{context}: {exc}', metadata)
+        self.emit_metric_sync(metric_object=metrics.MODEL_ANALYZE_ERROR_COUNT, tags=core_labels)
+
+    def get_drift_metrics(
+        self,
+        reference_data: DataFrame,
+        target_data: DataFrame,
+        target_name: str,
+        reference_columns: Index,
+        drift_metrics: list[str],
+        chunk_period: str,
+        metadata: dict[str, Any],
+    ) -> DataFrame:
+        """
+        Compute univariate and multivariate drift outputs and merge them into one dataframe.
+
+        The method orchestrates three analysis stages (univariate drift,
+        multivariate drift, and dataframe projection), emitting lag/count/error
+        metrics for each stage independently so failures are attributable.
+
+        Args:
+         - reference_data (DataFrame): Baseline dataset representing expected behavior.
+         - target_data (DataFrame): Current analysis dataset to compare against reference.
+         - target_name (str): Target column name used by ``DriftAnalysis`` config.
+         - reference_columns (Index): Feature columns evaluated for drift.
+         - drift_metrics (list[str]): Enabled univariate methods.
+         - chunk_period (str): Time bucket granularity used by analysis methods.
+         - metadata (dict[str, Any]): Workflow metadata for logs and notifications.
+
+        Return:
+         DataFrame: Consolidated drift dataframe from ``get_drift_metrics_dataframe`` using
+         ``method`` / ``value`` (and optional ``threshold``, ``drift_type``), ready for
+         activity-level formatting before Postgres export.
+        """
+        # ``DriftAnalysis`` uses truthiness checks on ``features`` (e.g. ``if not features``);
+        # a pandas ``Index`` is ambiguous in boolean context — normalize to a list.
+        feature_names: list[str] = list(reference_columns)
+
+        config = {
+            'target': target_name,
+            'prediction': 'prediction',
+            'timestamp': 'timestamp',
+            'features': feature_names,
+        }
+
+        drift_analysis = DriftAnalysis(config=config)
+
+        self._debug_dataframe(
+            f'Reference data: Size {reference_data.shape}', reference_data, metadata
+        )
+
+        self._debug_dataframe(f'Target data: Size {target_data.shape}', target_data, metadata)
+
+        core_labels = self.get_core_labels(metadata, operation_type='detect_univariate_drift')
+        start_time = time.time()
+        try:
+            univariate_drift = drift_analysis.detect_univariate_drift(
+                reference_df=reference_data,
+                analysis_df=target_data,
+                features=feature_names,
+                timestamp_col=config['timestamp'],
+                methods=drift_metrics,
+                chunk_period=chunk_period,
+            )
+        except Exception as e:
+            if isinstance(e, DriftInsufficientDataError):
+                raise
+            self._drift_analyze_stage_error(
+                e, 'Error detecting univariate drift', metadata, core_labels
+            )
+            raise
+        self.observe_lag_sync(start_time, metrics.MODEL_ANALYZE_LAG, core_labels)
+        self.emit_metric_sync(metric_object=metrics.MODEL_ANALYZE_COUNT, tags=core_labels)
+
+        core_labels = self.get_core_labels(metadata, operation_type='detect_multivariate_drift')
+        start_time = time.time()
+        try:
+            multivariate_drift = drift_analysis.detect_multivariate_drift(
+                reference_df=reference_data,
+                analysis_df=target_data,
+                features=feature_names,
+                timestamp_col=config['timestamp'],
+                chunk_period=chunk_period,
+            )
+        except Exception as e:
+            if isinstance(e, DriftInsufficientDataError):
+                raise
+            self._drift_analyze_stage_error(
+                e, 'Error detecting multivariate drift', metadata, core_labels
+            )
+            raise
+        self.observe_lag_sync(start_time, metrics.MODEL_ANALYZE_LAG, core_labels)
+        self.emit_metric_sync(metric_object=metrics.MODEL_ANALYZE_COUNT, tags=core_labels)
+
+        start_time = time.time()
+        core_labels = self.get_core_labels(metadata, operation_type='get_drift_metrics_dataframe')
+        try:
+            drift_df = drift_analysis.get_drift_metrics_dataframe(
+                univariate_drift=univariate_drift,
+                multivariate_drift=multivariate_drift,
+            )
+        except Exception as e:
+            if isinstance(e, DriftInsufficientDataError):
+                raise
+            self._drift_analyze_stage_error(
+                e, 'Error building drift metrics dataframe', metadata, core_labels
+            )
+            raise
+        self.observe_lag_sync(start_time, metrics.MODEL_ANALYZE_LAG, core_labels)
+        self.emit_metric_sync(metric_object=metrics.MODEL_ANALYZE_COUNT, tags=core_labels)
+
+        return drift_df
+
+    @staticmethod
+    def _to_naive_utc(series: Series) -> Series:
+        """
+        Parse ``series`` as datetime and return a TZ-naive UTC copy.
+
+        ``sientia_model.analytics.drift_analysis.DriftAnalysis`` preserves the
+        timezone of the input dataframe in its outputs, while target rows
+        loaded from PostgreSQL come in with ``+00:00``. Forcing both sides of
+        a comparison to TZ-naive UTC keeps ``isin`` / ``floor`` operations
+        deterministic regardless of how the analyzer constructs its
+        timestamps.
+
+        Args:
+         - series (Series): Input series containing datetime-parseable values.
+
+        Return:
+         Series: Datetime64 series with ``tz=None`` representing UTC instants.
+        """
+        parsed = to_datetime(series)
+        if getattr(parsed.dt, 'tz', None) is not None:
+            parsed = parsed.dt.tz_convert('UTC').dt.tz_localize(None)
+        return parsed
+
+    @activity.defn(name='calculate_drift')
+    def calculate_drift(self, input_data: dict[str, Any]) -> list[dict]:
+        """
+        Calculate drift metrics for a model.
+
+        Args:
+            input_data (dict[str, Any]): Input data containing:
+                - metadata (dict): Workflow execution metadata
+                - model_name (str): Name of the MLFlow model to calculate drift for
+                - reference_data (pd.DataFrame): Reference data for the model
+                - target_data (pd.DataFrame): Target data for calculating drift
+                - target_name (str): Name of the target column
+                - drift_metrics (list[str]): List of drift metrics to calculate
+        """
+        metadata = input_data['metadata']
+        model_name = input_data['model_name']
+        model_id = input_data['model_id']
+        reference_raw_data = input_data['reference_data']
+        target_data = DataFrame(input_data['target_data'])
+        target_name = input_data['target_name']
+        drift_metrics = input_data['drift_metrics']
+        chunk_period = input_data['chunk_period']
+
+        if chunk_period not in ['min', 's']:
+            self.error(f'Invalid chunk period: {chunk_period}', metadata)
+            raise ValueError(f'Invalid chunk period: {chunk_period}, must be "min" or "s"')
+
+        self.info(f'Calculating drift for model {model_name}', metadata)
+
+        target_data = target_data.pivot(index='timestamp', columns='variable', values='value')
+        target_data['timestamp'] = target_data.index
+        # Keep timestamps as datetime: DriftAnalysis._chunk_dataframe relies on
+        # ``pd.Grouper(freq=...)`` which rejects string timestamp columns.
+        target_data['timestamp'] = to_datetime(target_data['timestamp'])
+        target_data = target_data.reset_index(drop=True)
+        target_data.dropna(inplace=True)
+
+        if reference_raw_data is not None:
+            self.info('Using reference data', metadata)
+            reference_data = DataFrame(reference_raw_data)
+            if 'timestamp' in reference_data.columns:
+                reference_data['timestamp'] = to_datetime(reference_data['timestamp'])
+            accurate = True
+        else:
+            # Get 30% first rows of target_data
+            self.warning('Using 30% first rows of target data as reference data', metadata)
+            target_data.sort_values(by='timestamp', ascending=True, inplace=True)
+            reference_data = target_data.head(int(len(target_data) * 0.3))
+            accurate = False
+
+            self.send_notification(
+                metadata=metadata,
+                notification_id='MODEL_METRICS_REFERENCE_DATA_WARNING',
+                message='Using 30% first rows of target data as reference data',
+                block='model_metrics',
+                level=NotificationLevel.WARNING,
+                attachment_content=reference_data.to_csv(),
+            )
+
+        reference_columns = reference_data.drop(
+            columns=[target_name, 'timestamp', 'target', 'prediction'], errors='ignore'
+        ).columns
+
+        try:
+            drift_df = self.get_drift_metrics(
+                reference_data=reference_data,
+                target_data=target_data,
+                target_name=target_name,
+                reference_columns=reference_columns,
+                drift_metrics=drift_metrics,
+                chunk_period=chunk_period,
+                metadata=metadata,
+            )
+        except Exception as e:
+            if isinstance(e, DriftInsufficientDataError):
+                self.error(str(e), metadata)
+                notification_id = e.notification_id
+                notification_message = str(e)
+            else:
+                self.error(f'Error getting drift metrics: {e}', metadata)
+                notification_id = 'MODEL_METRICS_GET_DRIFT_METRICS_ERROR'
+                notification_message = f'Error getting drift metrics: {e}'
+            self.send_notification(
+                metadata=metadata,
+                notification_id=notification_id,
+                message=notification_message,
+                block='model_metrics',
+                level=NotificationLevel.ERROR,
+                attachment_content=traceback.format_exc(),
+            )
+            raise
+
+        # Drop chunks whose floored timestamp does not appear in the analysis window.
+        # ``DriftAnalysis`` chunks over ``analysis_df``; this only excludes rows that
+        # do not belong to the current target window (e.g. stray merged reference rows).
+        target_floor = self._to_naive_utc(target_data['timestamp']).dt.floor(chunk_period)
+        drift_floor = self._to_naive_utc(drift_df['timestamp']).dt.floor(chunk_period)
+        drift_df = drift_df[drift_floor.isin(target_floor)]
+
+        if drift_df.empty:
+            self.warning(
+                'No drift metrics found after dropping rows where timestamp is not in target data',
+                metadata,
+            )
+            return []
+
+        # Analyzer emits diagnostic columns that are not stored in ``sientia_data.drift_metrics``.
+        drift_df = drift_df.drop(columns=['threshold', 'drift_type'], errors='ignore')
+
+        drift_df['model_id'] = str(model_id)
+        drift_df['accurate'] = accurate
+
+        # ``timestamp`` is overridden with the most recent target instant so
+        # every persisted row shares a single business timestamp (the run's
+        # logical "now"), matching what downstream consumers expect.
+        latest_target_timestamp = self._to_naive_utc(target_data['timestamp']).max()
+        drift_df['timestamp'] = (
+            pd.Timestamp(latest_target_timestamp)
+            .tz_localize('UTC')
+            .strftime(DATETIME_FORMAT_WITH_TZ)
+        )
+
+        # ``chunk_start_date`` / ``chunk_end_date`` may carry nanosecond
+        # precision (beyond ``timestamptz`` microseconds), so serialize as ISO
+        # text for the ``text`` Postgres columns.
+        for column in ('chunk_start_date', 'chunk_end_date'):
+            drift_df[column] = drift_df[column].apply(
+                lambda value: pd.Timestamp(value).isoformat() if pd.notna(value) else None
+            )
+
+        self._debug_dataframe(f'Drift dataframe: Size {drift_df.shape}', drift_df, metadata)
+
+        return drift_df.to_dict(orient='records')
+
+    @activity.defn(name='calculate_simple_metrics')
+    def calculate_simple_metrics(self, input_data: dict[str, Any]) -> list[dict]:
+        """
+        Calculate simple metrics for a model. Metrics available are:
+            - rmse
+            - mse
+            - mae
+            - r2
+            - accuracy
+            - precision
+            - recall
+            - f1
+        Args:
+            input_data (dict[str, Any]): Input data containing:
+                - metadata (dict): Workflow execution metadata
+                - model_id (str): ID of the MLFlow model
+                - target_data (pd.DataFrame): Target data for calculating metrics, containing target and prediction columns
+                - metrics (list[str]): List of metrics to calculate
+        Returns:
+            dict[Hashable, Any]: Dictionary containing the calculated metrics
+        """
+
+        metadata = input_data['metadata']
+        model_id = input_data['model_id']
+        target_data = DataFrame(input_data['target_data'])
+        metric_names = input_data['metrics']
+        interval_minutes = input_data['interval_minutes']
+
+        data_size = target_data.shape[0]
+
+        output_data = []
+
+        diff = target_data['target'] - target_data['prediction']
+        diff_squared = diff**2
+
+        self.info(f'Calculating simple metrics for model {model_id}: {metric_names}', metadata)
+
+        for metric in metric_names:
+            if metric == 'rmse':
+                output_data.append({'metric': 'rmse', 'value': np.sqrt(np.mean(diff_squared))})
+            elif metric == 'mse':
+                output_data.append({'metric': 'mse', 'value': np.mean(diff_squared)})
+            elif metric == 'mae':
+                output_data.append({'metric': 'mae', 'value': np.mean(np.abs(diff))})
+            elif metric == 'r2':
+                y_true = target_data['target']
+                y_mean = np.mean(y_true)
+
+                ss_res = np.sum(diff_squared)
+                ss_tot = np.sum((y_true - y_mean) ** 2)
+
+                # Evita divisão por zero
+                if ss_tot == 0:
+                    r2_score = 0.0
+                else:
+                    r2_score = 1 - (ss_res / ss_tot)
+
+                output_data.append({'metric': 'r2', 'value': r2_score})
+
+        data = DataFrame(output_data)
+        data['model_id'] = model_id
+        data['timestamp'] = target_data['timestamp'].max()
+        data['data_size'] = data_size
+        data['interval_minutes'] = interval_minutes
+
+        self._debug_dataframe(f'Simple metrics dataframe: Size {data.shape}', data, metadata)
+
+        return data.to_dict(orient='records')

laborious/activities/opc.py

@@ -0,0 +1,514 @@
+from temporalio import activity, workflow
+
+with workflow.unsafe.imports_passed_through():
+    import traceback
+    from collections.abc import Hashable
+    from typing import Any
+
+    from pandas import DataFrame
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.notifications.models import NotificationLevel
+    from sientia_do.observability.logger import Logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+
+    from laborious.utils.repository.opc_repository import OpcRepository
+
+OPC_WRITTING_ERROR_CONFIDENCE = 12
+OPC_SESSION_BAD_CONFIDENCE = 14
+OPC_SESSION_BAD_COMMENT_PREFIX = 'OPC UA session/channel error:'
+OPC_WRITTING_ERROR_MESSAGE = 'Some data could not be written to OPC servers'
+OPC_RECONNECT_IN_PROGRESS_COMMENT = 'OPC UA reconnect in progress'
+OPC_COMMENT_SEPARATOR = ' | '
+
+
+def _opc_session_bad_comment(opc_status: str | None) -> str:
+    status = opc_status or 'Unknown'
+    return f'{OPC_SESSION_BAD_COMMENT_PREFIX} {status}'
+
+
+def _apply_opc_write_error(
+    error_info: dict[str, Any] | None,
+    session_bad_seen: bool,
+    session_bad_status: str | None,
+    reconnect_in_progress_seen: bool,
+) -> tuple[bool, str | None, bool]:
+    """
+    Update session/reconnect flags from an OPC write error payload.
+
+    Args:
+        error_info: Repository error details, or None when the write succeeded.
+        session_bad_seen: Whether a session_bad error was seen so far.
+        session_bad_status: Last known OPC status for session errors.
+        reconnect_in_progress_seen: Whether reconnect_in_progress was seen so far.
+
+    Return:
+        Updated (session_bad_seen, session_bad_status, reconnect_in_progress_seen).
+    """
+    if not error_info:
+        return session_bad_seen, session_bad_status, reconnect_in_progress_seen
+
+    kind = error_info.get('opc_error_kind')
+    if kind == 'session_bad':
+        return True, error_info.get('opc_status', session_bad_status), reconnect_in_progress_seen
+    if kind == 'reconnect_in_progress':
+        return session_bad_seen, session_bad_status, True
+    return session_bad_seen, session_bad_status, reconnect_in_progress_seen
+
+
+class OPC(SientiaMonitoring):
+    """
+    OPC server integration activities for real-time data export.
+
+    This class provides comprehensive OPC UA client functionality for connecting
+    to multiple OPC servers and writing prediction data in real-time. It implements
+    secure communication with certificate-based authentication and automatic
+    reconnection capabilities.
+
+    The class supports multiple OPC servers with individual configurations and
+    provides robust error handling and monitoring for production environments.
+
+    Attributes:
+        opc_servers (dict): Configuration for multiple OPC servers
+        opc_repository (dict): Active OPC repository connections
+        logger (Logger): Logging and observability instance
+        notification_handler (NotificationHandler): Notification management instance
+    """
+
+    def __init__(
+        self,
+        opc_servers: dict[str, dict[str, Any]],
+        logger: Logger,
+        notification_handler: NotificationHandler,
+        metrics_controller: MetricsController,
+    ):
+        self.opc_servers = opc_servers
+
+        SientiaMonitoring.__init__(self, logger, notification_handler, metrics_controller)
+
+        self.opc_repository: dict[str, OpcRepository] = {}
+
+    def init_opc(self):
+        """
+        Initialize OPC server connections and establish communication channels.
+
+        This method iterates through all configured OPC servers and attempts to
+        establish secure connections using certificate-based authentication.
+        Each server connection is managed independently, and connection failures
+        are reported through the notification system.
+
+        The method performs the following operations:
+        1. Creates OpcRepository instances for each configured server
+        2. Establishes secure connections with certificate validation
+        3. Reports connection success/failure through notifications
+        4. Logs connection status for operational visibility
+
+        Raises:
+            Exception: If OPC repository initialization fails or connection
+                establishment encounters critical errors
+
+        Note:
+            Connection failures are logged and reported but do not prevent
+            the initialization of other OPC servers. Each server is handled
+            independently to ensure maximum availability.
+        """
+        self.info('Initializing OPC servers...')
+        for opc_id, server in self.opc_servers.items():
+            self.opc_repository[opc_id] = OpcRepository(
+                opc_id=opc_id,
+                server_name=server['server_name'],
+                url=server['url'],
+                logger=self.logger,
+                server_uri=server['server_uri'],
+                cert_path=server['cert_path'],
+                private_key_path=server['private_key_path'],
+                server_cert_path=server['server_cert_path'],
+                notification_handler=self.notification_handler,
+                reconnection_interval=server.get('reconnection_interval', 60),
+                metrics_controller=self.metrics_controller,
+            )
+            is_connected, error_data = self.opc_repository[opc_id].connect()
+            if not is_connected:
+                self.send_notification(
+                    metadata={
+                        'model_id': '-',
+                        'model_name': '-',
+                        'workflow_name': '-',
+                        'schedule_name': 'INITIALIZATION',
+                    },
+                    notification_id=error_data['notification_id'],
+                    message=error_data['message'],
+                    block=error_data['block'],
+                    level=error_data.get('level', NotificationLevel.ERROR),
+                    attachment_content=error_data.get('attachment_content', None),
+                )
+            else:
+                self.info(f'OPC server {opc_id}:{server["server_name"]} connected successfully.')
+
+    def write_data(
+        self,
+        server_id: str,
+        tag: str,
+        data: Any,
+        data_type: str,
+        tag_type: str,
+        metadata: dict[str, Any],
+    ) -> tuple[float | None, dict[str, Any] | None]:
+        """
+        Write data to a specific OPC server tag with comprehensive error handling.
+
+        Return:
+            tuple[float | None, dict[str, Any] | None]: Response time on success, or
+                (None, error info_data) on repository failure.
+        """
+
+        try:
+            is_success, info_data = self.opc_repository[server_id].write_data(
+                tag, data, data_type, metadata
+            )
+            if not is_success:
+                self.send_notification(
+                    metadata=metadata,
+                    notification_id=info_data['notification_id'],
+                    message=info_data['message'],
+                    block=info_data['block'],
+                    level=info_data.get('level', NotificationLevel.ERROR),
+                    attachment_content=info_data.get('attachment_content', None),
+                )
+                return None, info_data
+            return info_data['response_time'], None
+        except Exception as e:
+            trace = traceback.format_exc()
+            self.send_notification(
+                metadata=metadata,
+                notification_id=f'WRITE_OPC_{tag_type.upper()}_ERROR',
+                message=f'Error writing data to OPC server: {e}',
+                block='write_opc_data',
+                level=NotificationLevel.ERROR,
+                attachment_content=trace,
+            )
+            raise
+
+    def validate_server(self, server_id: str, metadata: dict[str, Any]) -> bool:
+        """
+        Validate that an OPC server is available and configured for write operations.
+
+        This method checks if the specified OPC server exists in the active
+        repository and is available for data writing operations. It provides
+        immediate feedback for server availability and logs validation failures
+        for operational monitoring.
+
+        Args:
+            server_id (str): Unique identifier for the OPC server to validate
+            metadata (dict[str, Any]): Context metadata for logging and notifications
+
+        Returns:
+            bool: True if server is available, False otherwise
+
+        Note:
+            Server validation failures are automatically reported through the
+            notification system with detailed information about available servers.
+            This helps operators quickly identify configuration issues.
+        """
+        if self.opc_repository.get(server_id) is None:
+            message = f'OPC server {server_id} not found to perform write operation.'
+            self.send_notification(
+                metadata=metadata,
+                notification_id='OPC_SERVER_NOT_FOUND',
+                message=message,
+                block='write_opc_data',
+                level=NotificationLevel.ERROR,
+                attachment_content=f'OPC servers: {list(self.opc_repository.keys())}',
+            )
+            return False
+        return True
+
+    def _write_tags_from_config(
+        self,
+        server_id: str,
+        tags_config: dict[str, dict[str, Any]],
+        data: DataFrame,
+        data_column: str,
+        tag_type: str,
+        log_label: str,
+        metadata: dict[str, Any],
+    ) -> tuple[dict[str, float | None], bool, str | None, bool]:
+        """
+        Write a group of OPC tags and collect response times and error flags.
+
+        Args:
+            server_id: Target OPC server identifier.
+            tags_config: Tag name to configuration mapping.
+            data: DataFrame with prediction/confidence columns.
+            data_column: Column name whose first row value is written.
+            tag_type: Tag category passed to write_data ('prediction' or 'confidence').
+            log_label: Human-readable label for success logs.
+            metadata: Context metadata for logging and notifications.
+
+        Return:
+            (response_times, session_bad_seen, session_bad_status, reconnect_in_progress_seen)
+        """
+        response_times: dict[str, float | None] = {}
+        session_bad_seen = False
+        session_bad_status: str | None = None
+        reconnect_in_progress_seen = False
+
+        for tag, tag_config in tags_config.items():
+            response_time, error_info = self.write_data(
+                server_id=server_id,
+                tag=tag,
+                data=data.head(1)[data_column].values[0],
+                data_type=tag_config['data_type'],
+                tag_type=tag_type,
+                metadata=metadata,
+            )
+            session_bad_seen, session_bad_status, reconnect_in_progress_seen = (
+                _apply_opc_write_error(
+                    error_info,
+                    session_bad_seen,
+                    session_bad_status,
+                    reconnect_in_progress_seen,
+                )
+            )
+            if response_time is not None:
+                self.info(
+                    f'{log_label} written to OPC server {server_id} for tag {tag}.',
+                    metadata,
+                )
+            response_times[tag] = response_time
+
+        return response_times, session_bad_seen, session_bad_status, reconnect_in_progress_seen
+
+    def manage_output_tags(
+        self,
+        server_id: str,
+        config: dict[str, Any],
+        data: DataFrame,
+        metadata: dict[str, Any],
+    ) -> tuple[bool, dict[str, float | None], bool, str | None, bool]:
+        """
+        Manage the writing of prediction and confidence data to OPC server tags.
+
+        This method orchestrates the writing of multiple data types to OPC servers
+        based on configuration. It handles both prediction data and confidence
+        values independently, allowing for flexible tag configuration and
+        comprehensive error handling.
+
+        The method supports two main tag types:
+        1. Prediction tags: Write actual prediction values to configured OPC tags
+        2. Confidence tags: Write confidence scores to separate OPC tags
+
+        Args:
+            server_id (str): Unique identifier for the target OPC server
+            config (dict[str, Any]): OPC tag configuration containing:
+                - prediction_tags (dict, optional): Prediction tag configurations
+                - confidence_tags (dict, optional): Confidence tag configurations
+            data (DataFrame): DataFrame containing prediction and confidence data
+            metadata (dict[str, Any]): Context metadata for logging and notifications
+            success (bool): Current success status to maintain across operations
+
+        Returns:
+            tuple[bool, int]: (overall_success, total_tags_written)
+                - overall_success: True if all configured tags were written successfully
+                - total_tags_written: Count of successfully written tags
+        """
+        response_times: dict[str, float | None] = {}
+        session_bad_seen = False
+        session_bad_status: str | None = None
+        reconnect_in_progress_seen = False
+
+        tag_groups = (
+            ('prediction_tags', 'prediction', 'prediction', 'Prediction data'),
+            ('confidence_tags', 'prediction_confidence', 'confidence', 'Confidence data'),
+        )
+        for config_key, data_column, tag_type, log_label in tag_groups:
+            if config_key not in config:
+                continue
+            (
+                group_times,
+                group_session_bad,
+                group_status,
+                group_reconnect,
+            ) = self._write_tags_from_config(
+                server_id=server_id,
+                tags_config=config[config_key],
+                data=data,
+                data_column=data_column,
+                tag_type=tag_type,
+                log_label=log_label,
+                metadata=metadata,
+            )
+            response_times.update(group_times)
+            if group_session_bad:
+                session_bad_seen = True
+                session_bad_status = group_status or session_bad_status
+            if group_reconnect:
+                reconnect_in_progress_seen = True
+
+        success = None not in response_times.values()
+        return (
+            success,
+            response_times,
+            session_bad_seen,
+            session_bad_status,
+            reconnect_in_progress_seen,
+        )
+
+    @activity.defn(name='write_opc_data')
+    def write_opc_data(
+        self, input_data: dict[str, Any]
+    ) -> tuple[dict[Hashable, Any], dict[str, dict[str, float | None]]]:
+        """
+        Write prediction and confidence data to OPC servers. The two writing
+        operations are optional and independent of each other.
+
+        Args:
+            - input_data(dict[str, Any]): The input data. Contains the following keys:
+                - data(dict[str, Any]): The dataframe that contains the data to write
+                to the OPC servers.
+                - opc_output_config(dict[str, Any]): The OPC writing configuration.
+                The keys are the OPC server names and the values contain:
+                    - prediction_tags(dict[str, Any]): The tags to write to the OPC servers.
+                    - confidence_tags(dict[str, Any]): The tags to write to the OPC servers.
+
+        Returns:
+            - dict[Any, Any]: The data that was written to the OPC servers.
+
+        """
+        metadata = input_data['metadata']
+        self.info('Writing data to OPC servers...', metadata)
+        data = DataFrame(input_data['data'])
+        opc_output_config = input_data['opc_output_config']
+        self.info(f'Data to write: {data.size} rows', metadata)
+
+        success = True
+        session_bad_seen = False
+        session_bad_status: str | None = None
+        reconnect_in_progress_seen = False
+
+        opc_metrics: dict[str, dict[str, float | None]] = {}
+
+        for server_id, config in opc_output_config.items():
+            if not self.validate_server(server_id, metadata):
+                success = False
+                continue
+
+            (
+                local_success,
+                local_response_times,
+                local_session_bad,
+                local_status,
+                local_reconnect_in_progress,
+            ) = self.manage_output_tags(server_id, config, data, metadata)
+            opc_metrics[server_id] = local_response_times
+            local_count = len(local_response_times)
+            success = success and local_success
+            if local_session_bad:
+                session_bad_seen = True
+                session_bad_status = local_status or session_bad_status
+            if local_reconnect_in_progress:
+                reconnect_in_progress_seen = True
+
+            n_pred = len(config.get('prediction_tags') or {})
+            n_conf = len(config.get('confidence_tags') or {})
+            self.info(
+                f'Process completed for OPC server {server_id}: {local_count} of {n_pred} prediction tags and {n_conf} confidence tags',
+                metadata,
+            )
+
+        return (
+            self.process_confidence(
+                data,
+                success,
+                metadata,
+                session_bad=session_bad_seen,
+                opc_status=session_bad_status,
+                reconnect_in_progress=reconnect_in_progress_seen,
+            ),
+            opc_metrics,
+        )
+
+    def process_confidence(
+        self,
+        data: DataFrame,
+        success: bool,
+        metadata: dict[str, Any],
+        *,
+        session_bad: bool = False,
+        opc_status: str | None = None,
+        reconnect_in_progress: bool = False,
+    ) -> dict[Hashable, Any]:
+        """
+        Process prediction confidence based on OPC write operation success.
+
+        This method updates the prediction confidence values in the DataFrame
+        based on the success status of OPC server write operations. If any
+        write operations failed, it sets the confidence to a predefined error
+        value to indicate data quality issues.
+
+        The method implements a confidence degradation strategy:
+        - Success: Maintains original confidence values
+        - Failure: Sets confidence to error value for operational awareness
+
+        Args:
+            data (DataFrame): DataFrame containing prediction and confidence data
+            success (bool): Overall success status of OPC write operations
+            metadata (dict[str, Any]): Context metadata for logging and notifications
+
+        Returns:
+            dict[Any, Any]: Processed data as a dictionary with updated confidence values
+
+        Note:
+            The error confidence value (OPC_WRITTING_ERROR_CONFIDENCE = 12) is
+            used to indicate that data was not successfully exported to OPC servers.
+            This allows downstream systems to handle data quality appropriately.
+        """
+
+        if not success:
+            comment_parts: list[str] = []
+            confidence = OPC_WRITTING_ERROR_CONFIDENCE
+
+            if session_bad:
+                comment_parts.append(_opc_session_bad_comment(opc_status))
+                confidence = OPC_SESSION_BAD_CONFIDENCE
+            if reconnect_in_progress:
+                comment_parts.append(OPC_RECONNECT_IN_PROGRESS_COMMENT)
+                confidence = OPC_SESSION_BAD_CONFIDENCE
+            if not comment_parts:
+                comment_parts.append(OPC_WRITTING_ERROR_MESSAGE)
+
+            comments = OPC_COMMENT_SEPARATOR.join(comment_parts)
+            data['prediction_confidence'] = confidence
+            data['comments'] = comments
+            self.debug(
+                f'OPC write issues, confidence={confidence}, comments={comments}',
+                metadata,
+            )
+        else:
+            self.debug('Data written to OPC servers successfully.', metadata)
+
+        return data.to_dict()
+
+    def close(self):
+        """
+        Gracefully shutdown all OPC server connections and cleanup resources.
+
+        This method ensures proper cleanup of all active OPC server connections
+        by calling the disconnect method on each repository instance. It's
+        designed to be called during application shutdown to prevent resource
+        leaks and ensure clean termination.
+
+        The method performs the following cleanup operations:
+        1. Iterates through all active OPC repository connections
+        2. Calls disconnect() on each repository instance
+        3. Allows for graceful connection termination
+        4. Prevents resource leaks and connection hanging
+
+        Note:
+            This method should be called during application shutdown to ensure
+            proper cleanup. It handles all active connections regardless of
+            their current state and provides a clean shutdown experience.
+        """
+        for opc in self.opc_repository.values():
+            opc.disconnect()
+        self.opc_repository.clear()

laborious/activities/storage.py

@@ -0,0 +1,231 @@
+from temporalio import activity, workflow
+
+with workflow.unsafe.imports_passed_through():
+    # Extend the Temporal Postgres activities for convenient query -> MinIO export
+    import traceback
+    from datetime import timedelta
+    from typing import Any
+
+    import pandas as pd
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.notifications.models import NotificationLevel
+    from sientia_do.observability.logger import Logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+    from sientia_do.repository.minio_repository_sync import MinioRepository
+    from sientia_do.temporal.activities.postgres_sync import Postgres
+    from sientia_do.temporal.constants import now
+
+    from laborious.utils.models.minio_dataframe_payload import MinioDataFramePayload
+
+_LOAD_QUERY_OFFLOAD_SKIP_KEYS = frozenset({'model_name', 'key_prefix', 'size_threshold_bytes'})
+
+
+class Storage(Postgres, SientiaMonitoring):
+    """
+    Extensions for Postgres activities with a helper to export query results
+    directly to MinIO as Parquet and return the object name.
+    """
+
+    minio_repository: MinioRepository | None = None
+
+    def __init__(
+        self,
+        host: str,
+        port: int,
+        user: str,
+        password: str,
+        dbname: str,
+        min_connections: int,
+        max_connections: int,
+        retention_hours: int = 24,
+        minio_repository: MinioRepository | None = None,
+        logger: Logger | None = None,
+        notification_handler: NotificationHandler | None = None,
+        metrics_controller: MetricsController | None = None,
+    ):
+        self.retention_hours = retention_hours
+        Postgres.__init__(
+            self,
+            host=host,
+            port=port,
+            user=user,
+            password=password,
+            dbname=dbname,
+            min_connections=min_connections,
+            max_connections=max_connections,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=metrics_controller,
+        )
+
+        self.minio_repository = minio_repository
+        SientiaMonitoring.__init__(
+            self,
+            logger=logger,
+            notification_handler=notification_handler,
+            metrics_controller=metrics_controller,
+        )
+
+    @activity.defn(name='load_query_with_minio_offload')
+    def load_query_with_minio_offload(self, input_data: dict[str, Any]) -> MinioDataFramePayload:
+        """
+        Run the custom SQL load, then return a MinIO-aware dataframe wire dict.
+
+        Args (input_data):
+            metadata (dict): Workflow metadata (same as load_custom_query).
+            query (str): SQL query.
+            datetime_columns (list[str], optional): Datetime column names.
+            model_name (str): Model name for object key basename.
+            key_prefix (str, optional): Directory prefix inside the bucket.
+            size_threshold_bytes (int, optional): Override env offload threshold.
+
+        Returns:
+            dict[str, Any]: Flat ``MinioDataFramePayload`` dict or ``success: False`` on failure.
+        """
+        if self.minio_repository is None:
+            raise ValueError('Minio repository not initialized')
+
+        metadata: dict = input_data.get('metadata', {})
+        model_name = input_data['model_name']
+
+        rows = self.load_custom_query(
+            input_data,
+        )
+        if not rows:
+            self.error(
+                'load_query_with_minio_offload failed: No data returned from query', metadata
+            )
+            dataframe = None
+        else:
+            dataframe = pd.DataFrame(rows)
+
+        return MinioDataFramePayload.from_dataframe(
+            dataframe,
+            minio_repo=self.minio_repository,
+            workflow_metadata=metadata,
+            model_name=model_name,
+            operation='initial',
+            logger=self.logger,
+        )
+
+    @activity.defn(name='export_payload_to_postgres')
+    def export_payload_to_postgres(self, input_data: dict[str, Any]) -> dict:
+        """
+        Resolve a MinIO-aware payload into a DataFrame and persist it into PostgreSQL.
+
+        This activity accepts the serialized payload produced by previous steps
+        (inline dict or MinIO object reference), reconstructs the tabular data,
+        and delegates the final write to ``export_data_to_postgres`` using the
+        same input contract expected by the Postgres activity mixin.
+
+        Args:
+         - input_data (dict[str, Any]): Activity input containing ``data`` as a
+           ``MinioDataFramePayload``-compatible dict plus database write options
+           (schema/table/on_conflict/metadata and related fields).
+
+        Return:
+         dict: Result dictionary returned by ``export_data_to_postgres``, including
+         success status and optional write diagnostics.
+        """
+        metadata = input_data.get('metadata')
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        data = payload.retrieve(self.minio_repository, metadata)
+
+        return self.export_data_to_postgres(
+            {
+                **input_data,
+                'data': data,
+            }
+        )
+
+    @activity.defn(name='cleanup_minio_objects_expired')
+    def cleanup_minio_objects_expired(self, input_data: dict[str, Any]) -> dict[str, Any]:
+        """
+        Delete objects under the given prefixes that are older than the retention window.
+
+        Args (input_data):
+            metadata (dict): Workflow metadata for logging and metrics.
+            prefixes (list[str]): Key prefixes to scan (one level or subtree per prefix).
+
+        Returns:
+            dict[str, Any]: ``success``, ``deleted_count``, and optional ``message``.
+        """
+        if self.minio_repository is None:
+            raise ValueError('Minio repository not initialized')
+
+        metadata = input_data.get('metadata', {})
+        payload = MinioDataFramePayload.from_dict(input_data['data'])
+        prefix = payload.cleanup_prefix()
+        base = now()
+        cutoff = (base.replace(tzinfo=None) if base.tzinfo else base) - timedelta(
+            hours=self.retention_hours
+        )
+
+        report: dict[str, Any] = {
+            'failed': {},
+            'deleted': {},
+            'failed_count': 0,
+            'deleted_count': 0,
+        }
+        try:
+            keys = self.minio_repository.list_objects(
+                prefix=prefix,
+                recursive=True,
+                metadata=metadata,
+            )
+            for key in keys:
+                try:
+                    ts = MinioDataFramePayload.parse_object_timestamp(key)
+                    if ts is None:
+                        continue
+                    if ts >= cutoff:
+                        continue
+                    self.minio_repository.delete_file(
+                        object_name=key,
+                        metadata=metadata,
+                    )
+                except Exception as e:
+                    report['failed'][key] = {
+                        'success': False,
+                        'message': str(e),
+                    }
+                    report['failed_count'] += 1
+                    continue
+                report['deleted'][key] = {
+                    'success': True,
+                    'message': 'Deleted',
+                }
+                report['deleted_count'] += 1
+        except Exception as e:
+            trace = traceback.format_exc()
+            self.send_notification(
+                metadata=metadata,
+                notification_id='ERROR_CLEANUP_MINIO_OBJECTS_EXPIRED',
+                message=f'Error cleaning up MinIO objects: {e}',
+                block='cleanup_minio_objects_expired',
+                level=NotificationLevel.ERROR,
+                attachment_content=trace,
+            )
+            self.error(trace, metadata)
+        else:
+            # Cleanup success is expected in normal flow; avoid noisy INFO notifications
+            # that do not impact behavior and can flood observability in test runs.
+            self.info('MinIO objects cleaned up successfully', metadata)
+
+        return report
+
+    def close(self) -> None:
+        """
+        Shutdown Storage resources in deterministic order.
+
+        The method first closes Postgres resources via ``Postgres.close`` (engine,
+        sessions, and monitoring hooks), then closes the optional MinIO repository
+        and clears the local reference to avoid accidental reuse after shutdown.
+        """
+        Postgres.close(self)
+        if self.minio_repository is not None:
+            try:
+                self.minio_repository.close()
+            finally:
+                self.minio_repository = None

laborious/metrics.py

@@ -0,0 +1,200 @@
+"""
+Laborious Metrics Module
+
+This module defines all Prometheus metrics used by the Sientia DataOps Laborious system
+for monitoring and observability. The metrics provide insights into system performance,
+prediction quality, and operational health.
+
+The metrics are designed to be scraped by Prometheus and can be visualized in
+Grafana or other monitoring dashboards to provide real-time visibility into
+the system's operation.
+
+Key Metric Categories:
+- Application Health: Overall system status and availability
+- Prediction Operations: Count and performance of prediction operations
+- Data Quality: Confidence levels and validation results
+- Export Operations: Database and OPC export performance
+- Response Times: Performance monitoring for various operations
+
+Metric Labels:
+- pod_id: Kubernetes pod identifier for multi-instance deployments
+- runtime: Runtime / environment identifier (matches ``RUNTIME`` env, see ``SientiaMonitoring``)
+- model_name: Name of the ML model being used
+- workflow_name: Name of the prediction pipeline
+- opc_server_id: Identifier for OPC server operations
+"""
+
+from prometheus_client import Counter, Gauge, Histogram
+from sientia_do.observability.metrics import CORE_LABELS
+
+# Application health metric
+APP_UP = Gauge(
+    'app_up',
+    'Indicates if the application is running (1) or shutting down (0)',
+    ['pod_id'],
+)
+
+# Prediction operation metrics
+PREDICTIONS_WRITTEN_COUNT = Counter(
+    'laborious_predictions_written_count',
+    'Number of predictions written to the database table predictions',
+    CORE_LABELS,
+)
+
+# Prediction quality metrics
+PREDICTION_CONFIDENCE_MONITOR = Gauge(
+    'laborious_prediction_confidence_monitor',
+    'Current confidence of each prediction',
+    CORE_LABELS,
+)
+
+# Prediction total response time
+PREDICTION_RESPONSE_TIME_MONITOR = Histogram(
+    'laborious_prediction_response_time_monitor',
+    'Current response time of each prediction',
+    CORE_LABELS,
+    buckets=[0.01, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10.0],
+)
+
+
+# ================== OPC metrics ==================
+
+
+PREDICTION_OPC_WRITING_COUNT = Counter(
+    'laborious_prediction_opc_writing_count',
+    'Number of predictions written to the OPC server',
+    [*CORE_LABELS, 'opc_server_id', 'tag'],
+)
+
+PREDICTION_OPC_WRITING_RESPONSE_TIME_MONITOR = Histogram(
+    'laborious_prediction_opc_writing_response_time_monitor',
+    'Current response time of each prediction written to the OPC server',
+    [*CORE_LABELS, 'opc_server_id', 'tag'],
+    buckets=[0.01, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10.0],
+)
+
+OPC_CONNECTIONS_TOTAL = Counter(
+    'opc_connections_initiated_total',
+    'Total connection attempts to OPC servers',
+    ['pod_id', 'server_name'],
+)
+OPC_CONNECTIONS_FAILED = Counter(
+    'opc_connections_failed_total',
+    'Total failed connection attempts to OPC servers',
+    ['pod_id', 'server_name'],
+)
+OPC_CONNECTION_STATUS = Gauge(
+    'opc_connection_status',
+    'Connection status with the OPC server (1=connected, 0=disconnected)',
+    ['pod_id', 'server_name', 'server_url'],
+)
+
+_OPC_SESSION_DEBUG_LABELS = ['pod_id', 'server_name', 'runtime', 'opc_server_id', 'session_id']
+
+OPC_SESSION_CREATED_TOTAL = Counter(
+    'opc_session_created_total',
+    'OPC UA sessions established (after successful connect)',
+    _OPC_SESSION_DEBUG_LABELS,
+)
+
+OPC_SESSION_CLOSED_TOTAL = Counter(
+    'opc_session_closed_total',
+    'OPC UA client disconnects completed (session tear-down initiated)',
+    _OPC_SESSION_DEBUG_LABELS,
+)
+
+OPC_SESSION_REVISED_TIMEOUT_MS = Gauge(
+    'opc_session_revised_timeout_milliseconds',
+    'Server-revised OPC UA session timeout (RevisedSessionTimeout) in ms after connect',
+    _OPC_SESSION_DEBUG_LABELS,
+)
+
+OPC_WRITE_ATTEMPT_LABELS = [*_OPC_SESSION_DEBUG_LABELS, 'model_id', 'model_name', 'result']
+
+OPC_WRITE_ATTEMPTS_TOTAL = Counter(
+    'opc_write_attempts_total',
+    'OPC UA write attempts with session and outcome (result=OK or exception class name)',
+    OPC_WRITE_ATTEMPT_LABELS,
+)
+
+OPC_WRITE_INTER_ARRIVAL_OVER_SESSION_TIMEOUT_TOTAL = Counter(
+    'opc_write_inter_arrival_over_session_timeout_total',
+    'Successful writes where seconds since the previous successful write exceeded RevisedSessionTimeout (ms)',
+    _OPC_SESSION_DEBUG_LABELS,
+)
+
+# ================== Model metrics ==================
+
+MODEL_READ_LAG = Histogram(
+    'laborious_model_read_lag',
+    'Lag between the start and read of read operations',
+    CORE_LABELS,
+    buckets=[0.01, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10.0],
+)
+
+MODEL_WRITE_LAG = Histogram(
+    'laborious_model_write_lag',
+    'Lag between the start and end of write operations',
+    CORE_LABELS,
+    buckets=[0.01, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10.0],
+)
+
+MODEL_READ_COUNT = Counter(
+    'laborious_model_read_count',
+    'Number of reads from the model',
+    CORE_LABELS,
+)
+
+MODEL_WRITE_COUNT = Counter(
+    'laborious_model_write_count',
+    'Number of writes to the model',
+    CORE_LABELS,
+)
+
+MODEL_READ_ERROR_COUNT = Counter(
+    'laborious_model_read_error_count',
+    'Number of errors reading from the model',
+    CORE_LABELS,
+)
+
+MODEL_WRITE_ERROR_COUNT = Counter(
+    'laborious_model_write_error_count',
+    'Number of errors writing to the model',
+    CORE_LABELS,
+)
+
+MODEL_ANALYZE_LAG = Histogram(
+    'laborious_model_analyze_lag',
+    'Lag between the start and end of analyze operations',
+    CORE_LABELS,
+    buckets=[0.01, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10.0],
+)
+
+MODEL_ANALYZE_COUNT = Counter(
+    'laborious_model_analyze_count',
+    'Number of analyze operations',
+    CORE_LABELS,
+)
+
+MODEL_ANALYZE_ERROR_COUNT = Counter(
+    'laborious_model_analyze_error_count',
+    'Number of errors during analyze operations',
+    CORE_LABELS,
+)
+
+
+# ================== PI Web API metrics ==================
+
+PI_WEB_API_LABELS = [*CORE_LABELS, 'tag_name']
+
+PI_WEB_API_PREDICTION_WRITTEN_COUNT = Counter(
+    'laborious_pi_web_api_prediction_written_count',
+    'Number of predictions written to the PI Web API',
+    PI_WEB_API_LABELS,
+)
+
+PI_WEB_API_PREDICTION_WRITTEN_ERROR_COUNT = Counter(
+    'laborious_pi_web_api_prediction_written_error_count',
+    'Number of errors writing predictions to the PI Web API',
+    PI_WEB_API_LABELS,
+)

laborious/utils/connectors_config.py

@@ -0,0 +1,127 @@
+import json
+from os import getenv
+from typing import Any
+
+
+def build_mlflow_config() -> dict[str, Any]:
+    """
+    Read MLflow tracking and registry credentials from the environment.
+
+    Used by ``Activities`` when constructing ``SientiaMLflowRepository``. The ``url`` value is the
+    same string workers and notebooks should use for ``MLFLOW_TRACKING_URI``-style clients.
+
+    Environment Variables:
+        MLFLOW_URL: Host with scheme
+        MLFLOW_USERNAME: Basic-auth or service user (default: aignosi)
+        MLFLOW_PASSWORD: Password or token (default: aignosi)
+
+    Return:
+        dict[str, Any]: ``url``, ``username``, ``password``.
+    """
+    return {
+        'url': getenv('MLFLOW_URL', 'http://localhost:5080'),
+        'username': getenv('MLFLOW_USERNAME', 'aignosi'),
+        'password': getenv('MLFLOW_PASSWORD', 'aignosi'),
+    }
+
+
+def build_plugin_store_config() -> dict[str, Any]:
+    """
+    Collect settings for ``PluginStore`` (Git-backed catalog + runtime install via pip).
+
+    Mirrors the model-manager service: the worker passes these kwargs into ``PluginStore`` after
+    ``install_runtime`` resolves wheels from the configured PyPI index. Missing optional env vars
+    become ``None`` so the store can run without auth in local dev.
+
+    Environment Variables:
+        STORE_BASE_URL: Git HTTP(S) server (e.g. Gitea) base URL (default: http://localhost:3000)
+        STORE_OWNER: Namespace or org owning the store repo (default: sientia)
+        STORE_REPO: Repository name (default: model-library-store)
+        STORE_BRANCH: Checkout branch; unset lets the client use default
+        STORE_USERNAME / STORE_PASSWORD: HTTP basic credentials for Git fetch
+        STORE_CACHE_TTL_SECONDS: Optional integer seconds for metadata cache TTL
+        PYPI_SERVER: Index URL for ``pip install`` during runtime install (default: http://localhost:5000)
+        PYPI_USERNAME / PYPI_PASSWORD: Optional index authentication
+
+    Return:
+        dict[str, Any]: Keys aligned with ``PluginStore`` constructor parameter names.
+    """
+
+    cache_ttl_seconds = getenv('STORE_CACHE_TTL_SECONDS')
+    return {
+        'base_url': getenv('STORE_BASE_URL', 'http://localhost:3000'),
+        'owner': getenv('STORE_OWNER', 'sientia'),
+        'repo': getenv('STORE_REPO', 'model-library-store'),
+        'username': getenv('STORE_USERNAME'),
+        'password': getenv('STORE_PASSWORD'),
+        'branch': getenv('STORE_BRANCH'),
+        'cache_ttl_seconds': int(cache_ttl_seconds) if cache_ttl_seconds else None,
+        'pypi_index_url': getenv('PYPI_SERVER', 'http://localhost:5000'),
+        'pypi_username': getenv('PYPI_USERNAME'),
+        'pypi_password': getenv('PYPI_PASSWORD'),
+    }
+
+
+def build_opc_config() -> dict[str, Any]:
+    """
+    Build OPC server configuration from environment variables.
+
+    This function constructs an OPC server configuration dictionary from
+    environment variables. It supports both single server and multi-server
+    configurations with flexible parameter handling.
+
+    Environment Variables:
+        OPC_CONFIG: JSON string containing multiple OPC server configurations
+        OPC_ID: OPC server ID (fallback, default: 1)
+        OPC_URL: Single OPC server URL (fallback, default: opc.tcp://localhost:4840)
+        OPC_SERVER_URI: Single OPC server URI (fallback, default: opc.tcp://localhost:4840)
+        OPC_CERT_PATH: Client certificate path (fallback, default: None)
+        OPC_PRIVATE_KEY_PATH: Client private key path (fallback, default: None)
+        OPC_SERVER_CERT_PATH: Server certificate path (fallback, default: None)
+        OPC_RECONNECTION_INTERVAL: Reconnection interval in seconds (fallback, default: 120)
+
+    Returns:
+        dict: OPC server configuration dictionary
+    """
+    opc_raw = getenv('OPC_CONFIG', None)
+
+    if opc_raw:
+        return json.loads(opc_raw)
+
+    return {
+        getenv('OPC_ID', '1'): {
+            'id': getenv('OPC_ID', '1'),
+            'server_name': getenv('OPC_SERVER_NAME', 'default_server'),
+            'url': getenv('OPC_URL', 'opc.tcp://localhost:4840'),
+            'server_uri': getenv('OPC_SERVER_URI', 'opc.tcp://localhost:4840'),
+            'cert_path': getenv('OPC_CERT_PATH', None),
+            'private_key_path': getenv('OPC_PRIVATE_KEY_PATH', None),
+            'server_cert_path': getenv('OPC_SERVER_CERT_PATH', None),
+            'reconnection_interval': int(getenv('OPC_RECONNECTION_INTERVAL', '120')),
+        }
+    }
+
+
+def build_minio_config() -> dict[str, Any]:
+    """
+    Build MinIO (S3-compatible) configuration from environment variables.
+
+    Environment Variables:
+        MINIO_ENDPOINT_URL: Host:port or URL for the S3 API (default: http://localhost:9000)
+        MINIO_ACCESS_KEY: Access key (default: minioadmin)
+        MINIO_SECRET_KEY: Secret key (default: minioadmin)
+        MINIO_DEFAULT_BUCKET: Default bucket for Laborious payloads (default: laborious)
+        MINIO_RETENTION_HOURS: Offloaded object retention window (default: 24)
+        MINIO_SECURE: If ``true``, use HTTPS (default: false)
+
+    Return:
+        dict[str, Any]: Keys consumed by ``Activities`` / ``MinioRepository``.
+    """
+    return {
+        'endpoint_url': getenv('MINIO_ENDPOINT_URL', 'http://localhost:9000'),
+        'access_key': getenv('MINIO_ACCESS_KEY', 'minioadmin'),
+        'secret_key': getenv('MINIO_SECRET_KEY', 'minioadmin'),
+        'default_bucket': getenv('MINIO_DEFAULT_BUCKET', 'laborious'),
+        'retention_hours': int(getenv('MINIO_RETENTION_HOURS', '24')),
+        'secure': getenv('MINIO_SECURE', 'false') == 'true',
+    }

laborious/utils/dataframe_debug.py

@@ -0,0 +1,34 @@
+from typing import Any
+
+from pandas import DataFrame
+
+DEFAULT_MAX_DEBUG_DATAFRAME_ROWS = 100
+
+
+def build_dataframe_debug_message(
+    message: str,
+    data: Any,
+    max_rows: int = DEFAULT_MAX_DEBUG_DATAFRAME_ROWS,
+) -> str:
+    """
+    Build a safe debug message for dataframe payloads
+
+    Args:
+     - message (str): Base message to identify the logged payload
+     - data (Any): Payload to evaluate for dataframe-aware logging
+     - max_rows (int): Maximum dataframe row count allowed for full payload logging
+
+    Return:
+     Formatted debug message with full dataframe content or compact summary
+    """
+    if not isinstance(data, DataFrame):
+        return f'{message} {data}'
+
+    rows = data.shape[0]
+    if rows <= max_rows:
+        return f'{message}\n{data.to_csv()}'
+
+    return (
+        f'{message} skipped because dataframe has {rows} rows '
+        f'(max: {max_rows}). Shape: {data.shape}'
+    )

laborious/utils/filters/conditional_filters.py

@@ -0,0 +1,48 @@
+from pandas import DataFrame
+
+
+def filter_specific_variables_null_values(data: DataFrame, config: dict) -> bool:
+    """
+    Filter to check if specific variables contain null values.
+
+    This function examines a DataFrame to determine if any of the specified variables
+    contain null (NaN) values. It returns True if null values are found for any of
+    the specified variables, False otherwise.
+
+    Args:
+        data (DataFrame): The pandas DataFrame to be examined. Must contain columns
+            named 'variable' and 'value'.
+        config (dict): Configuration dictionary containing the following key:
+            - variables (list): List of variable names to check for null values
+
+    Returns:
+        bool: True if any of the specified variables contain null values,
+              False if none of the specified variables contain null values.
+
+    """
+
+    if data.empty:
+        return False
+
+    return not data[data['variable'].isin(config['variables']) & data['value'].isna()].empty
+
+
+def filter_empty_data(data: DataFrame, _config: dict) -> bool:
+    """
+    Filter to check if the DataFrame is empty.
+
+    This function determines whether the provided DataFrame contains any data.
+    It's a simple utility function that can be used in conditional logic to
+    handle cases where no data is available.
+
+    Args:
+        data (DataFrame): The pandas DataFrame to be checked for emptiness.
+        _config (dict): Configuration dictionary (unused in this function).
+            The underscore prefix indicates this parameter is required for
+            interface consistency but not used in the implementation.
+
+    Returns:
+        bool: True if the DataFrame is empty (has no rows), False if it contains data.
+
+    """
+    return data.empty

laborious/utils/filters/mlflow_filters.py

@@ -0,0 +1,64 @@
+import numpy as np
+from pandas import DataFrame
+
+
+def api_error_filter(response: dict, _config: dict) -> bool:
+    """
+    Filter MLFlow API responses for error conditions.
+
+    This function analyzes MLFlow API responses to detect error conditions
+    and determine if the response should be filtered out due to quality
+    or reliability issues.
+
+
+    Args:
+        response: MLFlow API response data (dict)
+        _config: Filter configuration dictionary
+            Required keys:
+                - error_codes (list, optional): List of error codes to detect
+                - error_keywords (list, optional): List of error keywords to detect
+                - check_structure (bool, optional): Whether to validate response structure
+
+    Returns:
+        bool: True if data should be filtered (contains errors), False otherwise
+
+    """
+    if not response:
+        return True
+
+    if not response['success']:
+        return True
+
+    return False
+
+
+def nan_values_filter(predictions: DataFrame, _config: dict) -> bool:
+    """
+    Filter data for NaN (Not a Number) values.
+
+    This function detects NaN values in MLFlow prediction results and
+    determines if the data quality is sufficient for further processing
+    or export operations.
+
+    Args:
+        predictions: DataFrame containing prediction data to check for NaN values
+        _config: Filter configuration dictionary
+            Required keys:
+                - max_nan_ratio (float, optional): Maximum allowed NaN value ratio (0.0 to 1.0)
+                - max_nan_count (int, optional): Maximum allowed NaN value count
+                - check_nested (bool, optional): Whether to check nested data structures
+
+    Returns:
+        bool: True if data should be filtered (too many NaN values), False otherwise
+
+    """
+    data = (
+        predictions.replace({None: np.nan})
+        .drop(columns=['timestamp'], errors='ignore')
+        .infer_objects()
+    )
+
+    if data.isna().all().all():
+        return True
+
+    return False

laborious/utils/models/minio_dataframe_payload.py

@@ -0,0 +1,362 @@
+"""
+MinIO-backed DataFrame payload for Temporal workflows.
+
+Data is never stored as a pandas ``DataFrame`` field on the dataclass.
+Instead, the DataFrame is only provided as an input to:
+`from_dataframe` / `from_dataframe_to_dict`.
+
+At build time, the DataFrame is evaluated for its serialized size; if it exceeds
+the configured threshold, it is serialized to parquet bytes and uploaded to MinIO.
+Otherwise, it is inlined as a Temporal-friendly ``dict``.
+"""
+
+import pickle
+import re
+from collections.abc import Hashable
+from dataclasses import dataclass
+from datetime import datetime
+from io import BytesIO
+from os import getenv
+from typing import Any, Literal
+
+from pandas import DataFrame, read_parquet
+from sientia_do.observability.logger import Logger
+from sientia_do.repository.minio_repository_sync import MinioRepository
+from sientia_do.temporal.constants import DATETIME_FORMAT_FILENAME, DATETIME_FORMAT_WITH_TZ, now
+
+# Keys that are part of the serialized wire format (not arbitrary metadata).
+_SERIALIZED_FIELD_KEYS = frozenset({'data', 'bucket', 'object_key', 'object_prefix', 'uri'})
+
+_OBJECT_TIMESTAMP_PATTERN = re.compile(
+    r'-(?:initial|transform)-(\d{4}-\d{2}-\d{2}_\d{2}-\d{2}-\d{2})\.parquet$'
+)
+
+OFFLOAD_THRESHOLD_BYTES = int(
+    float(getenv('SIENTIA_MINIO_OFFLOAD_THRESHOLD_MEGABYTES', '1.5')) * 1024 * 1024
+)
+
+# Relative prefix used for storing offloaded prediction datasets in MinIO.
+# It is also the root directory for retention cleanup listing.
+PREDICTION_DATASETS_PREFIX = 'prediction_datasets'
+
+OperationKind = Literal['initial', 'transform', 'predict']
+
+
+def _build_object_key(
+    model_name: str, operation: OperationKind, timestamp: str
+) -> tuple[str, str | None]:
+    """
+    Build the MinIO object key and the directory prefix used for retention listing.
+
+    Args:
+        model_name: Registered model name used in the pipeline.
+        operation: Either initial (pre-transform load) or transform (post-MLFlow transform).
+        timestamp: Filename timestamp segment from DATETIME_FORMAT_FILENAME.
+
+    Return:
+        tuple[str, str | None]: Full object key and normalized prefix (or None if at bucket root).
+    """
+    # Naming convention:
+    #   - Directory is always `prediction_datasets/<model_name>`
+    #   - Filename follows the retention-parsing pattern
+    basename = f'{model_name}-{operation}-{timestamp}.parquet'
+    model_dir = model_name.strip().strip('/')
+    prefix = f'{PREDICTION_DATASETS_PREFIX}/{model_dir}'
+    return f'{prefix}/{basename}', prefix
+
+
+@dataclass
+class MinioDataFramePayload:
+    """
+    Serializable payload after a DataFrame was evaluated: inline tabular dict and/or MinIO keys.
+
+    Build from a live DataFrame only via `from_dataframe` / `from_dataframe_to_dict`.
+    Rehydrate from Temporal via `from_dict`. The DataFrame is not a field on this class.
+    """
+
+    last_timestamp: str
+    status: dict[str, Any] | None = None
+    data: dict[Hashable, Any] | None = None
+    bucket: str | None = None
+    object_key: str | None = None
+    object_prefix: str | None = None
+    uri: str | None = None
+
+    @staticmethod
+    def _debug(
+        logger: Logger | None,
+        message: str,
+        metadata: dict[str, Any] | None = None,
+    ) -> None:
+        """
+        Emit a debug message only when a logger instance is available.
+
+        Args:
+         - logger (Logger | None): Logger instance used for debug messages
+         - message (str): Message to be logged
+         - metadata (dict[str, Any] | None): Optional workflow metadata context
+        """
+        if logger is None:
+            return
+        logger.custom_debug(message, metadata)
+
+    @classmethod
+    def from_dict(cls, raw: 'dict[str, Any] | MinioDataFramePayload') -> 'MinioDataFramePayload':
+        """
+        Reconstruct a MinioDataFramePayload from a plain dict produced by Temporal serialization.
+
+        Temporal converts dataclass return values into plain dicts when crossing
+        workflow/activity boundaries. This method rebuilds the typed instance so
+        that methods like ``retrieve``, ``cleanup_prefix`` and ``has_data`` are
+        available on the receiving side.
+
+        If the argument is already a MinioDataFramePayload, it is returned as-is.
+
+        Args:
+            raw: Dict with keys matching the dataclass fields
+                (last_timestamp, status, data, bucket, object_key, object_prefix, uri),
+                or an existing MinioDataFramePayload instance.
+
+        Return:
+            MinioDataFramePayload: Reconstructed (or original) instance.
+        """
+        if isinstance(raw, MinioDataFramePayload):
+            return raw
+        return cls(
+            last_timestamp=raw['last_timestamp'],
+            status=raw.get('status'),
+            data=raw.get('data'),
+            bucket=raw.get('bucket'),
+            object_key=raw.get('object_key'),
+            object_prefix=raw.get('object_prefix'),
+            uri=raw.get('uri'),
+        )
+
+    @staticmethod
+    def estimate_size_bytes(
+        df: DataFrame,
+        metadata: dict[str, Any] | None = None,
+        logger: Logger | None = None,
+    ) -> int:
+        """
+        Approximate serialized size of the DataFrame as the default-orient dict.
+
+        Args:
+            df: DataFrame whose tabular content size is estimated.
+
+        Return:
+            int: Estimated size in bytes (pickle of dict representation).
+        """
+        try:
+            size = len(pickle.dumps(df.to_dict()))
+        except Exception:
+            size = len(pickle.dumps(df))
+
+        MinioDataFramePayload._debug(
+            logger,
+            f'DataFrame size: {size} bytes',
+            metadata,
+        )
+        return size
+
+    @staticmethod
+    def parse_object_timestamp(object_key: str) -> datetime | None:
+        """
+        Parse the timestamp embedded in the object key basename (before .parquet).
+
+        Args:
+            object_key: S3/MinIO object key whose basename follows
+                ``{model}-{initial|transform}-{DATETIME_FORMAT_FILENAME}.parquet``.
+
+        Return:
+            datetime | None: Parsed UTC-naive datetime from the key, or None if not matched.
+        """
+        basename = object_key.rsplit('/', 1)[-1]
+        match = _OBJECT_TIMESTAMP_PATTERN.search(basename)
+        if not match:
+            return None
+        try:
+            return datetime.strptime(match.group(1), DATETIME_FORMAT_FILENAME)
+        except ValueError:
+            return None
+
+    def cleanup_prefix(self) -> str | None:
+        """
+        Return the MinIO prefix eligible for retention cleanup.
+
+        Cleanup is only applicable when payload data was offloaded to MinIO
+        (``object_key`` present and inline ``data`` absent). Inline-only payloads
+        return ``None`` because there is no object tree to prune.
+
+        Return:
+         str | None: Prefix used by cleanup listing, or ``None`` when cleanup does not apply.
+        """
+        if self.object_key is not None and self.data is None:
+            return self.object_prefix
+        return None
+
+    def has_data(self) -> bool:
+        """
+        Indicate whether the payload contains retrievable tabular content.
+
+        A payload is considered non-empty when either inline ``data`` exists
+        (and is not an empty dict) or an ``object_key`` is available for MinIO
+        download.
+
+        Return:
+         bool: ``True`` when data can be retrieved, ``False`` otherwise.
+        """
+        return (self.data is not None and self.data != {}) or self.object_key is not None
+
+    @classmethod
+    def from_dataframe(
+        cls,
+        dataframe: DataFrame | None,
+        minio_repo: MinioRepository,
+        model_name: str,
+        operation: OperationKind,
+        status: dict[str, Any] | None = None,
+        workflow_metadata: dict | None = None,
+        last_timestamp: str | None = None,
+        logger: Logger | None = None,
+    ) -> 'MinioDataFramePayload':
+        """
+        Evaluate the DataFrame size, then either inline dict or upload parquet to MinIO.
+
+        The DataFrame is not stored on the returned instance.
+
+        Args:
+            dataframe: Tabular data to evaluate and persist (inline or MinIO).
+            metadata: Small metadata dict merged into the payload (e.g. success, message).
+            minio_repo: sientia_do MinioRepository (or compatible) with `upload_file()`.
+            workflow_metadata: Metadata passed to MinIO store for logging/metrics.
+            model_name: Registered model name used in the object basename.
+            operation: Either ``initial`` (query load) or ``transform`` (post-transform).
+            key_prefix: Backward-compatible parameter (currently ignored for object naming).
+            size_threshold_bytes: Byte limit before offload. When None, the module-level
+                environment-derived default is used.
+
+        Return:
+            MinioDataFramePayload: Instance with data and/or MinIO fields set.
+        """
+
+        if dataframe is None or dataframe.empty:
+            cls._debug(
+                logger,
+                'MinioDataFramePayload.from_dataframe received empty dataframe, returning empty payload',
+                workflow_metadata,
+            )
+            return cls(
+                data=None, last_timestamp=now().strftime(DATETIME_FORMAT_WITH_TZ), status=status
+            )
+
+        if last_timestamp is None:
+            last_timestamp = max(dataframe['timestamp'].values.tolist())
+
+        dataframe_size = cls.estimate_size_bytes(dataframe, workflow_metadata, logger)
+        cls._debug(
+            logger,
+            (
+                f'MinioDataFramePayload.from_dataframe estimated size: {dataframe_size} bytes '
+                f'(threshold: {OFFLOAD_THRESHOLD_BYTES} bytes)'
+            ),
+            workflow_metadata,
+        )
+
+        if dataframe_size <= OFFLOAD_THRESHOLD_BYTES:
+            cls._debug(
+                logger,
+                'MinioDataFramePayload.from_dataframe using inline payload',
+                workflow_metadata,
+            )
+            return cls(data=dataframe.to_dict(), last_timestamp=last_timestamp, status=status)
+
+        timestamp = now().strftime(DATETIME_FORMAT_FILENAME)
+        object_key, object_prefix = _build_object_key(model_name, operation, timestamp)
+        cls._debug(
+            logger,
+            (
+                'MinioDataFramePayload.from_dataframe offloading payload to MinIO '
+                f'with key {object_key}'
+            ),
+            workflow_metadata,
+        )
+
+        # Upload using the relative object key. The upstream repository will
+        # prefix it internally under its MinIO namespace.
+        parquet_buffer = BytesIO()
+        dataframe.to_parquet(parquet_buffer, engine='pyarrow', index=True)
+        file_bytes = parquet_buffer.getvalue()
+
+        upload_result = minio_repo.upload_file(
+            file_bytes=file_bytes,
+            relative_key=object_key,
+            metadata=workflow_metadata,
+        )
+
+        bucket = minio_repo.bucket
+        object_key_full = upload_result.get('minio_object_name', object_key)
+        uri = f's3://{bucket}/{object_key_full}' if bucket else None
+        cls._debug(
+            logger,
+            f'MinioDataFramePayload.from_dataframe upload completed: {uri}',
+            workflow_metadata,
+        )
+
+        return cls(
+            data=None,
+            bucket=bucket,
+            object_key=object_key_full,
+            object_prefix=object_prefix,
+            uri=uri,
+            last_timestamp=last_timestamp,
+            status=status,
+        )
+
+    def retrieve(
+        self,
+        minio_repo: MinioRepository,
+        workflow_metadata: dict[str, Any] | None = None,
+        logger: Logger | None = None,
+    ) -> DataFrame:
+        """
+        Load parquet from MinIO when object_key is set and populate inline data.
+
+        Args:
+            minio_repo: sientia_do MinioRepository (or compatible) with download_file().
+            workflow_metadata: Metadata passed to MinIO read for logging/metrics.
+
+        Return:
+            dict[str, Any]: Flat dict with data filled (same keys as to_dict after load).
+        """
+        if self.data is not None:
+            self._debug(
+                logger,
+                'MinioDataFramePayload.retrieve using inline payload data',
+                workflow_metadata,
+            )
+            return DataFrame(self.data)
+
+        if not self.has_data():
+            self._debug(
+                logger,
+                'MinioDataFramePayload.retrieve found no payload data, returning empty dataframe',
+                workflow_metadata,
+            )
+            return DataFrame()
+
+        self._debug(
+            logger,
+            f'MinioDataFramePayload.retrieve downloading object from MinIO: {self.object_key}',
+            workflow_metadata,
+        )
+        file_bytes = minio_repo.download_file(
+            object_name=self.object_key, metadata=workflow_metadata
+        )
+        df = read_parquet(BytesIO(file_bytes))
+        self._debug(
+            logger,
+            f'MinioDataFramePayload.retrieve loaded dataframe from MinIO with shape {df.shape}',
+            workflow_metadata,
+        )
+        return df

laborious/utils/repository/opc_repository.py

@@ -0,0 +1,893 @@
+"""
+Synchronous OPC UA client repository using asyncua ``sync`` API.
+
+``asyncua.sync.Client`` runs the asyncio stack on a background thread so Temporal
+activities and other callers stay blocking while preserving the same session
+lifecycle, security policy, reconnect semantics, and write error classification
+as the async ``main`` implementation at ``fcc8920a8be4`` (async → sync/thread conversion).
+"""
+
+import json
+import threading
+import time
+import traceback
+from datetime import datetime
+from pathlib import Path
+from typing import Any
+
+from asyncua import ua
+from asyncua.crypto import security_policies
+from asyncua.sync import Client
+from asyncua.ua.uaerrors import UaStatusCodeError
+from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+from sientia_do.notifications.models import NotificationLevel
+from sientia_do.observability.logger import Logger
+from sientia_do.observability.metrics_controller import MetricsController
+from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+
+from laborious import metrics
+
+# Requested session and secure channel lifetime (ms) before server revision; 10 minutes.
+OPC_UA_SESSION_AND_CHANNEL_TIMEOUT_MS = 10 * 60 * 1000
+
+
+class OpcClientAlreadyExistsError(RuntimeError):
+    """Raised when _create_client is called while self.client is already set."""
+
+
+class OpcSessionAlreadyConnectedError(RuntimeError):
+    """Raised when _open_session is called while a UA session is already open."""
+
+
+class OpcClientNotInitializedError(RuntimeError):
+    """Raised when _open_session is called before _create_client."""
+
+
+RECONNECTABLE_OPC_BAD_NAMES: frozenset[str] = frozenset(
+    {
+        'BadSessionIdInvalid',
+        'BadSessionClosed',
+        'BadSessionNotActivated',
+        'BadSecureChannelIdInvalid',
+        'BadSecureChannelClosed',
+        'BadSecureChannelTokenUnknown',
+        'BadTcpSecureChannelUnknown',
+        'BadServerNotConnected',
+        'BadConnectionClosed',
+        'BadDisconnect',
+        'BadConnectionRejected',
+        'BadCommunicationError',
+        'BadRequestInterrupted',
+        'BadUnknownResponse',
+        'BadTimeout',
+        'BadRequestTimeout',
+        'BadSequenceNumberInvalid',
+        'BadSequenceNumberUnknown',
+        'BadSecurityModeInsufficient',
+        'BadRequestHeaderInvalid',
+        'BadInvalidState',
+    }
+)
+
+
+def _opc_authentication_token_str(client: Client | None) -> str:
+    """
+    Serialize the current OPC UA authentication token (session handle) for logging and metrics.
+
+    Return:
+        str: Token string, or "unknown" if unavailable.
+    """
+    if client is None:
+        return 'unknown'
+    try:
+        proto = client.aio_obj.uaclient.protocol
+        if proto is None:
+            return 'unknown'
+        tok = getattr(proto, 'authentication_token', None)
+        if tok is None:
+            return 'unknown'
+        return str(tok)
+    except Exception:
+        return 'unknown'
+
+
+def _opc_status_from_exception(exc: BaseException) -> str:
+    """
+    Resolve OPC UA status name from an exception, including chained UaStatusCodeError causes.
+
+    Args:
+        exc (BaseException): Raised error from asyncua.
+
+    Return:
+        str: Status class name or generic Python exception name.
+    """
+    current: BaseException | None = exc
+    while current is not None:
+        if isinstance(current, UaStatusCodeError):
+            return type(current).__name__
+        current = current.__cause__
+    return type(exc).__name__
+
+
+def is_reconnectable_opcua_bad(exc: BaseException) -> bool:
+    """
+    Return whether the exception is a Tier-1 OPC UA Bad* that should trigger reconnect.
+
+    Args:
+        exc (BaseException): Raised error from get_node or write_value.
+
+    Return:
+        bool: True if reconnect should be scheduled.
+    """
+    return _opc_status_from_exception(exc) in RECONNECTABLE_OPC_BAD_NAMES
+
+
+def _model_labels_from_write_metadata(metadata: dict[str, Any] | None) -> dict[str, str]:
+    """
+    Extract model_id and model_name from write metadata for Prometheus labels.
+
+    Args:
+        metadata (dict[str, Any] | None): Context passed into write_data; may omit keys.
+
+    Return:
+        dict[str, str]: Labels model_id and model_name, defaulting to "unknown".
+    """
+    if not metadata:
+        return {'model_id': 'unknown', 'model_name': 'unknown'}
+    return {
+        'model_id': str(metadata.get('model_id', 'unknown')),
+        'model_name': str(metadata.get('model_name', 'unknown')),
+    }
+
+
+data_type_map = {
+    'float': {
+        'converter': float,
+        'opc_type': ua.VariantType.Float,
+    },
+    'double': {
+        'converter': float,
+        'opc_type': ua.VariantType.Double,
+    },
+    'int': {
+        'converter': int,
+        'opc_type': ua.VariantType.Int32,
+    },
+    'bool': {
+        'converter': bool,
+        'opc_type': ua.VariantType.Boolean,
+    },
+    'str': {
+        'converter': str,
+        'opc_type': ua.VariantType.String,
+    },
+}
+
+
+class OpcRepository(SientiaMonitoring):
+    """
+    Synchronous OPC UA repository for connect/disconnect and typed writes.
+
+    Uses ``asyncua.sync.Client`` with the same session metrics, Tier-1 Bad* reconnect,
+    and structured write error payloads as the async repository on ``main``.
+    """
+
+    def __init__(
+        self,
+        opc_id: str,
+        url: str,
+        server_name: str,
+        logger: Logger,
+        notification_handler: NotificationHandler,
+        metrics_controller: MetricsController,
+        reconnection_interval: int = 60,
+        server_uri: str | None = None,
+        cert_path: str | None = None,
+        private_key_path: str | None = None,
+        server_cert_path: str | None = None,
+    ):
+        self.url = url
+        self.id = opc_id
+        self.server_name = server_name
+        self.server_uri = server_uri
+        self.cert_path = cert_path
+        self.private_key_path = private_key_path
+        self.server_cert_path = server_cert_path
+        self.reconnection_interval = reconnection_interval
+        self.last_reconnection_time: None | datetime = None
+        self.disconnection_interval = 10.0
+        self.notification_handler = notification_handler
+        self.client: None | Client = None
+
+        SientiaMonitoring.__init__(self, logger, notification_handler, metrics_controller)
+
+        self.metadata = {
+            'model_name': '-',
+            'model_id': '-',
+            'workflow_name': 'opc_repository',
+            'schedule_name': '-',
+        }
+        self._last_write_mono: float | None = None
+        self._connection_lock = threading.Lock()
+        self._session_ready = threading.Event()
+        self._reconnect_thread: threading.Thread | None = None
+        self._allow_reconnect = True
+
+    def _opc_debug_tags(self, session_id: str) -> dict[str, str]:
+        """
+        Build Prometheus/log label tags for OPC session-scoped metrics.
+
+        Args:
+            session_id (str): OPC UA session token string.
+
+        Return:
+            dict[str, str]: Labels pod_id, server_name, runtime, opc_server_id, session_id.
+        """
+        return {
+            'pod_id': str(getattr(self, 'pod_id', 'unknown')),
+            'server_name': self.server_name,
+            'runtime': str(getattr(self, 'runtime', 'unknown')),
+            'opc_server_id': self.id,
+            'session_id': session_id,
+        }
+
+    def _is_session_open(self) -> bool:
+        """
+        Return whether the asyncua client has an open transport session.
+
+        Return:
+            bool: True when protocol exists and is not closed.
+        """
+        if self.client is None:
+            return False
+        try:
+            proto = self.client.aio_obj.uaclient.protocol
+            return proto is not None and proto.state != 'closed'
+        except Exception:
+            return False
+
+    def _reconnection_window_elapsed(self) -> bool:
+        """
+        Return whether enough time has passed since the last reconnect attempt.
+
+        Return:
+            bool: True if a new reconnect is allowed.
+        """
+        if self.last_reconnection_time is None:
+            return True
+        return (
+            datetime.now() - self.last_reconnection_time
+        ).total_seconds() > self.reconnection_interval
+
+    def _not_connected_error(self) -> dict[str, Any]:
+        """
+        Build the standard error payload when validate_connection finds no open protocol.
+
+        Return:
+            dict[str, Any]: Notification fields for OPC_CONNECTION_NOT_READY.
+        """
+        return {
+            'notification_id': f'OPC_CONNECTION_NOT_READY_{self.id}',
+            'message': f'OPC server {self.id} is not connected',
+            'block': 'opc_repository',
+            'level': NotificationLevel.WARNING,
+        }
+
+    def set_security(self) -> None:
+        """
+        Configure certificates and timeouts on the sync asyncua client.
+
+        Raises:
+            ValueError: If cert paths or client are missing.
+        """
+        if self.cert_path is None or self.private_key_path is None:
+            raise ValueError(
+                'Certificate and private key paths must be provided for secure connection.'
+            )
+
+        cert = Path(self.cert_path)
+        private_key = Path(self.private_key_path)
+        server_cert = Path(self.server_cert_path) if self.server_cert_path else None
+
+        if self.client is None:
+            raise ValueError('Client must be initialized before setting security')
+
+        self.client.application_uri = self.server_uri
+        self.info('Setting security...', self.metadata)
+        self.client.set_security(
+            security_policies.SecurityPolicyBasic256,
+            str(cert),
+            str(private_key),
+            None,
+            str(server_cert) if server_cert else None,
+        )
+        aio = self.client.aio_obj
+        aio.secure_channel_timeout = OPC_UA_SESSION_AND_CHANNEL_TIMEOUT_MS
+        aio.session_timeout = OPC_UA_SESSION_AND_CHANNEL_TIMEOUT_MS
+
+    def _create_client(self) -> None:
+        """
+        Instantiate the sync Client and apply security when configured.
+
+        Caller must hold _connection_lock. Does not open a UA session.
+
+        Raises:
+            OpcClientAlreadyExistsError: If self.client is already set.
+        """
+        if self.client is not None:
+            raise OpcClientAlreadyExistsError(
+                f'OPC client already exists for server {self.id}; '
+                'call disconnect() before creating a new client'
+            )
+
+        self.client = Client(self.url, timeout=10)
+        aio = self.client.aio_obj
+        if hasattr(aio, 'watchdog_intervall'):
+            aio.watchdog_intervall = 50
+        aio.name = self.pod_id
+        aio.description = self.pod_id
+        pod_uri = self.pod_id.replace('-', ':')
+        self.client.application_uri = pod_uri
+        aio.product_uri = pod_uri
+        if self.cert_path:
+            self.set_security()
+
+    def _open_session(self) -> tuple[bool, dict[str, Any]]:
+        """
+        Open the OPC UA session on the existing client.
+
+        Caller must hold _connection_lock.
+
+        Raises:
+            OpcClientNotInitializedError: If self.client is None.
+            OpcSessionAlreadyConnectedError: If a session is already open.
+
+        Return:
+            tuple[bool, dict[str, Any]]: Success flag and error payload on connect failure.
+        """
+        if self.client is None:
+            raise OpcClientNotInitializedError(
+                f'OPC client is not initialized for server {self.id}; '
+                'call _create_client() before opening a session'
+            )
+        if self._is_session_open():
+            raise OpcSessionAlreadyConnectedError(
+                f'OPC session already connected for server {self.id}; '
+                'call disconnect() before connecting again'
+            )
+
+        tags = {
+            'pod_id': self.pod_id,
+            'server_name': self.server_name,
+        }
+        self.emit_metric_sync(metrics.OPC_CONNECTIONS_TOTAL, tags)
+
+        try:
+            self.client.connect()
+
+            aio = self.client.aio_obj
+            session_id = _opc_authentication_token_str(self.client)
+            revised_session_timeout_ms = int(aio.session_timeout)
+            revised_secure_channel_timeout_ms = int(aio.secure_channel_timeout)
+            self.info(
+                f'OPC new session connected opc_server_id={self.id} session_id={session_id} '
+                f'revised_session_timeout_ms={revised_session_timeout_ms} '
+                f'revised_secure_channel_timeout_ms={revised_secure_channel_timeout_ms}',
+                self.metadata,
+            )
+            self.emit_metric_sync(
+                metrics.OPC_SESSION_CREATED_TOTAL, self._opc_debug_tags(session_id)
+            )
+            self.emit_metric_sync(
+                metric_object=metrics.OPC_SESSION_REVISED_TIMEOUT_MS,
+                method='set',
+                tags=self._opc_debug_tags(session_id),
+                value=revised_session_timeout_ms,
+            )
+            self.emit_metric_sync(
+                metric_object=metrics.OPC_CONNECTION_STATUS,
+                method='set',
+                tags={**tags, 'server_url': self.url},
+                value=1,
+            )
+
+            self._last_write_mono = None
+            self._session_ready.set()
+            return True, {}
+
+        except Exception as e:
+            self._disconnect_locked()
+            trace = traceback.format_exc()
+            self.error(trace, self.metadata)
+            self.emit_metric_sync(metrics.OPC_CONNECTIONS_FAILED, tags)
+            return False, {
+                'notification_id': f'OPC_CONNECTION_ERROR_{self.id}',
+                'message': f'Failed to connect to OPC server: {e}',
+                'block': 'opc_repository',
+                'level': NotificationLevel.ERROR,
+                'attachment_content': trace,
+            }
+
+    def _connect_locked(self) -> tuple[bool, dict[str, Any]]:
+        """
+        Create the client when absent, then open a UA session.
+
+        Caller must hold _connection_lock.
+
+        Raises:
+            OpcSessionAlreadyConnectedError: If a session is already open.
+
+        Return:
+            tuple[bool, dict[str, Any]]: Result from _open_session on connect failure.
+        """
+        if self._is_session_open():
+            raise OpcSessionAlreadyConnectedError(
+                f'OPC session already connected for server {self.id}; '
+                'call disconnect() before connecting again'
+            )
+        if self.client is None:
+            self._create_client()
+        return self._open_session()
+
+    def _disconnection_fallback(self) -> list[dict[str, Any]]:
+        """
+        Try up to five times to disconnect from the OPC UA server.
+        """
+        assert self.client is not None
+        error_stack: list[dict[str, Any]] = []
+        for i in range(5):
+            try:
+                self.info(
+                    f'Disconnecting from OPC UA server, attempt {i + 1} of 5',
+                    self.metadata,
+                )
+                self.client.disconnect()
+                return []
+            except Exception as e:
+                self.error(
+                    f'Failed to disconnect from OPC UA server in attempt {i + 1} of 5: {e}',
+                    self.metadata,
+                )
+                error_stack.append(
+                    {
+                        'attempt': i + 1,
+                        'error': str(e),
+                        'traceback': traceback.format_exc(),
+                    }
+                )
+                time.sleep(self.disconnection_interval * i)
+        return error_stack
+
+    def _disconnect_locked(self) -> None:
+        """
+        Tear down the current session and client.
+
+        Caller must hold _connection_lock.
+        """
+        self._last_write_mono = None
+        self._session_ready.clear()
+
+        if self.client is None:
+            return
+
+        session_id = _opc_authentication_token_str(self.client)
+        self.info(
+            f'OPC disconnecting opc_server_id={self.id} session_id={session_id}',
+            self.metadata,
+        )
+        self.emit_metric_sync(metrics.OPC_SESSION_CLOSED_TOTAL, self._opc_debug_tags(session_id))
+
+        errors = self._disconnection_fallback()
+        if errors:
+            self.send_notification(
+                metadata=self.metadata,
+                notification_id=f'OPC_DISCONNECTION_ERROR_{self.id}',
+                message='Failed to disconnect from OPC server in 5 attempts.',
+                block='opc_repository',
+                level=NotificationLevel.ERROR,
+                attachment_content=json.dumps(errors, indent=4),
+            )
+        else:
+            self.warning(f'Disconnected from OPC server {self.id} successfully', self.metadata)
+
+        self.emit_metric_sync(
+            metric_object=metrics.OPC_CONNECTION_STATUS,
+            method='set',
+            tags={
+                'pod_id': self.pod_id,
+                'server_name': self.server_name,
+                'server_url': self.url,
+            },
+            value=0,
+        )
+        self.client = None
+
+    def _reconnect_locked(self) -> tuple[bool, dict[str, Any]]:
+        """
+        Close the current session and open a new one.
+
+        Caller must hold _connection_lock. Records last_reconnection_time for interval gating.
+
+        Return:
+            tuple[bool, dict[str, Any]]: Result from _connect_locked after teardown.
+        """
+        self.last_reconnection_time = datetime.now()
+        self._disconnect_locked()
+        return self._connect_locked()
+
+    def connect(self) -> tuple[bool, dict[str, Any]]:
+        """
+        Open an OPC UA session under the connection lock (worker initialization).
+        """
+        with self._connection_lock:
+            self.info(
+                f'Starting connection to OPC server {self.id}:{self.server_name}...',
+                self.metadata,
+            )
+            return self._connect_locked()
+
+    def disconnect(self) -> None:
+        """
+        Gracefully disconnect from the OPC server under the connection lock.
+
+        Disables background reconnect so late writes during worker shutdown do not
+        respawn sessions.
+        """
+        with self._connection_lock:
+            self._allow_reconnect = False
+            self._disconnect_locked()
+
+    def validate_connection(self) -> tuple[bool, dict[str, Any]]:
+        """
+        Read-only check that the asyncua protocol is open.
+
+        Caller must ensure _session_ready before writing. Does not connect or reconnect.
+
+        Return:
+            tuple[bool, dict[str, Any]]: (True, {}) when open, otherwise (False, error).
+        """
+        if self._is_session_open():
+            return True, {}
+        self.error(f'OPC server {self.id} is not connected', self.metadata)
+        return False, self._not_connected_error()
+
+    def _reconnect_thread_in_progress(self) -> bool:
+        """
+        Return whether a background reconnect thread is currently running.
+
+        Return:
+            bool: True when a reconnect thread exists and is alive.
+        """
+        return self._reconnect_thread is not None and self._reconnect_thread.is_alive()
+
+    def _start_reconnect(self, reason: str, session_id: str) -> None:
+        """
+        Schedule a background reconnect when allowed by interval and thread state.
+
+        Clears _session_ready before starting the thread. No-op when _allow_reconnect is
+        False, the reconnection window has not elapsed, or a reconnect is already running.
+
+        Args:
+            reason (str): Trigger for reconnect (OPC status name or synthetic reason).
+            session_id (str): Session token before failure.
+        """
+        if not self._allow_reconnect:
+            return
+        if not self._reconnection_window_elapsed():
+            self.warning(
+                f'OPC reconnect skipped reason=reconnection_window opc_server_id={self.id} '
+                f'reconnect_reason={reason}',
+                self.metadata,
+            )
+            return
+        if self._reconnect_thread_in_progress():
+            self.warning(
+                f'OPC reconnect skipped reason=in_progress opc_server_id={self.id} '
+                f'reconnect_reason={reason}',
+                self.metadata,
+            )
+            return
+
+        self._session_ready.clear()
+        self.info(
+            f'OPC reconnect scheduled reconnect_reason={reason} opc_server_id={self.id} '
+            f'old_session_id={session_id}',
+            self.metadata,
+        )
+        self._reconnect_thread = threading.Thread(
+            target=self._run_reconnect,
+            args=(reason, session_id),
+            daemon=True,
+        )
+        self._reconnect_thread.start()
+
+    def _run_reconnect(self, reason: str, session_id: str) -> None:
+        """
+        Tear down and re-establish the OPC UA session under the connection lock.
+
+        Args:
+            reason (str): Trigger for reconnect (OPC status or ProtocolClosed).
+            session_id (str): Previous session token string.
+        """
+        try:
+            with self._connection_lock:
+                self.info(
+                    f'OPC reconnect started reconnect_reason={reason} opc_server_id={self.id} '
+                    f'old_session_id={session_id}',
+                    self.metadata,
+                )
+                success, error = self._reconnect_locked()
+            if not success:
+                self.error(
+                    f'OPC reconnect failed reconnect_reason={reason} opc_server_id={self.id}',
+                    self.metadata,
+                )
+                if error:
+                    self.error(error.get('message', ''), self.metadata)
+        except Exception:
+            self.error(
+                f'OPC reconnect task failed opc_server_id={self.id} reconnect_reason={reason}',
+                self.metadata,
+            )
+            self.error(traceback.format_exc(), self.metadata)
+
+    def _log_write_inter_arrival(self, session_id: str, node: str) -> None:
+        """
+        Log elapsed wall time since the previous successful OPC write on this repository.
+
+        Args:
+            session_id (str): Current OPC UA session token string.
+            node (str): Node id written in this operation.
+        """
+        now = time.monotonic()
+        if self._last_write_mono is not None:
+            delta_s = now - self._last_write_mono
+            self.info(
+                f'OPC write inter-arrival_s={delta_s:.6f} opc_server_id={self.id} '
+                f'session_id={session_id} node={node}',
+                self.metadata,
+            )
+            if self.client is not None:
+                session_timeout_ms = float(self.client.aio_obj.session_timeout)
+                if session_timeout_ms > 0 and delta_s > (session_timeout_ms / 1000.0):
+                    self.emit_metric_sync(
+                        metrics.OPC_WRITE_INTER_ARRIVAL_OVER_SESSION_TIMEOUT_TOTAL,
+                        self._opc_debug_tags(session_id),
+                    )
+        self._last_write_mono = now
+
+    def _emit_opc_write_metric(
+        self, session_id: str, result: str, metadata: dict[str, Any] | None
+    ) -> None:
+        self.emit_metric_sync(
+            metrics.OPC_WRITE_ATTEMPTS_TOTAL,
+            {
+                **self._opc_debug_tags(session_id),
+                **_model_labels_from_write_metadata(metadata),
+                'result': result,
+            },
+        )
+
+    def _write_failure_payload(
+        self,
+        notification_id: str,
+        message: str,
+        level: NotificationLevel = NotificationLevel.ERROR,
+        attachment_content: str | None = None,
+        opc_error_kind: str | None = None,
+        opc_status: str | None = None,
+    ) -> dict[str, Any]:
+        payload: dict[str, Any] = {
+            'notification_id': notification_id,
+            'message': message,
+            'block': 'opc_repository',
+            'level': level,
+        }
+        if attachment_content is not None:
+            payload['attachment_content'] = attachment_content
+        if opc_error_kind is not None:
+            payload['opc_error_kind'] = opc_error_kind
+        if opc_status is not None:
+            payload['opc_status'] = opc_status
+        return payload
+
+    def _handle_tier1_bad(
+        self,
+        exc: BaseException,
+        session_id: str,
+        node: str,
+        metadata: dict[str, Any],
+        phase: str,
+    ) -> tuple[bool, dict[str, Any]]:
+        """
+        Record metrics/logs and schedule reconnect after a Tier-1 Bad* error.
+
+        Args:
+            exc (BaseException): Tier-1 OPC UA error.
+            session_id (str): Session token at failure time.
+            node (str): Node id being written.
+            metadata (dict[str, Any]): Write context.
+            phase (str): get_node or write_value.
+
+        Return:
+            tuple[bool, dict[str, Any]]: Always (False, error payload).
+        """
+        opc_status = _opc_status_from_exception(exc)
+        trace = traceback.format_exc()
+        self.error(trace, metadata)
+        self._emit_opc_write_metric(session_id, opc_status, metadata)
+        self.error(
+            f'OPC write failed opc_status={opc_status} opc_server_id={self.id} '
+            f'session_id={session_id} model_id={metadata.get("model_id", "unknown")} '
+            f'model_name={metadata.get("model_name", "unknown")} node={node} phase={phase}',
+            metadata,
+        )
+        self._start_reconnect(opc_status, session_id)
+        return False, self._write_failure_payload(
+            notification_id=f'OPC_WRITE_DATA_ERROR_{self.id}',
+            message=f'Failed to {phase} on OPC server: {exc} | metadata: {metadata}',
+            attachment_content=trace,
+            opc_error_kind='session_bad',
+            opc_status=opc_status,
+        )
+
+    @staticmethod
+    def _write_node_value(
+        node_obj: Any,
+        ua_data: ua.DataValue,
+        data: Any,
+        variant_type: ua.VariantType,
+    ) -> None:
+        """
+        Write a DataValue to a node, falling back to set_value when write_value is unavailable.
+
+        Args:
+            node_obj: Sync or async node wrapper from asyncua.
+            ua_data (ua.DataValue): Encoded value for write_value.
+            data: Scalar converted value for set_value fallback.
+            variant_type (ua.VariantType): OPC UA type for set_value fallback.
+        """
+        if hasattr(node_obj, 'write_value'):
+            try:
+                node_obj.write_value(ua_data)
+                return
+            except (AttributeError, TypeError):
+                pass
+        node_obj.set_value(data, variant_type)
+
+    def _write_reconnect_in_progress(self, metadata: dict[str, Any]) -> tuple[bool, dict[str, Any]]:
+        """
+        Fail a write because a background reconnect thread is already running.
+
+        Args:
+            metadata (dict[str, Any]): Write context passed through to the activity.
+
+        Return:
+            tuple[bool, dict[str, Any]]: (False, error info with opc_error_kind reconnect_in_progress).
+        """
+        self._emit_opc_write_metric('unknown', 'ReconnectInProgress', metadata)
+        self.warning(
+            f'OPC write rejected reconnect_in_progress opc_server_id={self.id} '
+            f'model_id={metadata.get("model_id", "unknown")} '
+            f'model_name={metadata.get("model_name", "unknown")}',
+            metadata,
+        )
+        return False, {
+            'notification_id': f'OPC_WRITE_RECONNECT_IN_PROGRESS_{self.id}',
+            'message': f'OPC write skipped: reconnect in progress | metadata: {metadata}',
+            'block': 'opc_repository',
+            'level': NotificationLevel.WARNING,
+            'opc_error_kind': 'reconnect_in_progress',
+        }
+
+    def _write_connection_lost(
+        self, metadata: dict[str, Any], opc_status: str
+    ) -> tuple[bool, dict[str, Any]]:
+        """
+        Fail a write after scheduling reconnect for a closed or stale session.
+
+        Args:
+            metadata (dict[str, Any]): Write context passed through to the activity.
+            opc_status (str): Synthetic reason (ProtocolClosed, SessionNotReady).
+
+        Return:
+            tuple[bool, dict[str, Any]]: (False, error info with opc_error_kind connection_lost).
+        """
+        self._emit_opc_write_metric('unknown', opc_status, metadata)
+        return False, self._write_failure_payload(
+            notification_id=f'OPC_WRITE_CONNECTION_LOST_{self.id}',
+            message=f'OPC write skipped: connection lost ({opc_status}) | metadata: {metadata}',
+            level=NotificationLevel.WARNING,
+            opc_error_kind='connection_lost',
+            opc_status=opc_status,
+        )
+
+    def write_data(
+        self, node: str, value: Any, data_type: str, metadata: dict[str, Any]
+    ) -> tuple[bool, dict[str, Any]]:
+        """
+        Write data to OPC server with a single attempt and background reconnect scheduling.
+
+        Reconnect is scheduled on Tier-1 Bad*, closed protocol, or stale session readiness.
+        There is no retry within the same call.
+
+        Args:
+            node (str): OPC UA node id to write.
+            value (Any): Value to convert and send.
+            data_type (str): Logical type key (float, int, bool, str, double).
+            metadata (dict[str, Any]): Activity context (model_id, model_name, etc.).
+
+        Return:
+            tuple[bool, dict[str, Any]]: (True, {response_time}) on success, or
+                (False, structured error info) on failure.
+        """
+        if self._reconnect_thread_in_progress():
+            return self._write_reconnect_in_progress(metadata)
+
+        if not self._session_ready.is_set():
+            session_id = _opc_authentication_token_str(self.client)
+            self._start_reconnect('SessionNotReady', session_id)
+            if self._reconnect_thread_in_progress():
+                return self._write_reconnect_in_progress(metadata)
+            return self._write_connection_lost(metadata, 'SessionNotReady')
+
+        is_connected, _error = self.validate_connection()
+        if not is_connected:
+            session_id = _opc_authentication_token_str(self.client)
+            self._start_reconnect('ProtocolClosed', session_id)
+            return self._write_connection_lost(metadata, 'ProtocolClosed')
+
+        start_time = time.time()
+        session_id = _opc_authentication_token_str(self.client)
+
+        try:
+            assert self.client is not None
+            node_obj = self.client.get_node(node)
+        except Exception as e:
+            if is_reconnectable_opcua_bad(e):
+                return self._handle_tier1_bad(e, session_id, node, metadata, 'get_node')
+            trace = traceback.format_exc()
+            self.error(trace, metadata)
+            self._emit_opc_write_metric(session_id, f'GetNodeError:{type(e).__name__}', metadata)
+            return False, self._write_failure_payload(
+                notification_id=f'OPC_WRITE_GET_NODE_ERROR_{self.id}',
+                message=f'Failed to get node from OPC server: {e} | metadata: {metadata}',
+                attachment_content=trace,
+            )
+
+        if data_type not in data_type_map:
+            self._emit_opc_write_metric(session_id, 'UnsupportedDataType', metadata)
+            return False, self._write_failure_payload(
+                notification_id=f'OPC_WRITE_DATA_TYPE_ERROR_{self.id}',
+                message=f'Unsupported data type: {data_type} | metadata: {metadata}',
+            )
+
+        data = data_type_map[data_type]['converter'](value)
+        self.info(f'Writing {data} - {type(data)} to {node}', metadata)
+        variant_type = data_type_map[data_type]['opc_type']
+        ua_data = ua.DataValue(
+            ua.Variant(data, variant_type),
+        )
+
+        try:
+            self._write_node_value(node_obj, ua_data, data, variant_type)
+            end_time = time.time()
+            response_time = end_time - start_time
+        except Exception as e:
+            if is_reconnectable_opcua_bad(e):
+                return self._handle_tier1_bad(e, session_id, node, metadata, 'write_value')
+            trace = traceback.format_exc()
+            self.error(trace, metadata)
+            self._emit_opc_write_metric(session_id, type(e).__name__, metadata)
+            return False, self._write_failure_payload(
+                notification_id=f'OPC_WRITE_DATA_ERROR_{self.id}',
+                message=f'Failed to write data to OPC server: {e} | metadata: {metadata}',
+                attachment_content=trace,
+            )
+
+        self._emit_opc_write_metric(session_id, 'OK', metadata)
+        self._log_write_inter_arrival(session_id, node)
+
+        return True, {
+            'response_time': response_time,
+        }

laborious/worker/worker.py

@@ -0,0 +1,313 @@
+"""
+Laborious Worker Module
+
+Entry process that connects to Temporal, registers Laborious activities, and runs four workers in
+parallel. Each worker shares the same ``Activities`` instance (single Postgres pool, single MLflow
+repository, single PluginStore handle) but polls a different task queue.
+
+Task queues (see ``sientia_do.temporal.worker.prepare_worker``):
+- ``predictions_batch-{runtime}-queue`` + sub-workflows on the same queue (ML-heavy path).
+- ``minimal_retrain-{runtime}-queue`` (retrain + promote + export).
+- ``drift-queue`` and ``simple_metrics-queue`` without a runtime suffix so existing schedulers
+  keep stable queue names.
+
+Bootstrap order:
+1. Prometheus app metrics and Mongo-backed notification handler.
+2. ``RUNTIME`` validation and ``PluginStore.install_runtime`` so ``SientiaModel`` code is importable.
+3. ``Activities`` construction (builds ``SientiaMLflowRepository`` internally from env).
+4. OPC client initialization inside activities.
+5. Temporal ``Runtime`` with SDK Prometheus bind, client connect, then ``prepare_worker`` per workflow.
+
+Shutdown closes workers, notifications, activities (pools + OPC), and clears ``app_up``.
+
+Environment Variables:
+- RUNTIME: Required non-empty string passed to ``install_runtime``.
+- STORE_* / PYPI_*: Plugin store and private index (see ``build_plugin_store_config``).
+- TEMPORAL_HOST, TEMPORAL_NAMESPACE: Cluster connection.
+- POD_ID, HTTP_METRICS_PORT, HTTP_SDK_METRICS_PORT: Observability.
+- PROJECT_NAME, MONGODB_*: Notifications (via ``build_mongodb_config`` in handler).
+- POSTGRES_*, MINIO_*, OPC_*, PI_WEB_API_*, MLFLOW_*: Passed through ``Activities`` helpers.
+"""
+
+from temporalio import client, workflow
+from temporalio.runtime import PrometheusConfig, Runtime, TelemetryConfig
+
+with workflow.unsafe.imports_passed_through():
+    import asyncio
+    import os
+    import sys
+
+    from prometheus_client import start_http_server
+    from sientia_do.notifications.handlers import CoreNotificationHandler as NotificationHandler
+    from sientia_do.observability.logger import get_logger
+    from sientia_do.observability.metrics_controller import MetricsController
+    from sientia_do.temporal.worker.prepare_worker import prepare_worker
+    from sientia_do.utils.connectors_config import (
+        build_api_config,
+        build_mongodb_config,
+        build_postgres_config,
+    )
+    from sientia_model.model_repository.plugin_store import PluginStore
+
+    from laborious import metrics
+    from laborious.activities.activities import Activities
+    from laborious.utils.connectors_config import (
+        build_minio_config,
+        build_opc_config,
+        build_plugin_store_config,
+    )
+    from laborious.workflows.drift import Drift
+    from laborious.workflows.minimal_retrain import MinimalRetrain
+    from laborious.workflows.predictions_batch import PredictionsBatch
+    from laborious.workflows.simple_metrics import SimpleMetrics
+    from laborious.workflows.sub_workflows.format_and_export_prediction import (
+        FormatAndExportPrediction,
+    )
+    from laborious.workflows.sub_workflows.prediction_process import PredictionProcess
+
+POD_ID = os.getenv('HOSTNAME')
+SDK_METRICS_PORT = int(os.getenv('HTTP_SDK_METRICS_PORT', '9091'))
+
+
+async def main():
+    """
+    Run the full worker lifecycle: metrics, notifications, runtime install, workers, gather.
+
+    Exits the process with code 0 on normal completion of all worker tasks, or 1 after logging
+    if any worker raises. ``finally`` always shuts down notifications and activities and sets
+    ``app_up`` to 0 before ``sys.exit``.
+
+    Raises:
+        Exception: Propagated from ``asyncio.gather`` only before ``finally`` handling; typically
+            workers run until cancelled.
+
+    Return:
+        None (process terminates via ``sys.exit`` from the ``finally`` block).
+    """
+    host = os.getenv('TEMPORAL_HOST', 'localhost:7233')
+    logger = get_logger(__name__)
+
+    metadata = {
+        'pod_id': POD_ID,
+        'model_name': '-',
+        'model_id': '-',
+        'workflow_name': '-',
+        'schedule_name': '-',
+    }
+
+    logger.custom_info(f'Starting Worker with POD_ID: {POD_ID}', metadata)
+
+    logger.custom_info('Starting prometheus client...', metadata)
+    start_prometheus_server()
+
+    logger.custom_info('Starting Notification Handler...', metadata)
+
+    mongo_config = build_mongodb_config()
+    notification_handler = NotificationHandler(
+        connection_string=mongo_config['connection_string'],
+        database=mongo_config['database_name'],
+        logger=logger,
+        project_name=os.getenv('PROJECT_NAME', 'laborious'),
+    )
+
+    metrics_controller = MetricsController(logger=logger)
+
+    runtime = os.getenv('RUNTIME', '').strip()
+    if not runtime:
+        logger.custom_critical(
+            'RUNTIME environment variable is required and must be non-empty',
+            metadata,
+        )
+        metrics.APP_UP.labels(pod_id=POD_ID).set(0)
+        sys.exit(1)
+
+    metadata_runtime = {**metadata, 'runtime': runtime}
+    logger.custom_info(f'Installing PluginStore runtime: {runtime}', metadata_runtime)
+
+    ps_cfg = build_plugin_store_config()
+    plugin_store = PluginStore(
+        base_url=ps_cfg['base_url'],
+        owner=ps_cfg['owner'],
+        repo=ps_cfg['repo'],
+        username=ps_cfg['username'],
+        password=ps_cfg['password'],
+        branch=ps_cfg['branch'],
+        cache_ttl_seconds=ps_cfg['cache_ttl_seconds'],
+        pypi_index_url=ps_cfg['pypi_index_url'],
+        pypi_username=ps_cfg['pypi_username'],
+        pypi_password=ps_cfg['pypi_password'],
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+    )
+
+    if runtime == 'legacy':
+        to_install_runtime = 'single'
+    else:
+        to_install_runtime = runtime
+
+    try:
+        await plugin_store.install_runtime(
+            runtime_name=to_install_runtime, metadata=metadata_runtime
+        )
+    except Exception as exc:
+        logger.custom_critical(
+            f'Failed to install runtime {to_install_runtime}: {exc}', metadata_runtime
+        )
+        metrics.APP_UP.labels(pod_id=POD_ID).set(0)
+        sys.exit(1)
+
+    logger.custom_info('Starting Activities...', metadata)
+
+    activities = Activities(
+        postgres_config=build_postgres_config(),
+        plugin_store=plugin_store,
+        minio_config=build_minio_config(),
+        opc_config=build_opc_config(),
+        pi_web_api_config=build_api_config(),
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+    )
+
+    logger.custom_info('Initializing OPC...', metadata)
+    activities.init_opc()
+
+    logger.custom_info(f'Starting SDK Metrics Server on port {SDK_METRICS_PORT}...', metadata)
+
+    new_runtime = Runtime(
+        telemetry=TelemetryConfig(
+            metrics=PrometheusConfig(bind_address=f'0.0.0.0:{SDK_METRICS_PORT}')
+        )
+    )
+
+    logger.custom_info(f'Starting Temporal Client at {host}...', metadata)
+
+    temporal_client = await client.Client.connect(
+        target_host=host,
+        namespace=os.getenv('TEMPORAL_NAMESPACE', 'laborious'),
+        runtime=new_runtime,
+    )
+
+    logger.custom_info('Starting Workers...', metadata)
+
+    workers = [
+        prepare_worker(
+            temporal_client=temporal_client,
+            main_workflow=MinimalRetrain,
+            other_workflows=[],
+            activities=[
+                activities.load_query_with_minio_offload,
+                activities.retrain_model,
+                activities.update_production_model,
+                activities.format_retrain_report,
+                activities.export_data_to_postgres,
+            ],
+            logger=logger,
+            runtime=runtime,
+        ),
+        prepare_worker(
+            temporal_client=temporal_client,
+            main_workflow=SimpleMetrics,
+            other_workflows=[],
+            activities=[
+                activities.load_custom_query,
+                activities.calculate_simple_metrics,
+                activities.export_data_to_postgres,
+            ],
+            logger=logger,
+            runtime='core',
+        ),
+        prepare_worker(
+            temporal_client=temporal_client,
+            main_workflow=Drift,
+            other_workflows=[],
+            activities=[
+                activities.load_custom_query,
+                activities.get_reference_data,
+                activities.calculate_drift,
+                activities.export_data_to_postgres,
+            ],
+            logger=logger,
+            runtime='core',
+        ),
+        prepare_worker(
+            temporal_client=temporal_client,
+            main_workflow=PredictionsBatch,
+            other_workflows=[PredictionProcess, FormatAndExportPrediction],
+            activities=[
+                # MLFlow
+                activities.request_predict,
+                activities.request_transform,
+                # Gates
+                activities.input_gate,
+                activities.mlflow_response_gate,
+                activities.mlflow_content_gate,
+                activities.format_transformed_data,
+                activities.format_prediction,
+                activities.format_default_prediction,
+                # OPC
+                activities.write_opc_data,
+                # Postgres / MinIO offload
+                activities.load_query_with_minio_offload,
+                activities.cleanup_minio_objects_expired,
+                activities.repeat_last_prediction,
+                activities.export_data_to_postgres,
+                activities.export_payload_to_postgres,
+                activities.write_metrics,
+                # Pi Web API
+                activities.write_pi_web_api_data,
+            ],
+            logger=logger,
+            runtime=runtime,
+        ),
+    ]
+
+    handlers = []
+    for w in workers:
+        handlers.append(w.run())
+
+    logger.custom_info('Workers started successfully', metadata)
+
+    exit_code = 0
+    try:
+        await asyncio.gather(*handlers)
+    except BaseException as e:  # NOSONAR
+        logger.custom_error(f'An unhandled exception occurred: {e}', metadata)
+        exit_code = 1
+    finally:
+        notification_handler.shutdown()
+        activities.shutdown()
+        metrics.APP_UP.labels(pod_id=POD_ID).set(0)  # Mark app as DOWN
+        sys.exit(exit_code)
+
+
+def start_prometheus_server():
+    """
+    Starts the Prometheus metrics server for monitoring and observability.
+
+    This function initializes the Prometheus HTTP server on the configured port
+    and sets the application health metric to indicate the service is running.
+
+    The server exposes metrics that can be scraped by Prometheus for monitoring
+    the health and performance of the Laborious worker.
+
+    Environment Variables:
+        HTTP_METRICS_PORT: Port for the metrics server (default: 9090)
+        POD_ID: Pod identifier for metrics labeling
+
+    Raises:
+        SystemExit: If the metrics server fails to start
+    """
+    try:
+        port = int(os.getenv('HTTP_METRICS_PORT', 9090))
+        start_http_server(port)
+        print(f'Prometheus server started on port {port}.')
+        metrics.APP_UP.labels(pod_id=POD_ID).set(1)  # Mark app as UP
+    except Exception as e:
+        print(f'Failed to start Prometheus server: {e}')
+        os._exit(1)
+
+
+if __name__ == '__main__':
+    asyncio.run(main())

laborious/workflows/drift.py

@@ -0,0 +1,107 @@
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from datetime import timedelta
+    from typing import Any
+
+    from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+    from sientia_do.temporal.policies import retry_policy
+
+    from laborious.activities.activities import Activities
+
+
+@workflow.defn(name='drift')
+class Drift:
+    @workflow.run
+    async def run(self, input_data: dict[str, Any]):
+        """
+        Execute the drift workflow.
+
+        This method orchestrates the complete drift process by:
+        1. Loading data using the provided custom SQL query
+        2. Preparing prediction configuration and filters
+        3. Delegating to the PredictionProcess workflow for ML operations
+        """
+
+        metadata = {
+            'metadata': {
+                'schedule_name': input_data['schedule_name'],
+                'model_name': input_data['model_name'],
+                'model_id': input_data['model_id'],
+                'workflow_name': 'drift',
+            }
+        }
+
+        print(f'Input data: {input_data}', metadata)
+
+        model_config = input_data['model_config']
+        target_name = model_config['target']
+
+        gathering_query = f"""
+            SELECT *
+            FROM "{input_data['schema']}"."{input_data['source_table_name']}"
+            WHERE
+                model_id = '{input_data['model_id']}' AND
+                timestamp > NOW() - INTERVAL '{input_data['interval']} minutes'
+            ORDER BY timestamp ASC
+        """  # nosec B608 - values come from internal Temporal workflow config, not user input
+
+        target_data_handler = workflow.start_activity_method(
+            Activities.load_custom_query,
+            {
+                **metadata,
+                'query': gathering_query,
+                'datetime_columns': ['timestamp', 'created_at'],
+                'orient': 'records',
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )
+
+        reference_data_handler = workflow.start_activity_method(
+            Activities.get_reference_data,
+            {**metadata, 'model_name': input_data['model_name']},
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )
+
+        target_data = await target_data_handler
+        reference_data = await reference_data_handler
+
+        if not target_data:
+            return
+
+        drift_data = await workflow.execute_local_activity_method(
+            Activities.calculate_drift,
+            {
+                **metadata,
+                'target_data': target_data,
+                'reference_data': reference_data,
+                'model_name': input_data['model_name'],
+                'model_id': input_data['model_id'],
+                'target_name': target_name,
+                'drift_metrics': input_data.get(
+                    'drift_metrics', ['kolmogorov_smirnov', 'jensen_shannon', 'wasserstein']
+                ),
+                'chunk_period': input_data.get('chunk_period', 'min'),
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )
+
+        if drift_data:
+            await workflow.execute_activity_method(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'data': drift_data,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['target_table_name'],
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                },
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(seconds=300),
+            )

laborious/workflows/minimal_retrain.py

@@ -0,0 +1,137 @@
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from datetime import timedelta
+    from typing import Any
+
+    from sientia_do.temporal.policies import retry_policy
+
+    from laborious.activities.activities import Activities
+    from laborious.utils.models.minio_dataframe_payload import MinioDataFramePayload
+
+
+@workflow.defn(name='minimal_retrain')
+class MinimalRetrain:
+    """
+    Automated model retraining workflow for the Laborious system.
+
+    This workflow implements a complete model retraining pipeline that loads
+    training data, executes model retraining, updates production models,
+    and maintains comprehensive audit trails. It's designed for automated
+    model lifecycle management with minimal manual intervention.
+
+    The workflow provides a robust retraining process with:
+    - Automated data loading from configured data sources
+    - MLFlow model retraining with quality validation
+    - Production model updates with version control
+    - Comprehensive reporting and audit trail maintenance
+    - Error handling and notification integration
+    """
+
+    @workflow.run
+    async def run(self, input_data: dict[str, Any]):
+        """
+        Execute the automated model retraining workflow.
+
+        This method orchestrates the complete model retraining process by:
+        1. Loading training data using the provided custom SQL query
+        2. Executing MLFlow model retraining with the loaded data
+        3. Updating production models with newly trained versions
+        4. Persisting comprehensive retraining reports to database
+
+        The method implements comprehensive error handling and ensures all
+        required parameters are properly configured before proceeding.
+
+        Args:
+            input_data: Complete configuration for the retraining workflow
+                Required keys:
+                    - schedule_name (str): Schedule identifier for the retraining
+                    - model_name (str): Name of the ML model to retrain
+                    - model_id (int): Unique identifier for the model version
+                    - query (str): SQL query for training data loading
+                    - schema (str, optional): Database schema for report storage
+                    - table_name (str, optional): Target table for retraining reports
+                    - datetime_columns (list[str], optional): Columns to treat as datetime
+
+        Returns:
+            None: The workflow completes successfully when all steps finish
+
+        Raises:
+            Exception: If any required parameters are missing or if the workflow fails
+                during data loading, retraining, or model update operations
+        """
+
+        metadata = {
+            'metadata': {
+                'schedule_name': input_data['schedule_name'],
+                'model_name': input_data['model_name'],
+                'model_id': input_data['model_id'],
+                'workflow_name': 'minimal_retrain',
+            }
+        }
+
+        model_name = input_data['model_name']
+        model_config = input_data.get('model_config', {})
+
+        storage_result = await workflow.execute_activity_method(
+            Activities.load_query_with_minio_offload,
+            {
+                **metadata,
+                'query': input_data['query'],
+                'datetime_columns': input_data.get('datetime_columns', []),
+                'model_name': model_name,
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=600),
+        )
+
+        storage_payload = MinioDataFramePayload.from_dict(storage_result)
+        if not storage_payload.has_data():
+            raise ValueError('No data returned from query')
+
+        experiment_response = await workflow.execute_activity_method(
+            Activities.retrain_model,
+            {
+                **metadata,
+                'data': storage_result,
+                'model_name': model_name,
+                'model_config': model_config,
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(hours=1),
+        )
+
+        if experiment_response['success']:
+            update_report = await workflow.execute_activity_method(
+                Activities.update_production_model,
+                {**metadata, 'model_name': model_name, **experiment_response},
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(seconds=60),
+            )
+        else:
+            update_report = {}
+
+        report = await workflow.execute_local_activity_method(
+            Activities.format_retrain_report,
+            {
+                **metadata,
+                'experiment_response': experiment_response,
+                'model_name': model_name,
+                'model_id': input_data['model_id'],
+                'update_report': update_report,
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=60),
+        )
+
+        await workflow.execute_activity_method(
+            Activities.export_data_to_postgres,
+            {
+                **metadata,
+                'data': report,
+                'schema': input_data['schema'],
+                'table_name': input_data['table_name'],
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=600),
+        )

laborious/workflows/predictions_batch.py

@@ -0,0 +1,127 @@
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from datetime import timedelta
+    from typing import Any
+
+    from sientia_do.temporal.policies import retry_policy
+
+    from laborious.activities.activities import Activities
+
+
+@workflow.defn(name='predictions_batch')
+class PredictionsBatch:
+    """
+    Main batch prediction workflow for the Laborious system.
+
+    This workflow orchestrates the complete batch prediction process, handling
+    data loading, configuration management, and workflow delegation. It serves
+    as the primary entry point for batch prediction operations and ensures
+    proper data preparation before ML model inference.
+
+    The workflow implements a robust data processing pipeline with:
+    - Custom SQL query execution for data loading
+    - Comprehensive configuration management
+    - Data quality filter application
+    - MLFlow model integration
+    - Workflow delegation to specialized sub-workflows
+
+    Workflow Execution:
+        1. Data Loading: Executes custom SQL query to load prediction data
+        2. Configuration Preparation: Sets up prediction parameters and filters
+        3. Workflow Delegation: Spawns PredictionProcess child workflow
+        4. Error Handling: Implements comprehensive error handling and retry policies
+    """
+
+    @workflow.run
+    async def run(self, input_data: dict[str, Any]):
+        """
+        Execute the batch prediction workflow.
+
+        This method orchestrates the complete batch prediction process by:
+        1. Loading data using the provided custom SQL query
+        2. Preparing prediction configuration and filters
+        3. Delegating to the PredictionProcess workflow for ML operations
+
+        The method implements comprehensive error handling and ensures all
+        required parameters are properly configured before proceeding.
+
+        Args:
+            input_data: Complete configuration for the batch prediction
+                Required keys:
+                    - schedule_name (str): Schedule identifier for the prediction
+                    - model_name (str): Name of the ML model to use
+                    - model_id (int): Unique identifier for the model
+                    - query (str): SQL query for data loading
+                    - schema (dict, optional): Data schema definition
+                    - table_name (str, optional): Target table for predictions
+                    - input_filters (dict, optional): Data quality filters
+                    - mlflow_transform_filters (dict, optional): MLFlow transform filters
+                    - mlflow_predict_filters (dict, optional): MLFlow prediction filters
+                    - model_retention (int, optional): Model retention period in minutes
+                    - path_priority (list[str]): Decision path priority configuration
+                    - opc_output_config (dict, optional): OPC server export configuration
+                    - pi_web_api_output_config (dict, optional): PI Web API export configuration
+                    - datetime_columns (list[str], optional): Columns to treat as datetime
+                    - save_transform (bool, optional): Whether to save transformed data (default: True)
+                    - prediction_store_policy (str, optional): Data retention policy (default: 'lts:1')
+
+        Returns:
+            None: The workflow completes successfully when the child workflow finishes
+
+        Raises:
+            Exception: If any required parameters are missing or if the workflow fails
+                during data loading or workflow delegation
+        """
+
+        metadata = {
+            'metadata': {
+                'schedule_name': input_data['schedule_name'],
+                'model_name': input_data['model_name'],
+                'model_id': input_data['model_id'],
+                'workflow_name': 'predictions_batch',
+            }
+        }
+
+        # Load data using custom query with optional MinIO offload for large frames
+        data = await workflow.execute_activity_method(
+            Activities.load_query_with_minio_offload,
+            {
+                **metadata,
+                'query': input_data['query'],
+                'datetime_columns': input_data.get('datetime_columns', []),
+                'model_name': input_data['model_name'],
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )
+
+        # Prepare input for prediction_process workflow
+        prediction_input = {
+            'metadata': metadata,
+            'data': data,
+            'schema': input_data['schema'],
+            'table_name': input_data['table_name'],
+            'transform_table_name': input_data['transform_table_name'],
+            'model_id': input_data['model_id'],
+            'model_name': input_data['model_name'],
+            'input_filters': input_data.get(
+                'input_filters', {'EMPTY_DATA': {'POLICY': 'STOP', 'CONFIG': {}}}
+            ),
+            'mlflow_transform_filters': input_data.get(
+                'mlflow_transform_filters', {'API_ERROR': {'POLICY': 'STOP', 'CONFIG': {}}}
+            ),
+            'mlflow_predict_filters': input_data.get(
+                'mlflow_predict_filters', {'API_ERROR': {'POLICY': 'STOP', 'CONFIG': {}}}
+            ),
+            'model_config': input_data.get('model_config', {}),
+            'path_priority': input_data.get('path_priority', ['STOP', 'CONTINUE', 'REPEAT']),
+            'opc_output_config': input_data.get('opc_output_config', {}),
+            'on_conflict': input_data.get('on_conflict', 'error'),
+            'pi_web_api_output_config': input_data.get('pi_web_api_output_config', {}),
+            'prediction_store_policy': input_data.get('prediction_store_policy', 'lts:1'),
+            'save_transform': input_data.get('save_transform', True),
+        }
+
+        # Execute prediction process workflow
+        await workflow.execute_child_workflow('subworkflow.prediction_process', prediction_input)

laborious/workflows/simple_metrics.py

@@ -0,0 +1,95 @@
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from datetime import timedelta
+    from typing import Any
+
+    from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+    from sientia_do.temporal.policies import retry_policy
+
+    from laborious.activities.activities import Activities
+
+
+@workflow.defn(name='simple_metrics')
+class SimpleMetrics:
+    @workflow.run
+    async def run(self, input_data: dict[str, Any]):
+        """
+        Execute the simple metrics workflow.
+        """
+        metadata = {
+            'metadata': {
+                'model_id': input_data['model_id'],
+                'model_name': input_data['model_name'],
+                'workflow_name': 'simple_metrics',
+                'schedule_name': input_data['schedule_name'],
+            }
+        }
+
+        model_id = input_data['model_id']
+        interval_minutes = input_data['interval_minutes']
+
+        model_config = input_data['model_config']
+        target_name = model_config['target']
+
+        query = f"""
+            select p."timestamp", p.prediction, ld.value as "target"
+            from "{input_data['schema']}"."{input_data['predictions_table_name']}" p
+            inner join "{input_data['schema']}"."{input_data['data_table_name']}" ld
+                on p."timestamp" = ld."timestamp"
+            where
+                p.model_id = '{model_id}' and
+                p.prediction is not null and
+                ld.variable = '{target_name}' and
+                ld.value is not null and
+                p."timestamp" >= NOW() - INTERVAL '{interval_minutes} minutes'
+            order by
+                p."timestamp" desc;
+        """  # nosec B608 - values come from internal Temporal workflow config, not user input
+
+        target_data = await workflow.execute_activity_method(
+            Activities.load_custom_query,
+            {
+                **metadata,
+                'query': query,
+                'datetime_columns': ['timestamp'],
+                'orient': 'records',
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )
+
+        if not target_data:
+            return
+
+        simple_metrics = await workflow.execute_local_activity_method(
+            Activities.calculate_simple_metrics,
+            {
+                **metadata,
+                'model_id': model_id,
+                'target_data': target_data,
+                'metrics': input_data.get('metrics', ['rmse', 'mse', 'mae', 'r2']),
+                'interval_minutes': interval_minutes,
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )
+
+        if not simple_metrics:
+            return
+
+        await workflow.execute_activity_method(
+            Activities.export_data_to_postgres,
+            {
+                **metadata,
+                'data': simple_metrics,
+                'schema': input_data['schema'],
+                'table_name': input_data['target_table_name'],
+                'timestamp_conversion': {
+                    'column': 'timestamp',
+                    'format': DATETIME_FORMAT_WITH_TZ,
+                },
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=300),
+        )

laborious/workflows/sub_workflows/format_and_export_prediction.py

@@ -0,0 +1,220 @@
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from datetime import timedelta
+    from typing import Any
+
+    from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+    from sientia_do.temporal.policies import retry_policy
+
+    from laborious.activities.activities import Activities
+
+
+@workflow.defn(name='subworkflow.format_and_export_prediction')
+class FormatAndExportPrediction:
+    """
+    Data formatting and export workflow for prediction results.
+
+    This workflow handles the final stages of the prediction pipeline, including
+    data formatting, database persistence, OPC server export, and metrics recording.
+    It implements flexible formatting based on prediction quality and provides
+    comprehensive export capabilities to multiple destinations.
+
+    The workflow supports two main prediction paths:
+    1. Normal Prediction: Formats and exports successful prediction results
+    2. Default Prediction: Creates fallback predictions for error conditions
+
+    Export Destinations:
+    - PostgreSQL Database: Persistent storage with timestamp conversion
+    - PI Web API: Real-time industrial system integration for prediction and confidence values
+    - OPC Servers: Real-time industrial system integration
+    - Prometheus Metrics: Performance monitoring and operational visibility
+    """
+
+    @workflow.run
+    async def run(self, input_data: dict[str, Any]):
+        """
+        Execute the prediction formatting and export workflow.
+
+        This method orchestrates the complete data export process by:
+        1. Determining the appropriate formatting strategy based on path_flag
+        2. Formatting prediction data according to quality and requirements
+        3. Exporting data to PI Web API for real-time industrial access (if configured)
+        4. Exporting data to OPC servers for real-time industrial access (if configured)
+        5. Persisting data to PostgreSQL database with comprehensive metadata
+        6. Recording performance metrics for operational monitoring
+
+        The method implements flexible formatting strategies:
+        - Normal predictions: Full data formatting with confidence scores
+        - Error predictions: Default formatting with error indicators
+        - Comprehensive export: Multi-destination data distribution
+
+        Args:
+            input_data: Complete configuration for the export workflow
+                Required keys:
+                    - metadata (dict): Workflow execution metadata
+                    - path_flag (str | None): Decision path flag for formatting strategy
+                        - None: Normal prediction path with full formatting
+                        - Any other value: Default prediction path for error conditions
+                    - data (dict[str, Any]): Prediction data to format and export
+                    - prediction_confidence (float): Confidence score for the prediction
+                    - timestamp (str): ISO-formatted timestamp for the prediction
+                    - model_id (int): Unique identifier for the ML model
+                    - model_name (str): Name of the ML model
+                    - schema (str): Database schema for data storage
+                    - table_name (str): Target table for data persistence
+                Optional keys:
+                    - opc_output_config (dict[str, Any]): OPC server export configuration
+                    - pi_web_api_output_config (dict[str, Any]): PI Web API export configuration
+                        Contains endpoint, prediction_tags, and confidence_tags mappings
+                    - transformed_data (dict[str, Any]): Transformed data to export separately
+                        Only processed when path_flag is None
+                    - transform_table_name (str): Target table for transformed data export
+                        Required if transformed_data is provided
+                    - prediction_store_policy (str): Data retention policy (e.g., 'lts:1', 'erl:2')
+                        Required when path_flag is None
+                    - comment (str): Operational comment or error description
+                        Required when path_flag is not None
+
+        Returns:
+            None: The workflow completes successfully when all export operations finish
+
+        Note:
+            When transformed_data is provided and path_flag is None, the workflow will:
+            1. Format the transformed data using format_transformed_data
+            2. Export it to a separate table (transform_table_name) asynchronously
+            3. Wait for both prediction and transformed data exports to complete
+        """
+        metadata = input_data['metadata']
+        path_flag = input_data['path_flag']
+        data = input_data['data']
+        transformed_data = input_data.get('transformed_data', None)
+        prediction_confidence = input_data['prediction_confidence']
+
+        opc_output_config = input_data.get('opc_output_config', None)
+        pi_web_api_output_config = input_data.get('pi_web_api_output_config', None)
+
+        if path_flag is None:
+            # Normal prediction path: format prediction data with full metadata
+            prediction = await workflow.execute_local_activity_method(
+                Activities.format_prediction,
+                {
+                    **metadata,
+                    'data': data,
+                    'timestamp': input_data['timestamp'],
+                    'model_id': input_data['model_id'],
+                    'prediction_confidence': prediction_confidence,
+                    'prediction_store_policy': input_data['prediction_store_policy'],
+                    'model_name': input_data['model_name'],
+                },
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(seconds=60),
+            )
+
+            # Optionally format and export transformed data to separate table
+            if transformed_data is not None:
+                transformed = await workflow.execute_local_activity_method(
+                    Activities.format_transformed_data,
+                    {
+                        **metadata,
+                        'data': transformed_data,
+                        'model_id': input_data['model_id'],
+                        'model_name': input_data['model_name'],
+                    },
+                    retry_policy=retry_policy,
+                    start_to_close_timeout=timedelta(seconds=60),
+                )
+
+                write_transformed_handler = workflow.start_activity_method(
+                    Activities.export_payload_to_postgres,
+                    {
+                        **metadata,
+                        'schema': input_data['schema'],
+                        'table_name': input_data['transform_table_name'],
+                        'data': transformed,
+                        'timestamp_conversion': {
+                            'column': 'timestamp',
+                            'format': DATETIME_FORMAT_WITH_TZ,
+                        },
+                    },
+                    retry_policy=retry_policy,
+                    start_to_close_timeout=timedelta(seconds=60),
+                )
+
+            else:
+                write_transformed_handler = None
+
+        else:
+            # Error path: create default prediction with error indicators
+            prediction = await workflow.execute_local_activity_method(
+                Activities.format_default_prediction,
+                {
+                    **metadata,
+                    'timestamp': input_data['timestamp'],
+                    'model_id': input_data['model_id'],
+                    'prediction_confidence': prediction_confidence,
+                    'comment': input_data['comment'],
+                },
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(seconds=60),
+            )
+
+            write_transformed_handler = None
+
+        opc_metrics: dict[str, dict[str, float | None]] = {}
+
+        # write to pi web api
+        if pi_web_api_output_config:
+            prediction = await workflow.execute_activity_method(
+                Activities.write_pi_web_api_data,
+                {
+                    'pi_web_api_output_config': pi_web_api_output_config,
+                    'data': prediction,
+                    **metadata,
+                },
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(seconds=60),
+            )
+
+        # write to opc
+        if opc_output_config:
+            prediction, opc_metrics = await workflow.execute_activity_method(
+                Activities.write_opc_data,
+                {
+                    'opc_output_config': opc_output_config,
+                    'data': prediction,
+                    **metadata,
+                },
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(seconds=60),
+            )
+
+        # write to postgres
+        await workflow.execute_activity_method(
+            Activities.export_data_to_postgres,
+            {
+                **metadata,
+                'schema': input_data['schema'],
+                'table_name': input_data['table_name'],
+                'data': prediction,
+                'timestamp_conversion': {'column': 'timestamp', 'format': DATETIME_FORMAT_WITH_TZ},
+                'on_conflict': input_data.get('on_conflict', 'error'),
+                'unique_columns': ['model_id', 'timestamp'],
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=180),
+        )
+
+        if write_transformed_handler is not None:
+            await write_transformed_handler
+
+        await workflow.execute_activity_method(
+            Activities.write_metrics,
+            {
+                **metadata,
+                'prediction': prediction,
+                'opc_metrics': opc_metrics,
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(seconds=60),
+        )

laborious/workflows/sub_workflows/prediction_process.py

@@ -0,0 +1,346 @@
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from datetime import timedelta
+    from typing import Any
+
+    from sientia_do.temporal.policies import retry_policy
+
+    from laborious.activities.activities import Activities
+
+
+@workflow.defn(name='subworkflow.prediction_process')
+class PredictionProcess:
+    """
+    Core prediction processing workflow for the Laborious system.
+
+    This workflow implements the complete ML model inference pipeline, handling
+    data quality validation, MLFlow model interactions, and prediction processing.
+    It serves as the central orchestrator for all prediction operations and ensures
+    data quality throughout the entire process.
+
+    The workflow implements a robust data processing pipeline with:
+    - Data quality validation using configurable filters
+    - MLFlow model transformation and prediction
+    - Response validation and quality assurance
+    - Flexible decision path handling
+    - Comprehensive error handling and retry policies
+
+    Workflow Execution:
+        1. Timestamp Retrieval: Gets last processed timestamp for incremental processing
+        2. Input Data Gate: Applies data quality filters
+        3. Path Decision: Determines processing path based on filter results
+        4. MLFlow Transform: Requests data transformation using MLFlow models
+        5. Response Validation: Filters transform responses for quality assurance
+        6. MLFlow Prediction: Executes prediction using transformed data
+        7. Content Validation: Filters prediction responses for final quality check
+        8. Export Delegation: Delegates to FormatAndExportPrediction workflow
+    """
+
+    @workflow.run
+    async def run(self, input_data: dict[str, Any]):
+        """
+        Execute the prediction process workflow.
+
+        This method orchestrates the complete prediction processing pipeline by:
+        1. Retrieving the last processed timestamp for incremental processing
+        2. Applying data quality filters to validate input data
+        3. Executing MLFlow model transformation and prediction
+        4. Validating all responses for quality assurance
+        5. Delegating to export workflow for data persistence
+
+        The method implements comprehensive error handling and ensures all
+        data quality requirements are met before proceeding with ML operations.
+
+        Args:
+            input_data: Complete configuration for the prediction process
+                Required keys:
+                    - metadata (dict): Workflow execution metadata
+                    - data (dict): Input data for prediction processing
+                    - schema (dict): Data schema definition
+                    - table_name (str): Target table for predictions
+                    - model_id (str): ML model identifier
+                    - model_name (str): ML model name
+                    - input_filters (dict): Data quality filters
+                    - mlflow_transform_filters (dict): MLFlow transform filters
+                    - mlflow_predict_filters (dict): MLFlow prediction filters
+                    - model_retention (int): Model retention period in minutes
+                    - path_priority (list[str]): Decision path priority configuration
+                    - opc_output_config (dict, optional): OPC server export configuration
+                    - pi_web_api_output_config (dict, optional): PI Web API export configuration
+                    - save_transform (bool, optional): Whether to save transformed data (default: True)
+                    - prediction_store_policy (str, optional): Data retention policy (default: 'lts:1')
+
+        Returns:
+            None: The workflow completes successfully when export workflow finishes
+
+        Raises:
+            Exception: If any required parameters are missing or if the workflow fails
+                during data processing, MLFlow operations, or workflow delegation
+
+        """
+
+        metadata = input_data['metadata']
+        data = input_data['data']
+        model_id = input_data['model_id']
+        model_name = input_data['model_name']
+        model_config = input_data.get('model_config', {})
+        save_transform = input_data.get('save_transform', True)
+
+        try:
+            await self._run_prediction_pipeline(
+                input_data,
+                metadata,
+                data,
+                model_id,
+                model_name,
+                model_config,
+                save_transform,
+            )
+            await workflow.execute_activity_method(
+                Activities.cleanup_minio_objects_expired,
+                {**metadata, 'data': data},
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(minutes=5),
+            )
+        except Exception as e:
+            await workflow.execute_activity_method(
+                Activities.cleanup_minio_objects_expired,
+                {**metadata, 'data': data},
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(minutes=5),
+            )
+            raise e
+
+    async def _run_prediction_pipeline(
+        self,
+        input_data: dict[str, Any],
+        metadata: dict[str, Any],
+        data: dict[str, Any],
+        model_id: str,
+        model_name: str,
+        model_config: dict[str, Any],
+        save_transform: bool,
+    ) -> None:
+        last_timestamp = data['last_timestamp']
+
+        # Apply input data quality gates
+        gate_input = {
+            **metadata,
+            'filters': input_data['input_filters'],
+            'data': data,
+            'path_priority': input_data['path_priority'],
+        }
+
+        path_flag, confidence, comment = await workflow.execute_local_activity_method(
+            Activities.input_gate,
+            gate_input,
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(minutes=1),
+        )
+
+        # Handle path decision based on filter results
+        if await self.path_flag_handler(
+            data, path_flag, input_data, confidence, last_timestamp, comment
+        ):
+            return
+
+        # Request MLFlow model transformation
+        transformed_data = await workflow.execute_activity_method(
+            Activities.request_transform,
+            {**metadata, 'data': data, 'model_name': model_name, 'model_config': model_config},
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(minutes=5),
+        )
+
+        # Validate MLFlow transform response
+        path_flag, confidence, comment = await workflow.execute_local_activity_method(
+            Activities.mlflow_response_gate,
+            {
+                **metadata,
+                'filters': input_data['mlflow_transform_filters'],
+                'data': transformed_data,
+                'type': 'transform',
+                'path_priority': input_data['path_priority'],
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(minutes=1),
+        )
+
+        # Handle path decision based on transform validation
+        if await self.path_flag_handler(
+            data, path_flag, input_data, confidence, last_timestamp, comment
+        ):
+            return
+
+        path_flag, confidence, comment = await workflow.execute_local_activity_method(
+            Activities.mlflow_content_gate,
+            {
+                **metadata,
+                'filters': input_data['mlflow_transform_filters'],
+                'data': transformed_data,
+                'type': 'transform',
+                'path_priority': input_data['path_priority'],
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(minutes=1),
+        )
+
+        if await self.path_flag_handler(
+            data, path_flag, input_data, confidence, last_timestamp, comment
+        ):
+            return
+
+        predicted_data = await workflow.execute_activity_method(
+            Activities.request_predict,
+            {
+                **metadata,
+                'data': transformed_data,
+                'model_name': model_name,
+                'model_config': model_config,
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(minutes=5),
+        )
+
+        # Validate MLFlow prediction response
+        path_flag, confidence, comment = await workflow.execute_local_activity_method(
+            Activities.mlflow_response_gate,
+            {
+                **metadata,
+                'filters': input_data['mlflow_predict_filters'],
+                'data': predicted_data,
+                'type': 'predict',
+                'path_priority': input_data['path_priority'],
+            },
+            retry_policy=retry_policy,
+            start_to_close_timeout=timedelta(minutes=1),
+        )
+
+        # Handle path decision based on prediction validation
+        if await self.path_flag_handler(
+            data, path_flag, input_data, confidence, last_timestamp, comment
+        ):
+            return
+
+        # Delegate to export workflow for data persistence
+        await workflow.execute_child_workflow(
+            'subworkflow.format_and_export_prediction',
+            {
+                'metadata': metadata,
+                'on_conflict': input_data.get('on_conflict', 'error'),
+                'path_flag': path_flag,
+                'data': predicted_data,
+                'transformed_data': transformed_data if save_transform else None,
+                'prediction_confidence': confidence,
+                'timestamp': last_timestamp,
+                'model_id': model_id,
+                'model_name': model_name,
+                'model_config': model_config,
+                'opc_output_config': input_data['opc_output_config'],
+                'pi_web_api_output_config': input_data['pi_web_api_output_config'],
+                'schema': input_data['schema'],
+                'table_name': input_data['table_name'],
+                'transform_table_name': input_data['transform_table_name'],
+                'comment': comment,
+                'prediction_store_policy': input_data['prediction_store_policy'],
+            },
+        )
+
+    async def path_flag_handler(
+        self,
+        data: dict[str, Any],
+        path_flag: str | None,
+        input_data: dict,
+        confidence: int,
+        last_timestamp: str,
+        comment: str,
+    ) -> bool:
+        """
+        Handle path decisions based on filter results and confidence levels.
+
+        This method determines the appropriate action based on the path flag
+        returned by data quality filters. It can stop processing, continue,
+        or repeat operations based on the configured path priority.
+
+        Args:
+            data: Input data for processing
+            path_flag: Path decision from filter (STOP, CONTINUE, REPEAT)
+            input_data: Complete workflow input configuration including:
+                - metadata (dict): Workflow execution metadata
+                - schema (str): Database schema
+                - table_name (str): Target table for predictions
+                - transform_table_name (str): Target table for transformed data
+                - model_id (str): ML model identifier
+                - model_name (str): ML model name
+                - model_config (dict, optional): Model configuration
+                - opc_output_config (dict, optional): OPC server export configuration
+                - pi_web_api_output_config (dict, optional): PI Web API export configuration
+                - prediction_store_policy (str, optional): Data retention policy
+            confidence: Confidence level from filter validation
+            last_timestamp: Last processed timestamp
+            comment: Additional information about the filter result
+
+        Returns:
+            bool: True if processing should stop, False to continue
+
+        Path Handling:
+            - STOP: Terminates workflow execution
+            - CONTINUE: Delegates to FormatAndExportPrediction workflow with current data
+            - REPEAT: Repeats last prediction if available
+        """
+        metadata = input_data['metadata']
+
+        schema = input_data['schema']
+        table_name = input_data['table_name']
+        transform_table_name = input_data['transform_table_name']
+        model_id = input_data['model_id']
+        model_name = input_data['model_name']
+        model_config = input_data.get('model_config', {})
+
+        path_flag = path_flag.upper() if path_flag else ''
+
+        if path_flag == 'STOP':
+            # Stop processing and exit workflow
+            return True
+        elif path_flag == 'REPEAT':
+            # Repeat last prediction if available
+            await workflow.execute_activity_method(
+                Activities.repeat_last_prediction,
+                {
+                    **metadata,
+                    'schema': schema,
+                    'table_name': table_name,
+                    'model': model_id,
+                    'last_timestamp': last_timestamp,
+                },
+                retry_policy=retry_policy,
+                start_to_close_timeout=timedelta(minutes=1),
+            )
+            return True
+        elif path_flag == 'CONTINUE':
+            # call write workflow
+            await workflow.execute_child_workflow(
+                'subworkflow.format_and_export_prediction',
+                {
+                    'metadata': metadata,
+                    'path_flag': path_flag,
+                    'data': data,
+                    'prediction_confidence': confidence,
+                    'timestamp': last_timestamp,
+                    'model_id': model_id,
+                    'model_name': model_name,
+                    'model_config': model_config,
+                    'schema': schema,
+                    'table_name': table_name,
+                    'transform_table_name': transform_table_name,
+                    'comment': comment,
+                    'opc_output_config': input_data['opc_output_config'],
+                    'pi_web_api_output_config': input_data['pi_web_api_output_config'],
+                    'prediction_store_policy': input_data['prediction_store_policy'],
+                    'on_conflict': input_data.get('on_conflict', 'error'),
+                },
+            )
+            return True
+
+        return False

model_convert.ipynb

@@ -0,0 +1,106 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "id": "e838ff21",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import csv\n",
+    "\n",
+    "def csv_to_tag_lists(csv_path: str) -> dict:\n",
+    "    read_tags = []\n",
+    "    write_tags = []\n",
+    "\n",
+    "    def to_float(val):\n",
+    "        try:\n",
+    "            return float(str(val).strip())\n",
+    "        except Exception:\n",
+    "            return None\n",
+    "\n",
+    "    with open(csv_path, newline=\"\", encoding=\"utf-8\") as f:\n",
+    "        reader = csv.DictReader(f)\n",
+    "        for row in reader:\n",
+    "            # Basic normalization\n",
+    "            op = (row.get(\"operation\") or \"\").strip()\n",
+    "\n",
+    "            if op == \"READ\":\n",
+    "                # Build common tag payload with required mappings\n",
+    "                tag = {\n",
+    "                    \"server_id\": \"1\",\n",
+    "                    \"tag_address\": row.get(\"opc_tag\"),\n",
+    "                    \"tag_name\": row.get(\"name\"),\n",
+    "                    \"data_range\": [to_float(row.get(\"min_value\")), to_float(row.get(\"max_value\"))],\n",
+    "                    \"aggr_func\": row.get(\"aggregation_func\").lower(),\n",
+    "                    # keep other fields with their original names\n",
+    "                    \"frequency\": row.get(\"frequency\"),\n",
+    "                    \"local\": row.get(\"local\"),\n",
+    "                    \"area\": row.get(\"area\"),\n",
+    "                    \"description\": row.get(\"description\"),\n",
+    "                }\n",
+    "\n",
+    "                read_tags.append(tag)\n",
+    "\n",
+    "            else:\n",
+    "                tag = {\n",
+    "                    \"server_id\": \"1\",\n",
+    "                    \"addr\": row.get(\"opc_tag\"),\n",
+    "                    \"tag_name\": row.get(\"name\"),\n",
+    "                    \"local\": row.get(\"local\"),\n",
+    "                    \"area\": row.get(\"area\"),\n",
+    "                    \"description\": row.get(\"description\"),\n",
+    "                }\n",
+    "                \n",
+    "            if op == \"WRITE_PREDICTION\":\n",
+    "                tag[\"type\"] = \"prediction\"\n",
+    "                write_tags.append(tag)\n",
+    "            elif op == \"WRITE_CONFIDENCE\":\n",
+    "                tag[\"type\"] = \"confidence\"\n",
+    "                write_tags.append(tag)\n",
+    "            # ignore any other operation values silently\n",
+    "\n",
+    "    return {\"read_tags\": read_tags, \"write_tags\": write_tags}"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "4621cd43",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import json\n",
+    "\n",
+    "file_names = [\"Courier - Página1.csv\"]\n",
+    "\n",
+    "for file_name in file_names:\n",
+    "    write_file = file_name.replace(\".csv\", \".json\")\n",
+    "\n",
+    "    with open(write_file, \"w\", encoding=\"utf-8\") as f:\n",
+    "        json.dump(csv_to_tag_lists(file_name), f, indent=2, ensure_ascii=False)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "venv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

pyproject.toml

@@ -0,0 +1,158 @@
+[build-system]
+requires = ["setuptools>=61.0"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "laborious"
+version = "0.0.0"
+description = "Sientia DataOps Laborious - ML Model Orchestration System"
+readme = "README.md"
+requires-python = ">=3.11"
+authors = [
+    {name = "Aignosi", email = "dev@aignosi.com"}
+]
+
+[tool.ruff]
+line-length = 100
+target-version = "py311"
+exclude = [
+    ".git",
+    ".venv",
+    "venv",
+    "__pycache__",
+    "*.pyc",
+    ".pytest_cache",
+    "htmlcov",
+    "tests/laborious/workflows/subworkflows/test_prediction_process.py",
+]
+
+[tool.ruff.lint]
+select = [
+    "E",   # pycodestyle errors
+    "W",   # pycodestyle warnings
+    "F",   # pyflakes
+    "I",   # isort
+    "B",   # flake8-bugbear
+    "C4",  # flake8-comprehensions
+    "UP",  # pyupgrade
+    "N",   # pep8-naming
+    "YTT", # flake8-2020
+    "S",   # flake8-bandit
+    "BLE", # flake8-blind-except
+    "A",   # flake8-builtins
+    "C90", # mccabe complexity
+]
+
+ignore = [
+    "BLE001",  # ignore blind except, we need to send notifications with any error
+    "E501",  # line too long (handled by formatter)
+    "S101",  # use of assert (needed for tests)
+    "S105",  # possible hardcoded password (false positives)
+    "S106",  # possible hardcoded password (false positives)
+    "S608",  # potential sql injection (false positives)
+    "N802",  # function name should be lowercase (temporal decorators)
+    "N806",  # variable in function should be lowercase
+]
+
+[tool.ruff.lint.per-file-ignores]
+"tests/**/*.py" = [
+    "S101",  # assert allowed in tests
+    "S105",  # hardcoded passwords ok in tests
+    "S106",  # hardcoded passwords ok in tests
+]
+
+[tool.ruff.lint.mccabe]
+max-complexity = 15
+
+[tool.ruff.format]
+quote-style = "single"
+indent-style = "space"
+line-ending = "auto"
+
+[tool.mypy]
+python_version = "3.11"
+warn_return_any = false
+warn_unused_configs = true
+disallow_untyped_defs = false
+disallow_incomplete_defs = false
+check_untyped_defs = true
+no_implicit_optional = true
+warn_redundant_casts = true
+warn_unused_ignores = false
+warn_no_return = true
+strict_equality = true
+ignore_missing_imports = true
+
+# Ignore missing imports for external packages
+[[tool.mypy.overrides]]
+module = "temporalio.*"
+ignore_missing_imports = true
+
+[[tool.mypy.overrides]]
+module = "sientia_do.*"
+ignore_missing_imports = true
+
+[[tool.mypy.overrides]]
+module = "mlflow.*"
+ignore_missing_imports = true
+
+[[tool.mypy.overrides]]
+module = "prometheus_client.*"
+ignore_missing_imports = true
+
+[[tool.mypy.overrides]]
+module = "sientia.*"
+ignore_missing_imports = true
+
+[[tool.mypy.overrides]]
+module = "pandas.*"
+ignore_missing_imports = true
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+python_files = ["test_*.py"]
+python_classes = ["Test*"]
+python_functions = ["test_*"]
+addopts = [
+    "-v",
+    "--strict-markers",
+]
+markers = [
+    "asyncio: marks tests as async",
+    "integration: marks tests as integration tests",
+    "unit: marks tests as unit tests",
+    "opc: marks E2E tests that use in-process asyncua + real OpcRepository",
+]
+
+[tool.coverage.run]
+source = ["laborious"]
+omit = [
+    "*/tests/*",
+    "*/venv/*",
+    "*/__pycache__/*",
+    "*/site-packages/*",
+]
+branch = true
+
+[tool.coverage.report]
+precision = 2
+show_missing = true
+skip_covered = false
+exclude_lines = [
+    "pragma: no cover",
+    "def __repr__",
+    "def __str__",
+    "raise AssertionError",
+    "raise NotImplementedError",
+    "if __name__ == .__main__.:",
+    "if TYPE_CHECKING:",
+    "class .*\\bProtocol\\):",
+    "@(abc\\.)?abstractmethod",
+]
+
+[tool.coverage.html]
+directory = "htmlcov"
+
+[tool.bandit]
+exclude_dirs = ["tests", "venv", ".venv"]
+skips = ["B101", "B601", "B608"]  # Skip assert, shell injection, and SQL injection (false positives)

requirements-dev.txt

@@ -0,0 +1,21 @@
+# Development and Testing Dependencies
+# These packages are only needed for development, testing, and code quality checks
+# Install with: pip install -r requirements-dev.txt
+
+# Code Quality & Linting
+ruff>=0.1.0              # Fast Python linter and formatter (replaces flake8, black, isort)
+mypy>=1.7.0              # Static type checker
+bandit>=1.7.5            # Security vulnerability scanner
+pandas-stubs>=2.0.0      # Type stubs for pandas
+types-requests>=2.31.0   # Type stubs for requests
+
+# Testing
+pytest>=7.4.0            # Testing framework
+pytest-cov>=4.1.0        # Coverage plugin for pytest
+pytest-asyncio>=0.21.0   # Async test support (already in main requirements)
+testcontainers[postgres,minio] # PostgreSQL and MinIO containers for E2E tests
+
+# Development Tools
+ipython>=8.12.0          # Enhanced Python shell
+ipdb>=0.13.13            # IPython debugger
+ipykernel==6.30.1        # IPython kernel for Jupyter notebooks

requirements-local.txt

@@ -0,0 +1,18 @@
+temporalio
+psycopg2-binary
+sqlalchemy
+asyncua==1.0.6
+redis
+git+ssh://git@github.com/Aignosi/sientia-dataops-library.git@1.12.1
+git+ssh://git@github.com/Aignosi/sientia-model-library.git@0.10.0
+prometheus-client   
+botocore
+boto3
+s3fs
+pyarrow
+kaleido
+hyperopt
+shap
+pycurl
+scipy<1.14.0
+scikit-learn==1.5.2

requirements.txt

@@ -0,0 +1,18 @@
+temporalio
+psycopg2-binary
+sqlalchemy
+asyncua==1.0.6
+redis
+sientia_do>=1.12.1
+sientia_model>=0.8.2
+prometheus-client
+botocore
+boto3
+s3fs
+pyarrow
+kaleido
+hyperopt
+shap
+pycurl
+scipy<1.14.0
+scikit-learn==1.5.2

run_coverage.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+
+# Exit on any error
+set -e
+
+echo "Activating virtual environment..."
+source ./venv/bin/activate
+
+pytest --cov=laborious --cov-report=html
+
+xdg-open htmlcov/index.html

run_local.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+
+# Exit on any error
+set -e
+
+echo "Activating virtual environment..."
+source ./venv/bin/activate
+
+echo "Loading environment variables from .env..."
+if [ -f .env ]; then
+    export $(cat .env | grep -v '^#' | xargs)
+    echo "Environment variables loaded from .env"
+else
+    echo "Warning: .env file not found. Continuing without environment variables."
+fi
+
+echo "Starting ingestor application..."
+python -m laborious.worker.worker

sonar-project.properties

@@ -0,0 +1,11 @@
+sonar.projectKey=Aignosi_sientia-dataops-laborious_temporal_beaec423-6c42-4f26-8134-b676287b499d
+sonar.projectName=sientia-dataops-laborious_temporal
+sonar.sources=laborious
+sonar.tests=tests
+sonar.projectVersion=1.0.0
+sonar.coverage.exclusions=laborious/worker/*
+sonar.qualitygate.wait=true
+sonar.qualitygate.timeout=300
+sonar.python.coverage.reportPaths=coverage.xml
+sonar.python.xunit.reportPath=pytest.xml
+sonar.python.version=3.11

tests/conftest.py

@@ -0,0 +1,66 @@
+import os
+
+from sientia_do.temporal.activities.postgres_sync import Postgres
+
+
+def _noop_postgres_del(_self):
+    """
+    Unit tests use MagicMock metrics controllers; postgres_sync.Postgres.__del__ calls
+    close() during GC and triggers async shutdown. Explicit ``close()`` is covered in tests.
+    """
+    return None
+
+
+Postgres.__del__ = _noop_postgres_del  # type: ignore[method-assign]
+
+# The production code converts SIENTIA_MINIO_OFFLOAD_THRESHOLD_MEGABYTES to int at import-time.
+# Tests must set it to a valid integer string to avoid import errors.
+os.environ.setdefault('SIENTIA_MINIO_OFFLOAD_THRESHOLD_MEGABYTES', '1')
+
+
+class DummyMinioDataFramePayload:
+    """
+    Minimal payload double used by unit tests.
+
+    The production workflow/gates expect a MinioDataFramePayload-like object with:
+    - async retrieve(minio_repo, workflow_metadata) -> DataFrame | dict
+    - has_data() -> bool
+    - cleanup_prefix() -> str | None
+    - last_timestamp: attribute
+    - status: attribute
+    """
+
+    def __init__(
+        self,
+        *,
+        retrieve_return=None,
+        has_data: bool = True,
+        cleanup_prefix: str | None = None,
+        last_timestamp: str = '2024-01-01',
+        status: dict | None = None,
+    ):
+        self._retrieve_return = retrieve_return
+        self._has_data = has_data
+        self._cleanup_prefix = cleanup_prefix
+        self.last_timestamp = last_timestamp
+        self.status = status
+
+    async def retrieve(self, _minio_repo, _workflow_metadata=None):
+        return self._retrieve_return
+
+    def has_data(self) -> bool:
+        return self._has_data
+
+    def cleanup_prefix(self) -> str | None:
+        return self._cleanup_prefix
+
+
+"""
+Pytest configuration file with global mocks for external dependencies.
+
+The historical ``sientia`` package is no longer imported by the codebase;
+drift analysis lives in ``sientia_model.analytics.drift_analysis`` and is
+imported lazily inside Temporal activities. No global module-level mock is
+required here — unit tests that need to control ``DriftAnalysis`` outputs
+should patch ``laborious.activities.model_metrics.DriftAnalysis`` directly.
+"""

tests/laborious/activities/test_activities.py

@@ -0,0 +1,299 @@
+from unittest.mock import ANY, MagicMock, patch
+
+from laborious.activities.activities import Activities
+from laborious.activities.api import API
+from laborious.activities.gates import Gates
+from laborious.activities.mlflow import MLFlow
+from laborious.activities.model_metrics import ModelMetrics
+from laborious.activities.opc import OPC
+from laborious.activities.storage import Storage
+
+
+@patch('laborious.activities.activities.Storage.__init__')
+@patch('laborious.activities.activities.MLFlow.__init__')
+@patch('laborious.activities.activities.OPC.__init__')
+@patch('laborious.activities.activities.Gates.__init__')
+@patch('laborious.activities.activities.ModelMetrics.__init__')
+@patch('laborious.activities.activities.API.__init__')
+@patch('laborious.activities.activities.MinioRepository')
+@patch('laborious.activities.activities.MetricsController')
+def test___init__(
+    mock_metrics_controller,
+    mock_minio_repository,
+    mock_api_init,
+    mock_model_metrics_init,
+    mock_gates_init,
+    mock_opc_init,
+    mock_mlflow_init,
+    mock_storage_init,
+):
+    postgres_config = {
+        'host': 'localhost',
+        'port': 5432,
+        'user': 'postgres',
+        'password': 'postgres',
+        'dbname': 'postgres',
+        'min_connections': 1,
+        'max_connections': 10,
+    }
+
+    minio_config = {
+        'endpoint_url': 'localhost:9000',
+        'access_key': 'minio',
+        'secret_key': 'minio123',
+        'default_bucket': 'test',
+        'retention_hours': 24,
+        'secure': False,
+    }
+
+    mlflow_repository = MagicMock()
+    plugin_store = MagicMock()
+
+    opc_config = {
+        'bootstrap_servers': 'localhost:9092',
+        'polling_time': 1000,
+        'group_id': 'test-group',
+    }
+
+    pi_web_api_config = {
+        'base_url': 'https://test-pi-server.com',
+        'auth_type': 'bearer',
+        'auth_token': 'test_token',
+    }
+
+    logger = MagicMock()
+    notification_handler = MagicMock()
+
+    activities = Activities(
+        postgres_config=postgres_config,
+        plugin_store=plugin_store,
+        minio_config=minio_config,
+        opc_config=opc_config,
+        pi_web_api_config=pi_web_api_config,
+        logger=logger,
+        notification_handler=notification_handler,
+        mlflow_repository=mlflow_repository,
+    )
+
+    assert isinstance(activities, Activities)
+    assert isinstance(activities, Storage)
+    assert isinstance(activities, MLFlow)
+    assert isinstance(activities, OPC)
+    assert isinstance(activities, Gates)
+    assert isinstance(activities, ModelMetrics)
+    assert isinstance(activities, API)
+
+    mock_storage_init.assert_called_once_with(
+        ANY,
+        host=postgres_config['host'],
+        port=postgres_config['port'],
+        user=postgres_config['user'],
+        password=postgres_config['password'],
+        dbname=postgres_config['dbname'],
+        min_connections=postgres_config['min_connections'],
+        max_connections=postgres_config['max_connections'],
+        retention_hours=minio_config['retention_hours'],
+        minio_repository=mock_minio_repository.return_value,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )
+
+    mock_mlflow_init.assert_called_once_with(
+        ANY,
+        mlflow_repository=mlflow_repository,
+        plugin_store=plugin_store,
+        minio_repository=mock_minio_repository.return_value,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )
+
+    mock_opc_init.assert_called_once_with(
+        ANY,
+        opc_servers=opc_config,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )
+
+    mock_gates_init.assert_called_once_with(
+        ANY,
+        minio_repository=mock_minio_repository.return_value,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )
+
+    mock_model_metrics_init.assert_called_once_with(
+        ANY,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )
+
+    mock_api_init.assert_called_once_with(
+        ANY,
+        base_url=pi_web_api_config['base_url'],
+        auth_type=pi_web_api_config['auth_type'],
+        auth_token=pi_web_api_config['auth_token'],
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )
+
+    mock_minio_repository.assert_called_once_with(
+        endpoint=minio_config['endpoint_url'],
+        access_key=minio_config['access_key'],
+        secret_key=minio_config['secret_key'],
+        bucket=minio_config['default_bucket'],
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+        secure=minio_config['secure'],
+    )
+
+
+@patch('laborious.activities.activities.Storage')
+@patch('laborious.activities.activities.MLFlow')
+@patch('laborious.activities.activities.OPC')
+@patch('laborious.activities.activities.Gates')
+@patch('laborious.activities.activities.ModelMetrics')
+@patch('laborious.activities.activities.API')
+@patch('laborious.activities.activities.MinioRepository')
+def test_shutdown(
+    _mock_minio_repository,
+    mock_api_init,
+    mock_model_metrics_init,
+    mock_gates_init,
+    mock_opc_init,
+    mock_mlflow_init,
+    mock_storage_init,
+):
+    mock_opc_init.close = MagicMock()
+    postgres_config = {
+        'host': 'localhost',
+        'port': 5432,
+        'user': 'postgres',
+        'password': 'postgres',
+        'dbname': 'postgres',
+        'min_connections': 1,
+        'max_connections': 10,
+    }
+
+    minio_config = {
+        'endpoint_url': 'localhost:9000',
+        'access_key': 'minio',
+        'secret_key': 'minio123',
+        'default_bucket': 'test',
+        'retention_hours': 24,
+        'secure': False,
+    }
+
+    mlflow_repository = MagicMock()
+    plugin_store = MagicMock()
+
+    opc_config = {
+        'bootstrap_servers': 'localhost:9092',
+        'polling_time': 1000,
+        'group_id': 'test-group',
+    }
+
+    pi_web_api_config = {
+        'base_url': 'https://test-pi-server.com',
+        'auth_type': 'bearer',
+        'auth_token': 'test_token',
+    }
+
+    logger = MagicMock()
+    notification_handler = MagicMock()
+
+    activities = Activities(
+        postgres_config=postgres_config,
+        plugin_store=plugin_store,
+        minio_config=minio_config,
+        opc_config=opc_config,
+        pi_web_api_config=pi_web_api_config,
+        logger=logger,
+        notification_handler=notification_handler,
+        mlflow_repository=mlflow_repository,
+    )
+
+    activities.shutdown()
+    mock_opc_init.close.assert_called_once()
+    mock_storage_init.close.assert_called_once()
+    mock_mlflow_init.close.assert_called_once()
+    mock_gates_init.close.assert_called_once()
+    mock_model_metrics_init.close.assert_called_once()
+    mock_api_init.close.assert_called_once()
+
+
+@patch('laborious.activities.activities.SientiaMLflowRepository')
+@patch('laborious.activities.activities.build_mlflow_config')
+@patch('laborious.activities.activities.Storage.__init__')
+@patch('laborious.activities.activities.MLFlow.__init__')
+@patch('laborious.activities.activities.OPC.__init__')
+@patch('laborious.activities.activities.Gates.__init__')
+@patch('laborious.activities.activities.ModelMetrics.__init__')
+@patch('laborious.activities.activities.API.__init__')
+@patch('laborious.activities.activities.MinioRepository')
+@patch('laborious.activities.activities.MetricsController')
+def test___init___builds_mlflow_repository_when_not_provided(
+    mock_metrics_controller,
+    mock_minio_repository,
+    _mock_api_init,
+    _mock_model_metrics_init,
+    _mock_gates_init,
+    _mock_opc_init,
+    _mock_mlflow_init,
+    _mock_storage_init,
+    mock_build_mlflow_config,
+    mock_mlflow_repository_cls,
+):
+    postgres_config = {
+        'host': 'localhost',
+        'port': 5432,
+        'user': 'postgres',
+        'password': 'postgres',
+        'dbname': 'postgres',
+        'min_connections': 1,
+        'max_connections': 10,
+    }
+    minio_config = {
+        'endpoint_url': 'localhost:9000',
+        'access_key': 'minio',
+        'secret_key': 'minio123',
+        'default_bucket': 'test',
+        'retention_hours': 24,
+        'secure': False,
+    }
+    opc_config = {'bootstrap_servers': 'localhost:9092', 'polling_time': 1000, 'group_id': 'test'}
+    pi_web_api_config = {'base_url': 'https://pi', 'auth_type': 'bearer', 'auth_token': 'token'}
+    logger = MagicMock()
+    notification_handler = MagicMock()
+    plugin_store = MagicMock()
+    mock_build_mlflow_config.return_value = {
+        'url': 'http://mlflow:80',
+        'username': 'u',
+        'password': 'p',
+    }
+
+    Activities(
+        postgres_config=postgres_config,
+        plugin_store=plugin_store,
+        minio_config=minio_config,
+        opc_config=opc_config,
+        pi_web_api_config=pi_web_api_config,
+        logger=logger,
+        notification_handler=notification_handler,
+    )
+
+    mock_build_mlflow_config.assert_called_once()
+    mock_mlflow_repository_cls.assert_called_once_with(
+        host='http://mlflow:80',
+        username='u',
+        password='p',
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=mock_metrics_controller.return_value,
+    )

tests/laborious/activities/test_api.py

@@ -0,0 +1,503 @@
+from unittest.mock import ANY, MagicMock, call, patch
+
+from pytest import fixture
+from sientia_do.notifications.models import NotificationLevel
+
+from laborious.activities.api import API, PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schema_name': 'test_schedule',
+    },
+}
+
+
+def _create_mock_dataframe(to_dict_return=None):
+    """Helper function to create a mocked DataFrame for testing."""
+    mock_df = MagicMock()
+    mock_head = MagicMock()
+
+    def get_column_values(key):
+        if key == 'prediction':
+            return MagicMock(values=[0.75])
+        elif key == 'prediction_confidence':
+            return MagicMock(values=[0.95])
+        else:
+            return MagicMock(values=['2024-01-01T00:00:00+00:00'])
+
+    mock_head.__getitem__.side_effect = get_column_values
+    mock_df.head.return_value = mock_head
+
+    if to_dict_return is None:
+        to_dict_return = {
+            'prediction': [0.75],
+            'prediction_confidence': [0.95],
+            'timestamp': ['2024-01-01T00:00:00+00:00'],
+        }
+    mock_df.to_dict.return_value = to_dict_return
+
+    return mock_df
+
+
+@fixture
+def base_input_data():
+    """Base input data for PI Web API tests."""
+    return {
+        **metadata,
+        'data': {
+            'prediction': [0.75],
+            'prediction_confidence': [0.95],
+            'timestamp': ['2024-01-01T00:00:00+00:00'],
+        },
+        'pi_web_api_output_config': {
+            'endpoint': 'https://test-pi-server.com/piwebapi',
+            'prediction_tags': {'tag1': 'web_id_1'},
+            'confidence_tags': {'tag2': 'web_id_2'},
+        },
+    }
+
+
+@patch('laborious.activities.api.PIWebAPIClient')
+def test_get_pi_web_api_core_labels_without_operation_type(mock_pi_web_api_client):
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+
+    api_instance = API(
+        base_url='https://test-pi-server.com',
+        auth_type='bearer',
+        auth_token='test_token',
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+    with patch.object(
+        SientiaMonitoring,
+        'get_core_labels',
+        return_value={
+            'pod_id': 'test_pod',
+            'runtime': 'local',
+            'model_name': 'test_model',
+            'workflow_name': 'test_workflow',
+            'operation_type': 'write_pi_web_api_data',
+        },
+    ):
+        labels = api_instance.get_pi_web_api_core_labels(metadata=metadata['metadata'])
+    assert labels['operation_type'] == 'write_pi_web_api_data'
+    assert labels == {
+        'pod_id': 'test_pod',
+        'runtime': 'local',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'operation_type': 'write_pi_web_api_data',
+    }
+
+
+@patch('laborious.activities.api.PIWebAPIClient')
+def test_get_pi_web_api_core_labels_with_operation_type(mock_pi_web_api_client):
+    from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+
+    api_instance = API(
+        base_url='https://test-pi-server.com',
+        auth_type='bearer',
+        auth_token='test_token',
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+    with patch.object(
+        SientiaMonitoring,
+        'get_core_labels',
+        return_value={
+            'pod_id': 'test_pod',
+            'runtime': 'k8s',
+            'model_name': 'test_model',
+            'workflow_name': 'test_workflow',
+            'operation_type': 'write',
+        },
+    ):
+        labels = api_instance.get_pi_web_api_core_labels(
+            metadata=metadata['metadata'], operation_type='write'
+        )
+    assert labels['operation_type'] == 'write'
+    assert labels['runtime'] == 'k8s'
+
+
+def test__init__():
+    api = API(
+        base_url='https://test-pi-server.com',
+        auth_type='bearer',
+        auth_token='test_token',
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+
+    assert api.pi_web_api_client is not None
+
+
+@fixture
+@patch('laborious.activities.api.PIWebAPIClient')
+def api(mock_pi_web_api_client):
+    mock_client = MagicMock()
+    mock_client.write_value = MagicMock()
+    mock_client.close = MagicMock()
+    mock_client.base_url = 'https://test-pi-server.com'
+    mock_pi_web_api_client.return_value = mock_client
+
+    api_instance = API(
+        base_url='https://test-pi-server.com',
+        auth_type='bearer',
+        auth_token='test_token',
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+    api_instance.send_notification = MagicMock()
+    api_instance.info = MagicMock()
+    api_instance.error = MagicMock()
+    api_instance.emit_metric_sync = MagicMock()
+    api_instance.get_core_labels = MagicMock(
+        return_value={
+            'pod_id': 'test_pod',
+            'runtime': 'local',
+            'model_name': 'test_model',
+            'workflow_name': 'test_workflow',
+        }
+    )
+    return api_instance
+
+
+@patch('laborious.activities.api.DataFrame')
+def test_write_pi_web_api_data_success(mock_dataframe, api, base_input_data):
+    input_data = {
+        **base_input_data,
+        'pi_web_api_output_config': {
+            'endpoint': 'https://test-pi-server.com/piwebapi',
+            'prediction_tags': {'tag1': 'web_id_1', 'tag2': 'web_id_2'},
+            'confidence_tags': {'tag3': 'web_id_3', 'tag4': 'web_id_4'},
+        },
+    }
+
+    mock_dataframe.return_value = _create_mock_dataframe()
+
+    # Mock successful responses
+    api.pi_web_api_client.write_value.side_effect = [
+        [{'WebId': 'web_id_1', 'Errors': []}, {'WebId': 'web_id_2', 'Errors': []}],
+        [{'WebId': 'web_id_3', 'Errors': []}, {'WebId': 'web_id_4', 'Errors': []}],
+    ]
+
+    result = api.write_pi_web_api_data(input_data)
+
+    api.pi_web_api_client.write_value.assert_has_calls(
+        [
+            call(
+                web_ids=['web_id_1', 'web_id_2'],
+                value={
+                    'Timestamp': '2024-01-01T00:00:00+00:00',
+                    'Value': 0.75,
+                },
+                metadata=metadata['metadata'],
+            ),
+            call(
+                web_ids=['web_id_3', 'web_id_4'],
+                value={
+                    'Timestamp': '2024-01-01T00:00:00+00:00',
+                    'Value': 0.95,
+                },
+                metadata=metadata['metadata'],
+            ),
+        ]
+    )
+
+    assert result == {
+        'prediction': [0.75],
+        'prediction_confidence': [0.95],
+        'timestamp': ['2024-01-01T00:00:00+00:00'],
+    }
+
+
+@patch('laborious.activities.api.DataFrame')
+def test_write_pi_web_api_data_prediction_error(mock_dataframe, api, base_input_data):
+    mock_dataframe.return_value = _create_mock_dataframe(
+        {
+            'prediction': [0.75],
+            'prediction_confidence': [PI_WEB_API_PREDICTION_ERROR_CONFIDENCE],
+            'timestamp': ['2024-01-01T00:00:00+00:00'],
+        }
+    )
+
+    api.pi_web_api_client.write_value.side_effect = Exception('Prediction write failed')
+
+    result = api.write_pi_web_api_data(base_input_data)
+
+    api.send_notification.assert_called_once_with(
+        metadata=metadata['metadata'],
+        notification_id='WRITE_PI_WEB_API_PREDICTION_ERROR',
+        message="Error writing prediction data to PI Web API: Prediction write failed\n Tags: {'tag1': 'web_id_1'}",
+        block='write_pi_web_api_data',
+        level=NotificationLevel.ERROR,
+        attachment_content=ANY,
+    )
+
+    assert result['prediction_confidence'][0] == PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+    assert api.pi_web_api_client.write_value.call_count == 1
+
+
+@patch('laborious.activities.api.DataFrame')
+def test_write_pi_web_api_data_confidence_error(mock_dataframe, api, base_input_data):
+    mock_dataframe.return_value = _create_mock_dataframe()
+
+    # First call succeeds, second fails
+    api.pi_web_api_client.write_value.side_effect = [
+        [{'WebId': 'web_id_1', 'Errors': []}],
+        Exception('Confidence write failed'),
+    ]
+
+    result = api.write_pi_web_api_data(base_input_data)
+
+    api.send_notification.assert_called_once_with(
+        metadata=metadata['metadata'],
+        notification_id='WRITE_PI_WEB_API_CONFIDENCE_ERROR',
+        message="Error writing confidence data to PI Web API: Confidence write failed\n Tags: {'tag2': 'web_id_2'}",
+        block='write_pi_web_api_data',
+        level=NotificationLevel.ERROR,
+        attachment_content=ANY,
+    )
+
+    assert result == {
+        'prediction': [0.75],
+        'prediction_confidence': [0.95],
+        'timestamp': ['2024-01-01T00:00:00+00:00'],
+    }
+    assert api.pi_web_api_client.write_value.call_count == 2
+
+
+@patch('laborious.activities.api.DataFrame')
+def test_write_pi_web_api_data_empty_tags(mock_dataframe, api, base_input_data):
+    input_data = {
+        **base_input_data,
+        'pi_web_api_output_config': {
+            'endpoint': 'https://test-pi-server.com/piwebapi',
+            'prediction_tags': {},
+            'confidence_tags': {},
+        },
+    }
+
+    mock_dataframe.return_value = _create_mock_dataframe()
+
+    # Mock empty responses
+    api.pi_web_api_client.write_value.side_effect = [
+        [],
+        [],
+    ]
+
+    result = api.write_pi_web_api_data(input_data)
+
+    api.pi_web_api_client.write_value.assert_has_calls(
+        [
+            call(
+                web_ids=[],
+                value={
+                    'Timestamp': '2024-01-01T00:00:00+00:00',
+                    'Value': 0.75,
+                },
+                metadata=metadata['metadata'],
+            ),
+            call(
+                web_ids=[],
+                value={
+                    'Timestamp': '2024-01-01T00:00:00+00:00',
+                    'Value': 0.95,
+                },
+                metadata=metadata['metadata'],
+            ),
+        ]
+    )
+
+    assert result == {
+        'prediction': [0.75],
+        'prediction_confidence': [0.95],
+        'timestamp': ['2024-01-01T00:00:00+00:00'],
+    }
+
+
+@patch('laborious.activities.api.DataFrame')
+def test_write_pi_web_api_data_updates_confidence_and_comments(
+    mock_dataframe, api, base_input_data
+):
+    mock_dataframe.return_value = _create_mock_dataframe()
+    api.pi_web_api_client.write_value.side_effect = [
+        [{'WebId': 'web_id_1', 'Errors': []}],
+        [{'WebId': 'web_id_2', 'Errors': []}],
+    ]
+    with patch.object(
+        api,
+        'process_pi_web_api_response',
+        new=MagicMock(side_effect=[(0.33, 'PI warning'), (0, '')]),
+    ) as process_mock:
+        result = api.write_pi_web_api_data(base_input_data)
+
+    assert process_mock.call_count == 2
+    assert result is not None
+
+
+@patch('laborious.activities.api.SientiaMonitoring.shutdown')
+def test_close(mock_shutdown, api):
+    api.close()
+
+    api.pi_web_api_client.close.assert_called_once()
+    mock_shutdown.assert_called_once_with(api)
+
+
+def test_process_pi_web_api_response_success(api):
+    """Test successful processing of PI Web API response with all tags written."""
+    response_data = [
+        {'WebId': 'web_id_1', 'Errors': []},
+        {'WebId': 'web_id_2', 'Errors': []},
+    ]
+    tags = {'tag1': 'web_id_1', 'tag2': 'web_id_2'}
+    core_labels = {
+        'pod_id': 'test_pod',
+        'runtime': 'local',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+    }
+
+    confidence, message = api.process_pi_web_api_response(
+        response_data=response_data,
+        tags=tags,
+        core_labels=core_labels,
+        metadata=metadata['metadata'],
+    )
+
+    assert confidence == 0
+    assert message == ''
+    assert api.emit_metric_sync.call_count == 2
+    # Verify that emit_metric_sync was called with correct tags structure
+    call_args_list = api.emit_metric_sync.call_args_list
+    assert len(call_args_list) == 2
+    # Check that all calls include core_labels and tag_name
+    for call_args in call_args_list:
+        assert 'tag_name' in call_args.kwargs['tags']
+        assert call_args.kwargs['tags']['tag_name'] in ['tag1', 'tag2']
+
+
+def test_process_pi_web_api_response_with_errors(api):
+    """Test processing response with errors in some tags."""
+    response_data = [
+        {'WebId': 'web_id_1', 'Errors': ['Error writing tag']},
+        {'WebId': 'web_id_2', 'Errors': []},
+    ]
+    tags = {'tag1': 'web_id_1', 'tag2': 'web_id_2'}
+    core_labels = {
+        'pod_id': 'test_pod',
+        'runtime': 'local',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+    }
+
+    confidence, message = api.process_pi_web_api_response(
+        response_data=response_data,
+        tags=tags,
+        core_labels=core_labels,
+        metadata=metadata['metadata'],
+    )
+
+    assert confidence == PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+    assert (
+        message
+        == "The number of written tags does not match the number of tag names: Expected ['tag1', 'tag2'] tags, but ['tag2'] tags were written."
+    )
+    assert api.emit_metric_sync.call_count == 2
+
+
+def test_process_pi_web_api_response_missing_tags(api):
+    """Test processing response when number of written tags doesn't match expected."""
+    response_data = [
+        {'WebId': 'web_id_1', 'Errors': []},
+    ]
+    tags = {'tag1': 'web_id_1', 'tag2': 'web_id_2'}
+    core_labels = {
+        'pod_id': 'test_pod',
+        'runtime': 'local',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+    }
+
+    confidence, message = api.process_pi_web_api_response(
+        response_data=response_data,
+        tags=tags,
+        core_labels=core_labels,
+        metadata=metadata['metadata'],
+    )
+
+    assert confidence == PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+    assert (
+        message
+        == "The number of written tags does not match the number of tag names: Expected ['tag1', 'tag2'] tags, but ['tag1'] tags were written."
+    )
+    api.send_notification.assert_called_once()
+    call_args = api.send_notification.call_args
+    assert call_args.kwargs['notification_id'] == 'WRITE_PI_WEB_API_PREDICTION_ERROR'
+    assert call_args.kwargs['level'] == NotificationLevel.ERROR
+
+
+def test_process_pi_web_api_response_missing_webid(api):
+    """Test processing response when WebId is missing in response item."""
+    response_data = [
+        {'Errors': []},
+        {'WebId': 'web_id_2', 'Errors': []},
+    ]
+    tags = {'tag1': 'web_id_1', 'tag2': 'web_id_2'}
+    core_labels = {
+        'pod_id': 'test_pod',
+        'runtime': 'local',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+    }
+
+    confidence, message = api.process_pi_web_api_response(
+        response_data=response_data,
+        tags=tags,
+        core_labels=core_labels,
+        metadata=metadata['metadata'],
+    )
+
+    assert confidence == PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+    assert (
+        message
+        == "The number of written tags does not match the number of tag names: Expected ['tag1', 'tag2'] tags, but ['tag2'] tags were written."
+    )
+    api.error.assert_any_call('The response did not contain some WebIds', metadata['metadata'])
+
+
+def test_process_pi_web_api_response_missing_tag_name(api):
+    """Test processing response when tag name is not found for WebId."""
+    response_data = [
+        {'WebId': 'unknown_web_id', 'Errors': []},
+    ]
+    tags = {'tag1': 'web_id_1'}
+    core_labels = {
+        'pod_id': 'test_pod',
+        'runtime': 'local',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+    }
+
+    confidence, message = api.process_pi_web_api_response(
+        response_data=response_data,
+        tags=tags,
+        core_labels=core_labels,
+        metadata=metadata['metadata'],
+    )
+
+    assert confidence == PI_WEB_API_PREDICTION_ERROR_CONFIDENCE
+    assert (
+        message
+        == "The number of written tags does not match the number of tag names: Expected ['tag1'] tags, but [] tags were written."
+    )
+    api.error.assert_any_call(
+        'The response did not contain the tag name for WebId unknown_web_id', metadata['metadata']
+    )

tests/laborious/activities/test_mlflow.py

@@ -0,0 +1,856 @@
+from datetime import datetime
+from unittest.mock import ANY, MagicMock, patch
+
+import numpy as np
+import pandas as pd
+from pytest import fixture, raises
+from sientia_do.notifications.models import NotificationLevel
+from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+
+from laborious.activities.mlflow import MLFlow
+
+
+@fixture(autouse=True)
+def _passthrough_from_dict():
+    with patch(
+        'laborious.activities.mlflow.MinioDataFramePayload.from_dict', side_effect=lambda x: x
+    ):
+        yield
+
+
+@patch('laborious.activities.mlflow.MinioRepository')
+def test___init__(mock_minio_repository):
+    logger = MagicMock()
+    notification_handler = MagicMock()
+    metrics_controller = MagicMock()
+    mlflow_repo = MagicMock()
+    plugin_store = MagicMock()
+
+    minio_repo = mock_minio_repository(
+        endpoint='localhost:9000',
+        access_key='minio',
+        secret_key='minio123',
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+        bucket='test',
+    )
+    mlflow = MLFlow(
+        mlflow_repository=mlflow_repo,
+        plugin_store=plugin_store,
+        minio_repository=minio_repo,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+    )
+
+    assert mlflow.mlflow_repository is mlflow_repo
+    assert mlflow.plugin_store is plugin_store
+
+    mock_minio_repository.assert_called_once_with(
+        endpoint='localhost:9000',
+        access_key='minio',
+        secret_key='minio123',
+        logger=ANY,
+        notification_handler=ANY,
+        metrics_controller=ANY,
+        bucket='test',
+    )
+
+
+@fixture
+@patch('laborious.activities.mlflow.MinioRepository')
+def mlflow(mock_minio_repository):
+    logger = MagicMock()
+    notification_handler = MagicMock()
+    metrics_controller = MagicMock()
+    mlflow_repo = MagicMock()
+    plugin_store = MagicMock()
+
+    minio_repo = mock_minio_repository(
+        endpoint='localhost:9000',
+        access_key='minio',
+        secret_key='minio123',
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+        bucket='test',
+    )
+    mlflow = MLFlow(
+        mlflow_repository=mlflow_repo,
+        plugin_store=plugin_store,
+        minio_repository=minio_repo,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+    )
+
+    mlflow.minio_repository = MagicMock()
+
+    mlflow.send_notification = MagicMock()
+    mlflow.emit_metric = MagicMock()
+    mlflow.error = MagicMock()
+    mlflow.debug = MagicMock()
+    mlflow.info = MagicMock()
+    mlflow.warning = MagicMock()
+    mlflow.critical = MagicMock()
+
+    return mlflow
+
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schema_name': 'test_schedule',
+    },
+}
+
+
+def test_detect_and_parse_datetime_index_empty(mlflow):
+    df = pd.DataFrame()
+    out = mlflow._detect_and_parse_datetime_index(df, metadata['metadata'])
+    assert out.empty
+
+
+def test_detect_and_parse_datetime_index_mixed_types_error(mlflow):
+    idx = pd.Index([pd.Timestamp('2020-01-01', tz='UTC'), 'x'])
+    df = pd.DataFrame({'a': [1, 2]}, index=idx)
+    with raises(ValueError):
+        mlflow._detect_and_parse_datetime_index(df, metadata['metadata'])
+
+
+def test_detect_and_parse_datetime_index_invalid_string_error(mlflow):
+    idx = pd.Index(['bad-format'])
+    df = pd.DataFrame({'a': [1]}, index=idx)
+    with raises(ValueError):
+        mlflow._detect_and_parse_datetime_index(df, metadata['metadata'])
+
+
+def test_detect_and_parse_datetime_index_unsupported_type_error(mlflow):
+    idx = pd.Index([pd.Period('2020-01', freq='M')])
+    df = pd.DataFrame({'a': [1]}, index=idx)
+    with raises(ValueError):
+        mlflow._detect_and_parse_datetime_index(df, metadata['metadata'])
+
+
+def test_detect_and_parse_datetime_index_datetime_success(mlflow):
+    idx = pd.Index([datetime(2020, 1, 1, 0, 0, 0)])
+    df = pd.DataFrame({'a': [1]}, index=idx)
+    out = mlflow._detect_and_parse_datetime_index(df, metadata['metadata'])
+    assert out.index[0].endswith('+0000')
+
+
+def test_detect_and_parse_datetime_index_timestamp_with_tz_success(mlflow):
+    idx = pd.DatetimeIndex([pd.Timestamp('2020-01-01 00:00:00', tz='UTC')])
+    df = pd.DataFrame({'a': [1]}, index=idx)
+    out = mlflow._detect_and_parse_datetime_index(df, metadata['metadata'])
+    assert out.index[0].endswith('+0000')
+
+
+@patch(
+    'laborious.activities.mlflow.MinioDataFramePayload.from_dataframe',
+    new_callable=MagicMock,
+)
+def test_request_transform_success(mock_from_dataframe, mlflow):
+    ts = pd.Timestamp('2020-01-01', tz='UTC')
+    raw = pd.DataFrame(
+        {
+            'variable': ['v1', 'v1'],
+            'timestamp': [ts, ts],
+            'value': [1.0, 2.0],
+            'created_at': [ts, ts],
+        }
+    )
+    pivoted = raw.sort_values('created_at', ascending=False).drop_duplicates(
+        subset=['variable', 'timestamp'], keep='first'
+    )
+    pivoted = pivoted.pivot(index='timestamp', columns='variable', values='value')
+    pivoted = pivoted.fillna(np.nan)
+    pivoted.columns.name = None
+    pivoted.index.name = None
+    pivoted['timestamp'] = pivoted.index
+
+    out_idx = pd.Index([ts.strftime(DATETIME_FORMAT_WITH_TZ)], name=None)
+    out_df = pd.DataFrame({'v1': [1.0]}, index=out_idx)
+    wrapper = MagicMock()
+    wrapper.transform.return_value = (out_df, {'meta': True})
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw)
+
+    input_data = {
+        **metadata,
+        'data': payload,
+        'model_name': 'test_model',
+        'model_config': {},
+    }
+
+    response_data = mlflow.request_transform(input_data)
+
+    mlflow.mlflow_repository.get_cached_model.assert_called_once_with(
+        model_name='test_model',
+        alias='production',
+        retention_minutes=0,
+        metadata=metadata['metadata'],
+    )
+    mock_from_dataframe.assert_called_once()
+    assert response_data == mock_from_dataframe.return_value
+
+
+@patch(
+    'laborious.activities.mlflow.MinioDataFramePayload.from_dataframe',
+    new_callable=MagicMock,
+)
+def test_request_transform_success_without_transform_meta(mock_from_dataframe, mlflow):
+    ts = pd.Timestamp('2020-01-01', tz='UTC')
+    raw = pd.DataFrame(
+        {
+            'variable': ['v1'],
+            'timestamp': [ts],
+            'value': [1.0],
+            'created_at': [ts],
+        }
+    )
+    out_idx = pd.Index([ts.strftime(DATETIME_FORMAT_WITH_TZ)], name=None)
+    out_df = pd.DataFrame({'v1': [1.0]}, index=out_idx)
+    wrapper = MagicMock()
+    wrapper.transform.return_value = (out_df, {})
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw)
+
+    input_data = {
+        **metadata,
+        'data': payload,
+        'model_name': 'test_model',
+        'model_config': {},
+    }
+
+    mlflow.request_transform(input_data)
+    mock_from_dataframe.assert_called_once()
+
+
+@patch(
+    'laborious.activities.mlflow.MinioDataFramePayload.from_dataframe',
+    new_callable=MagicMock,
+)
+def test_request_transform_failure(mock_from_dataframe, mlflow):
+    mlflow.mlflow_repository.get_cached_model.side_effect = RuntimeError('boom')
+
+    data_mock = MagicMock()
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=data_mock)
+
+    input_data = {
+        **metadata,
+        'data': payload,
+        'model_name': 'test_model',
+        'model_config': {},
+    }
+
+    data_mock.sort_values.return_value = data_mock
+    data_mock.drop_duplicates.return_value = data_mock
+    data_mock.pivot.return_value = data_mock
+
+    mlflow.request_transform(input_data)
+
+    mock_from_dataframe.assert_called_once_with(
+        dataframe=None,
+        minio_repo=mlflow.minio_repository,
+        model_name='test_model',
+        operation='transform',
+        status={'success': False, 'content': ANY},
+        workflow_metadata=metadata['metadata'],
+        last_timestamp=payload.last_timestamp,
+        logger=mlflow.logger,
+    )
+
+
+@patch(
+    'laborious.activities.mlflow.MinioDataFramePayload.from_dataframe',
+    new_callable=MagicMock,
+)
+@patch('laborious.activities.mlflow.to_datetime')
+def test_request_predict(mock_to_datetime, mock_from_dataframe, mlflow):
+    wrapper = MagicMock()
+    pred_df = MagicMock()
+    wrapper.predict.return_value = (pred_df, {})
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+
+    data_mock = MagicMock()
+    data_mock.index = pd.DatetimeIndex([pd.Timestamp('2020-01-01', tz='UTC')])
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=data_mock)
+
+    input_data = {
+        **metadata,
+        'data': payload,
+        'model_name': 'test_model',
+        'model_config': {},
+    }
+
+    pred_df.columns = MagicMock()
+    pred_df.__setitem__ = MagicMock()
+
+    response_data = mlflow.request_predict(input_data)
+
+    data_mock.replace.assert_called_once_with(np.nan, None, inplace=True)
+    mock_to_datetime.assert_called()
+    mlflow.mlflow_repository.get_cached_model.assert_called_once()
+    mock_from_dataframe.assert_called_once()
+    assert response_data == mock_from_dataframe.return_value
+
+
+@patch(
+    'laborious.activities.mlflow.MinioDataFramePayload.from_dataframe',
+    new_callable=MagicMock,
+)
+@patch('laborious.activities.mlflow.to_datetime')
+def test_request_predict_success_dataframe_and_meta(mock_to_datetime, mock_from_dataframe, mlflow):
+    mock_to_datetime.side_effect = lambda x, **kwargs: x
+    wrapper = MagicMock()
+    pred_df = pd.DataFrame({'raw': [0.3]})
+    wrapper.predict.return_value = (pred_df, {'m': 1})
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+
+    data_mock = MagicMock()
+    data_mock.index = pd.DatetimeIndex([pd.Timestamp('2020-01-01', tz='UTC')])
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=data_mock)
+
+    input_data = {
+        **metadata,
+        'data': payload,
+        'model_name': 'test_model',
+        'model_config': {},
+    }
+
+    mlflow.request_predict(input_data)
+
+    assert list(pred_df.columns) == ['prediction', 'response_time']
+    mlflow.info.assert_any_call("Wrapper predict metadata: {'m': 1}", metadata['metadata'])
+    mock_from_dataframe.assert_called_once()
+
+
+@patch(
+    'laborious.activities.mlflow.MinioDataFramePayload.from_dataframe',
+    new_callable=MagicMock,
+)
+@patch('laborious.activities.mlflow.to_datetime')
+def test_request_predict_failure(mock_to_datetime, mock_from_dataframe, mlflow):
+    mlflow.mlflow_repository.get_cached_model.side_effect = RuntimeError('predict boom')
+
+    data_mock = MagicMock()
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=data_mock)
+
+    input_data = {
+        **metadata,
+        'data': payload,
+        'model_name': 'test_model',
+        'model_config': {},
+    }
+
+    mlflow.request_predict(input_data)
+
+    mock_from_dataframe.assert_called_once_with(
+        dataframe=None,
+        minio_repo=mlflow.minio_repository,
+        model_name='test_model',
+        operation='predict',
+        status={'success': False, 'content': ANY},
+        workflow_metadata=metadata['metadata'],
+        last_timestamp=payload.last_timestamp,
+        logger=mlflow.logger,
+    )
+
+
+@patch('laborious.activities.mlflow.mlflow.log_artifact')
+@patch('laborious.activities.mlflow.tempfile.mkdtemp')
+@patch('laborious.activities.mlflow.rmtree')
+@patch('laborious.activities.mlflow.to_datetime')
+def test_retrain_model_success_data_success_retrain(
+    mock_to_datetime, mock_rmtree, mock_mkdtemp, mock_log_artifact, mlflow
+):
+    mock_mkdtemp.return_value = 'tmp'
+    mock_to_datetime.side_effect = lambda idx, **kwargs: pd.DatetimeIndex(idx)
+
+    mv_alias = MagicMock()
+    mv_alias.run_id = 'source-run'
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv_alias
+
+    wrapper = MagicMock()
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+
+    mock_cm = MagicMock()
+    mock_cm.__enter__.return_value = MagicMock(run_id='new-run', experiment_id='exp-1')
+    mock_cm.__exit__.return_value = False
+    mlflow.mlflow_repository.start_run.return_value = mock_cm
+
+    ts = pd.Timestamp('2020-01-01', tz='UTC')
+    raw_data = pd.DataFrame(
+        {
+            'variable': ['target', 'f1'],
+            'timestamp': [ts, ts],
+            'value': [1.0, 2.0],
+        }
+    )
+    pivoted_index = pd.DatetimeIndex([ts])
+    wrapper.retrain.return_value = _retrain_prediction_frame(pivoted_index, 'target', [1.0])
+
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw_data)
+
+    response = mlflow.retrain_model(
+        {
+            **metadata,
+            'data': payload,
+            'model_name': 'test_model',
+            'model_config': {
+                'target': 'target',
+            },
+        }
+    )
+
+    wrapper.retrain.assert_called_once()
+    wrapper.store_model.assert_called_once_with(name='test_model')
+    assert response['success'] is True
+    assert response['experiment']['run_id'] == 'new-run'
+
+
+@patch('laborious.activities.mlflow.mlflow.log_artifact')
+@patch('laborious.activities.mlflow.tempfile.mkdtemp')
+@patch('laborious.activities.mlflow.rmtree')
+@patch('laborious.activities.mlflow.to_datetime')
+def test_retrain_model_success_with_payload_data(
+    mock_to_datetime, mock_rmtree, mock_mkdtemp, mock_log_artifact, mlflow
+):
+    mock_mkdtemp.return_value = 'tmp'
+    mock_to_datetime.side_effect = lambda idx, **kwargs: pd.DatetimeIndex(idx)
+    mv_alias = MagicMock(run_id='src')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv_alias
+    wrapper = MagicMock()
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+    mock_cm = MagicMock()
+    mock_cm.__enter__.return_value = MagicMock(run_id='r', experiment_id='e')
+    mock_cm.__exit__.return_value = False
+    mlflow.mlflow_repository.start_run.return_value = mock_cm
+
+    ts = pd.Timestamp('2020-01-01', tz='UTC')
+    raw_data = pd.DataFrame(
+        {
+            'variable': ['target', 'f1', 'target', 'f1'],
+            'timestamp': [ts, ts, ts + pd.Timedelta(hours=1), ts + pd.Timedelta(hours=1)],
+            'value': [1.0, 2.0, 3.0, 4.0],
+            'created_at': [ts, ts, ts + pd.Timedelta(hours=1), ts + pd.Timedelta(hours=1)],
+        }
+    )
+    pivoted_index = pd.DatetimeIndex([ts, ts + pd.Timedelta(hours=1)])
+    wrapper.retrain.return_value = _retrain_prediction_frame(pivoted_index, 'target', [1.0, 3.0])
+
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw_data)
+
+    response = mlflow.retrain_model(
+        {
+            **metadata,
+            'data': payload,
+            'model_name': 'test_model',
+            'model_config': {'target': 'target'},
+        }
+    )
+
+    assert response['success'] is True
+
+
+@patch('laborious.activities.mlflow.mlflow.log_artifact')
+@patch('laborious.activities.mlflow.tempfile.mkdtemp')
+@patch('laborious.activities.mlflow.rmtree')
+@patch('laborious.activities.mlflow.to_datetime')
+def test_retrain_model_always_uses_retrain_even_with_full_retrain_flag(
+    mock_to_datetime, mock_rmtree, mock_mkdtemp, mock_log_artifact, mlflow
+):
+    mock_mkdtemp.return_value = 'tmp'
+    mock_to_datetime.side_effect = lambda idx, **kwargs: pd.DatetimeIndex(idx)
+    mv_alias = MagicMock(run_id='src')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv_alias
+    wrapper = MagicMock()
+    mlflow.mlflow_repository.get_cached_model.return_value = wrapper
+    mock_cm = MagicMock()
+    mock_cm.__enter__.return_value = MagicMock(run_id='r', experiment_id='e')
+    mock_cm.__exit__.return_value = False
+    mlflow.mlflow_repository.start_run.return_value = mock_cm
+
+    ts = pd.Timestamp('2020-01-01', tz='UTC')
+    ts_next = ts + pd.Timedelta(days=1)
+    raw_data = pd.DataFrame(
+        {
+            'variable': ['target', 'f1', 'target', 'f1'],
+            'timestamp': [ts, ts, ts_next, ts_next],
+            'value': [1.0, 2.0, 3.0, 4.0],
+        }
+    )
+    pivoted_index = pd.DatetimeIndex([ts, ts_next])
+    wrapper.retrain.return_value = _retrain_prediction_frame(pivoted_index, 'target', [1.0, 3.0])
+
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw_data)
+
+    response = mlflow.retrain_model(
+        {
+            **metadata,
+            'data': payload,
+            'model_name': 'test_model',
+            'model_config': {'target': 'target', 'full_retrain': True, 'validation_fraction': 0.5},
+        }
+    )
+
+    wrapper.retrain.assert_called_once()
+    wrapper.train.assert_not_called()
+    assert response['success'] is True
+
+
+@patch('laborious.activities.mlflow.to_datetime')
+def test_retrain_model_success_data_fail_retrain(mock_to_datetime, mlflow):
+    mv_alias = MagicMock(run_id='src')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv_alias
+    mlflow.mlflow_repository.get_cached_model.side_effect = RuntimeError('retrain failed')
+
+    raw_data = MagicMock(columns=['variable', 'timestamp', 'value', 'created_at'])
+    raw_data.__getitem__.return_value.max.return_value = 'tsmax'
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw_data)
+
+    pivoted = MagicMock()
+    raw_data.sort_values.return_value = raw_data
+    raw_data.drop_duplicates.return_value = raw_data
+    raw_data.pivot.return_value = pivoted
+    pivoted.fillna = MagicMock()
+    pivoted.columns.name = None
+    pivoted.index = MagicMock()
+    pivoted.__setitem__ = MagicMock()
+
+    response = mlflow.retrain_model(
+        {
+            **metadata,
+            'data': payload,
+            'model_name': 'test_model',
+            'model_config': {'target': 'target'},
+        }
+    )
+
+    assert response['success'] is False
+    assert 'retrain failed' in response['message']
+
+
+def test_retrain_model_data_error(mlflow):
+    response = mlflow.retrain_model(
+        {
+            **metadata,
+            'model_name': 'test_model',
+            'model_config': {
+                'target': 'target',
+            },
+        }
+    )
+
+    assert response['success'] is False
+    assert 'data' in response['message'].lower() or 'loading' in response['message'].lower()
+
+
+def test_retrain_model_missing_target(mlflow):
+    ts = pd.Timestamp('2020-01-01', tz='UTC')
+    raw_data = pd.DataFrame(
+        {
+            'variable': ['f1', 'f2'],
+            'timestamp': [ts, ts],
+            'value': [1.0, 2.0],
+        }
+    )
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=raw_data)
+
+    response = mlflow.retrain_model(
+        {
+            **metadata,
+            'data': payload,
+            'model_name': 'test_model',
+            'model_config': {},
+        }
+    )
+
+    assert response['success'] is False
+    assert 'target' in response['message']
+
+
+def test_retrain_model_data_error_no_minio_repository(mlflow):
+    mlflow.minio_repository = None
+
+    with raises(ValueError) as e:
+        mlflow.retrain_model(
+            {
+                **metadata,
+                'object_key': 'test_object_key',
+                'model_name': 'test_model',
+                'model_config': {
+                    'target': 'target',
+                },
+            }
+        )
+
+    assert str(e.value) == 'Minio repository not initialized'
+
+
+def test_update_production_model(mlflow):
+    mlflow.mlflow_repository._client.search_model_versions.return_value = [
+        MagicMock(version='3', run_id='run-x'),
+        MagicMock(version='2', run_id='run-x'),
+    ]
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 1,
+        'experiment': {'run_id': 'run-x', 'experiment_id': 'e1'},
+        'timestamp': 2,
+        'status': 'success',
+    }
+
+    response = mlflow.update_production_model(input_data)
+
+    mlflow.mlflow_repository.promote_to_alias.assert_called_once_with(
+        model_name='test_model',
+        version='3',
+        alias='production',
+        metadata=metadata['metadata'],
+    )
+
+    assert response['model_name'] == 'test_model'
+    assert response['version'] == '3'
+
+
+def test_update_production_model_error(mlflow):
+    mlflow.mlflow_repository._client.search_model_versions.return_value = []
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 1,
+        'experiment': {'run_id': 'run-x', 'experiment_id': 'e1'},
+        'timestamp': 2,
+        'status': 'success',
+    }
+
+    try:
+        mlflow.update_production_model(input_data)
+    except Exception:
+        mlflow.send_notification.assert_called_once_with(
+            metadata=metadata['metadata'],
+            notification_id='UPDATE_PRODUCTION_MODEL_ERROR',
+            message=ANY,
+            block='update_production_model',
+            level=NotificationLevel.ERROR,
+            attachment_content=ANY,
+        )
+    else:
+        raise AssertionError('Expected exception')
+
+
+def _artifact_file_info(path: str) -> MagicMock:
+    file_info = MagicMock()
+    file_info.path = path
+    return file_info
+
+
+def _retrain_prediction_frame(
+    index: pd.DatetimeIndex, target: str, values: list[float]
+) -> pd.DataFrame:
+    return pd.DataFrame({target: values}, index=index)
+
+
+@patch('laborious.activities.mlflow.to_datetime')
+def test_get_reference_data_success(mock_to_datetime, mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+
+    mv = MagicMock(run_id='run1')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv
+    mlflow.mlflow_repository._client.list_artifacts.return_value = [
+        _artifact_file_info('evaluation_data.csv'),
+    ]
+
+    mock_reference_data = MagicMock()
+    mock_reference_data.__getitem__.return_value = MagicMock()
+    mock_to_datetime.return_value.dt.strftime.return_value = MagicMock()
+    mock_reference_data.to_dict.return_value = [
+        {'timestamp': '2023-05-26 11:12:27', 'value': 1.0},
+    ]
+
+    with patch('laborious.activities.mlflow.pd.read_csv', return_value=mock_reference_data):
+        with patch('laborious.activities.mlflow.tempfile.mkdtemp', return_value='/t'):
+            with patch('laborious.activities.mlflow.rmtree'):
+                with patch.object(
+                    mlflow,
+                    '_find_downloaded_csv',
+                    return_value=MagicMock(),
+                ):
+                    result = mlflow.get_reference_data(input_data)
+
+    mlflow.mlflow_repository.download_artifacts.assert_called_once_with(
+        run_id='run1',
+        artifact_path='evaluation_data.csv',
+        dst_path='/t',
+        metadata=metadata['metadata'],
+    )
+    mock_reference_data.to_dict.assert_called_once_with(orient='records')
+    assert result == mock_reference_data.to_dict.return_value
+
+
+def test_get_reference_data_not_found(mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+
+    mlflow.mlflow_repository._client.get_model_version_by_alias.side_effect = Exception('missing')
+
+    result = mlflow.get_reference_data(input_data)
+
+    mlflow.warning.assert_called()
+    assert result is None
+
+
+def test_get_reference_data_only_test_data_csv(mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+    mv = MagicMock(run_id='run1')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv
+    mlflow.mlflow_repository._client.list_artifacts.return_value = [
+        _artifact_file_info('test_data.csv'),
+    ]
+
+    mock_reference_data = MagicMock()
+    mock_reference_data.__getitem__.return_value = MagicMock()
+    mock_reference_data.to_dict.return_value = [{'timestamp': '2023-05-26 11:12:27', 'value': 1.0}]
+
+    with patch('laborious.activities.mlflow.to_datetime') as mock_to_datetime:
+        mock_to_datetime.return_value.dt.strftime.return_value = MagicMock()
+        with patch('laborious.activities.mlflow.pd.read_csv', return_value=mock_reference_data):
+            with patch('laborious.activities.mlflow.tempfile.mkdtemp', return_value='/t'):
+                with patch('laborious.activities.mlflow.rmtree'):
+                    with patch.object(
+                        mlflow,
+                        '_find_downloaded_csv',
+                        return_value=MagicMock(),
+                    ):
+                        result = mlflow.get_reference_data(input_data)
+
+    mlflow.mlflow_repository.download_artifacts.assert_called_once_with(
+        run_id='run1',
+        artifact_path='test_data.csv',
+        dst_path='/t',
+        metadata=metadata['metadata'],
+    )
+    assert result == mock_reference_data.to_dict.return_value
+
+
+def test_get_reference_data_prefers_evaluation_data_when_both_listed(mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+    mv = MagicMock(run_id='run1')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv
+    mlflow.mlflow_repository._client.list_artifacts.return_value = [
+        _artifact_file_info('test_data.csv'),
+        _artifact_file_info('evaluation_data.csv'),
+    ]
+
+    mock_reference_data = MagicMock()
+    mock_reference_data.__getitem__.return_value = MagicMock()
+    mock_reference_data.to_dict.return_value = [{'timestamp': '2023-05-26 11:12:27', 'value': 1.0}]
+
+    with patch('laborious.activities.mlflow.to_datetime') as mock_to_datetime:
+        mock_to_datetime.return_value.dt.strftime.return_value = MagicMock()
+        with patch('laborious.activities.mlflow.pd.read_csv', return_value=mock_reference_data):
+            with patch('laborious.activities.mlflow.tempfile.mkdtemp', return_value='/t'):
+                with patch('laborious.activities.mlflow.rmtree'):
+                    with patch.object(
+                        mlflow,
+                        '_find_downloaded_csv',
+                        return_value=MagicMock(),
+                    ):
+                        result = mlflow.get_reference_data(input_data)
+
+    mlflow.mlflow_repository.download_artifacts.assert_called_once_with(
+        run_id='run1',
+        artifact_path='evaluation_data.csv',
+        dst_path='/t',
+        metadata=metadata['metadata'],
+    )
+    assert result == mock_reference_data.to_dict.return_value
+
+
+def test_get_reference_data_no_candidate_artifacts_returns_none(mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+    mv = MagicMock(run_id='run1')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv
+    mlflow.mlflow_repository._client.list_artifacts.return_value = [
+        _artifact_file_info('other_artifact.csv'),
+    ]
+
+    result = mlflow.get_reference_data(input_data)
+
+    mlflow.mlflow_repository.download_artifacts.assert_not_called()
+    mlflow.warning.assert_called()
+    assert result is None
+
+
+def test_get_reference_data_missing_csv_file_returns_none(mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+    mv = MagicMock(run_id='run1')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv
+    mlflow.mlflow_repository._client.list_artifacts.return_value = [
+        _artifact_file_info('evaluation_data.csv'),
+    ]
+
+    with patch('laborious.activities.mlflow.tempfile.mkdtemp', return_value='tmp'):
+        with patch('laborious.activities.mlflow.rmtree'):
+            with patch.object(mlflow, '_find_downloaded_csv', return_value=None):
+                result = mlflow.get_reference_data(input_data)
+
+    mlflow.mlflow_repository.download_artifacts.assert_called_once()
+    mlflow.warning.assert_called()
+    assert result is None
+
+
+def test_get_reference_data_exception(mlflow):
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+    }
+
+    mv = MagicMock(run_id='run1')
+    mlflow.mlflow_repository._client.get_model_version_by_alias.return_value = mv
+    mlflow.mlflow_repository._client.list_artifacts.return_value = [
+        _artifact_file_info('evaluation_data.csv'),
+    ]
+    mlflow.mlflow_repository.download_artifacts.side_effect = Exception('dl fail')
+
+    result = mlflow.get_reference_data(input_data)
+
+    assert result is None

tests/laborious/activities/test_model_metrics.py

@@ -0,0 +1,954 @@
+from unittest.mock import ANY, MagicMock, patch
+
+from pandas import DataFrame, Timestamp
+from pytest import fixture, raises
+from sientia_do.notifications.models import NotificationLevel
+from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+from sientia_model.analytics.drift_analysis import DriftInsufficientDataError
+
+from laborious.activities.model_metrics import ModelMetrics
+
+
+@fixture
+def model_metrics_activity():
+    model_metrics = ModelMetrics(
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+    model_metrics.error = MagicMock()
+    model_metrics.debug = MagicMock()
+    model_metrics.info = MagicMock()
+    model_metrics.warning = MagicMock()
+    model_metrics.critical = MagicMock()
+    model_metrics.send_notification = MagicMock()
+    model_metrics.emit_metric_sync = MagicMock()
+    model_metrics.get_core_labels = MagicMock(
+        return_value={
+            'pod_id': 'test_pod',
+            'model_name': 'test_model',
+            'workflow_name': 'test_workflow',
+        }
+    )
+    model_metrics.observe_lag_sync = MagicMock()
+    model_metrics.pod_id = 'test_pod'
+    return model_metrics
+
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schema_name': 'test_schedule',
+    },
+}
+
+
+def test_calculate_drift_invalid_chunk_period(model_metrics_activity):
+    # Arrange
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': None,
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'invalid',
+    }
+
+    # Act & Assert
+    try:
+        model_metrics_activity.calculate_drift(input_data)
+    except ValueError as e:
+        assert str(e) == 'Invalid chunk period: invalid, must be "min" or "s"'
+        model_metrics_activity.error.assert_called_once_with(
+            'Invalid chunk period: invalid', metadata['metadata']
+        )
+    else:
+        raise AssertionError('Expected ValueError')
+
+
+def _sample_drift_metrics_df(ts: Timestamp) -> DataFrame:
+    """Minimal analyzer-shaped dataframe (univariate row + columns the activity expects)."""
+    return DataFrame(
+        {
+            'timestamp': [ts],
+            'feature': ['feature1'],
+            'method': ['ks_test'],
+            'value': [0.5],
+            'alert': [False],
+            'chunk_index': [0],
+            'chunk_start_date': [ts],
+            'chunk_end_date': [ts],
+            'threshold': [0.1],
+            'drift_type': ['univariate'],
+        }
+    )
+
+
+def test_calculate_drift_with_reference_data(model_metrics_activity):
+    ts = Timestamp('2023-05-26 11:12:27')
+    drift_df = _sample_drift_metrics_df(ts)
+    model_metrics_activity.get_drift_metrics = MagicMock(return_value=drift_df)
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict('list'),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'min',
+    }
+
+    result = model_metrics_activity.calculate_drift(input_data)
+
+    expected_timestamp = ts.tz_localize('UTC').strftime(DATETIME_FORMAT_WITH_TZ)
+    assert result == [
+        {
+            'timestamp': expected_timestamp,
+            'feature': 'feature1',
+            'method': 'ks_test',
+            'value': 0.5,
+            'alert': False,
+            'chunk_index': 0,
+            'chunk_start_date': ts.isoformat(),
+            'chunk_end_date': ts.isoformat(),
+            'model_id': 'test_model_id',
+            'accurate': True,
+        }
+    ]
+    model_metrics_activity.info.assert_called()
+    model_metrics_activity.get_drift_metrics.assert_called_once()
+
+
+def test_calculate_drift_without_reference_data(model_metrics_activity):
+    # Ten rows so int(len * 0.3) >= 1 for the built-in reference slice.
+    ts_last = Timestamp('2023-05-26 11:12:36')
+    drift_df = _sample_drift_metrics_df(ts_last)
+    model_metrics_activity.get_drift_metrics = MagicMock(return_value=drift_df)
+
+    timestamps = [f'2023-05-26 11:12:{27 + i:02d}' for i in range(10)]
+    target_data_dict = {
+        'timestamp': timestamps,
+        'variable': ['feature1'] * 10,
+        'value': [float(i) for i in range(10)],
+    }
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': None,
+        'target_data': target_data_dict,
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 's',
+    }
+
+    result = model_metrics_activity.calculate_drift(input_data)
+
+    expected_timestamp = ts_last.tz_localize('UTC').strftime(DATETIME_FORMAT_WITH_TZ)
+    assert result == [
+        {
+            'timestamp': expected_timestamp,
+            'feature': 'feature1',
+            'method': 'ks_test',
+            'value': 0.5,
+            'alert': False,
+            'chunk_index': 0,
+            'chunk_start_date': ts_last.isoformat(),
+            'chunk_end_date': ts_last.isoformat(),
+            'model_id': 'test_model_id',
+            'accurate': False,
+        }
+    ]
+    model_metrics_activity.warning.assert_called()
+    model_metrics_activity.send_notification.assert_called_once_with(
+        metadata=metadata['metadata'],
+        notification_id='MODEL_METRICS_REFERENCE_DATA_WARNING',
+        message='Using 30% first rows of target data as reference data',
+        block='model_metrics',
+        level=NotificationLevel.WARNING,
+        attachment_content=ANY,
+    )
+
+
+def test_calculate_drift_empty_drift_df(model_metrics_activity):
+    """Empty analyzer merge yields no rows and no insufficient-data alert (lib owns that failure mode)."""
+    ts = Timestamp('2023-05-26 11:12:27')
+    empty_df = _sample_drift_metrics_df(ts).iloc[0:0]
+    model_metrics_activity.get_drift_metrics = MagicMock(return_value=empty_df)
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict(),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'min',
+    }
+
+    result = model_metrics_activity.calculate_drift(input_data)
+    assert result == []
+    model_metrics_activity.send_notification.assert_not_called()
+
+
+def test_calculate_drift_empty_after_timestamp_filter(model_metrics_activity):
+    """Rows dropped by target-window alignment yield an empty export list, not an insufficient-data error."""
+    drift_df = _sample_drift_metrics_df(Timestamp('2020-01-01'))
+    model_metrics_activity.get_drift_metrics = MagicMock(return_value=drift_df)
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict(),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'min',
+    }
+
+    result = model_metrics_activity.calculate_drift(input_data)
+    assert result == []
+    model_metrics_activity.send_notification.assert_not_called()
+
+
+@patch('laborious.activities.model_metrics.DataFrame')
+@patch('laborious.activities.model_metrics.to_datetime')
+def test_calculate_drift_drift_insufficient_data_error_from_lib(
+    mock_to_datetime, mock_dataframe, model_metrics_activity
+):
+    """``DriftInsufficientDataError`` maps to MODEL_METRICS_DRIFT_INSUFFICIENT_DATA, not GET error."""
+    mock_to_datetime.return_value.dt.strftime.return_value = '2023-05-26 11:12:27'
+
+    lib_msg = (
+        '[MODEL_METRICS_DRIFT_INSUFFICIENT_DATA] Drift analysis produced no time chunks '
+        "(chunk_period='min', analysis_rows=1)."
+    )
+    model_metrics_activity.get_drift_metrics = MagicMock(
+        side_effect=DriftInsufficientDataError(lib_msg, analysis_rows=1)
+    )
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    mock_target_df = MagicMock()
+    mock_target_df.pivot.return_value = mock_target_df
+    mock_target_df.index = ['2023-05-26 11:12:27']
+    mock_target_df.reset_index.return_value = mock_target_df
+    mock_target_df.dropna.return_value = mock_target_df
+    mock_target_df.__getitem__.return_value.apply.return_value = ['2023-05-26 11:12:27']
+    mock_target_df.drop.return_value.columns = ['feature1']
+    mock_dataframe.return_value = mock_target_df
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict(),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'min',
+    }
+
+    with raises(DriftInsufficientDataError, match='MODEL_METRICS_DRIFT_INSUFFICIENT_DATA'):
+        model_metrics_activity.calculate_drift(input_data)
+
+    model_metrics_activity.error.assert_called_once_with(lib_msg, metadata['metadata'])
+    model_metrics_activity.send_notification.assert_called_once_with(
+        metadata=metadata['metadata'],
+        notification_id='MODEL_METRICS_DRIFT_INSUFFICIENT_DATA',
+        message=lib_msg,
+        block='model_metrics',
+        level=NotificationLevel.ERROR,
+        attachment_content=ANY,
+    )
+
+
+def test_calculate_drift_success_min(model_metrics_activity):
+    ts = Timestamp('2023-05-26 11:12:27')
+    drift_df = _sample_drift_metrics_df(ts)
+    model_metrics_activity.get_drift_metrics = MagicMock(return_value=drift_df)
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict('list'),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'min',
+    }
+
+    result = model_metrics_activity.calculate_drift(input_data)
+
+    expected_timestamp = ts.tz_localize('UTC').strftime(DATETIME_FORMAT_WITH_TZ)
+    assert result == [
+        {
+            'timestamp': expected_timestamp,
+            'feature': 'feature1',
+            'method': 'ks_test',
+            'value': 0.5,
+            'alert': False,
+            'chunk_index': 0,
+            'chunk_start_date': ts.isoformat(),
+            'chunk_end_date': ts.isoformat(),
+            'model_id': 'test_model_id',
+            'accurate': True,
+        }
+    ]
+    model_metrics_activity.info.assert_called()
+    model_metrics_activity.get_drift_metrics.assert_called_once()
+
+
+def test_calculate_drift_success_s(model_metrics_activity):
+    ts = Timestamp('2023-05-26 11:12:27')
+    drift_df = _sample_drift_metrics_df(ts)
+    model_metrics_activity.get_drift_metrics = MagicMock(return_value=drift_df)
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict('list'),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 's',
+    }
+
+    result = model_metrics_activity.calculate_drift(input_data)
+
+    expected_timestamp = ts.tz_localize('UTC').strftime(DATETIME_FORMAT_WITH_TZ)
+    assert result == [
+        {
+            'timestamp': expected_timestamp,
+            'feature': 'feature1',
+            'method': 'ks_test',
+            'value': 0.5,
+            'alert': False,
+            'chunk_index': 0,
+            'chunk_start_date': ts.isoformat(),
+            'chunk_end_date': ts.isoformat(),
+            'model_id': 'test_model_id',
+            'accurate': True,
+        }
+    ]
+    model_metrics_activity.info.assert_called()
+    model_metrics_activity.get_drift_metrics.assert_called_once()
+
+
+@patch('laborious.activities.model_metrics.DataFrame')
+@patch('laborious.activities.model_metrics.to_datetime')
+def test_calculate_drift_get_drift_metrics_error(
+    mock_to_datetime, mock_dataframe, model_metrics_activity
+):
+    # Arrange
+    mock_to_datetime.return_value.dt.strftime.return_value = '2023-05-26 11:12:27'
+
+    model_metrics_activity.get_drift_metrics = MagicMock(
+        side_effect=Exception('Get drift metrics error')
+    )
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    mock_target_df = MagicMock()
+    mock_target_df.pivot.return_value = mock_target_df
+    mock_target_df.index = ['2023-05-26 11:12:27']
+    mock_target_df.reset_index.return_value = mock_target_df
+    mock_target_df.dropna.return_value = mock_target_df
+    mock_target_df.__getitem__.return_value.apply.return_value = ['2023-05-26 11:12:27']
+    mock_target_df.drop.return_value.columns = ['feature1']
+    mock_dataframe.return_value = mock_target_df
+
+    input_data = {
+        **metadata,
+        'model_name': 'test_model',
+        'model_id': 'test_model_id',
+        'reference_data': reference_data.to_dict(),
+        'target_data': {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'variable': ['feature1'],
+            'value': [1.0],
+        },
+        'target_name': 'target',
+        'drift_metrics': ['ks_test'],
+        'chunk_period': 'min',
+    }
+
+    # Act / Assert
+    with raises(Exception, match='Get drift metrics error'):
+        model_metrics_activity.calculate_drift(input_data)
+
+    model_metrics_activity.error.assert_called_once_with(
+        'Error getting drift metrics: Get drift metrics error', metadata['metadata']
+    )
+    model_metrics_activity.send_notification.assert_called_once_with(
+        metadata=metadata['metadata'],
+        notification_id='MODEL_METRICS_GET_DRIFT_METRICS_ERROR',
+        message='Error getting drift metrics: Get drift metrics error',
+        block='model_metrics',
+        level=NotificationLevel.ERROR,
+        attachment_content=ANY,
+    )
+
+
+@patch('laborious.activities.model_metrics.to_datetime')
+@patch('laborious.activities.model_metrics.time.time')
+@patch('laborious.activities.model_metrics.DriftAnalysis')
+@patch('laborious.activities.model_metrics.metrics')
+def test_get_drift_metrics_success(
+    mock_metrics, mock_model_analysis, mock_time, mock_to_datetime, model_metrics_activity
+):
+    # Arrange
+    mock_time.return_value = 1000.0
+
+    mock_drift_df = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'method': ['ks_test'],
+            'value': [0.5],
+            'feature': ['feature1'],
+        }
+    )
+
+    mock_model_analysis.return_value.detect_univariate_drift.return_value = MagicMock()
+    mock_model_analysis.return_value.detect_multivariate_drift.return_value = MagicMock()
+    mock_model_analysis.return_value.get_drift_metrics_dataframe.return_value = mock_drift_df
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    reference_columns = reference_data.drop(
+        columns=['target', 'timestamp'], errors='ignore'
+    ).columns
+
+    # Act
+    result = model_metrics_activity.get_drift_metrics(
+        reference_data=reference_data,
+        target_data=target_data,
+        target_name='target',
+        reference_columns=reference_columns,
+        drift_metrics=['ks_test'],
+        chunk_period='min',
+        metadata=metadata['metadata'],
+    )
+
+    # Assert
+    assert isinstance(result, DataFrame)
+    model_metrics_activity.debug.assert_called()
+    model_metrics_activity.observe_lag_sync.assert_called()
+    model_metrics_activity.emit_metric_sync.assert_called()
+
+
+@patch('laborious.activities.model_metrics.to_datetime')
+@patch('laborious.activities.model_metrics.time.time')
+@patch('laborious.activities.model_metrics.DriftAnalysis')
+@patch('laborious.activities.model_metrics.metrics')
+def test_get_drift_metrics_univariate_error(
+    mock_metrics, mock_model_analysis, mock_time, mock_to_datetime, model_metrics_activity
+):
+    # Arrange
+    mock_time.return_value = 1000.0
+
+    mock_model_analysis.return_value.detect_univariate_drift.side_effect = Exception(
+        'Univariate drift error'
+    )
+
+    reference_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27'],
+            'target': [1.0],
+            'feature1': [1.0],
+        }
+    )
+
+    reference_columns = reference_data.drop(
+        columns=['target', 'timestamp'], errors='ignore'
+    ).columns
+
+    # Act & Assert
+    try:
+        model_metrics_activity.get_drift_metrics(
+            reference_data=reference_data,
+            target_data=target_data,
+            target_name='target',
+            reference_columns=reference_columns,
+            drift_metrics=['ks_test'],
+            chunk_period='min',
+            metadata=metadata['metadata'],
+        )
+    except Exception as e:
+        assert str(e) == 'Univariate drift error'
+        model_metrics_activity.error.assert_called_once_with(
+            'Error detecting univariate drift: Univariate drift error', metadata['metadata']
+        )
+        model_metrics_activity.emit_metric_sync.assert_called_with(
+            metric_object=mock_metrics.MODEL_ANALYZE_ERROR_COUNT, tags=ANY
+        )
+    else:
+        raise AssertionError('Expected Exception')
+
+
+@patch('laborious.activities.model_metrics.to_datetime')
+@patch('laborious.activities.model_metrics.time.time')
+@patch('laborious.activities.model_metrics.DriftAnalysis')
+@patch('laborious.activities.model_metrics.metrics')
+def test_get_drift_metrics_multivariate_error(
+    mock_metrics, mock_model_analysis, mock_time, mock_to_datetime, model_metrics_activity
+):
+    mock_time.return_value = 1000.0
+
+    mock_model_analysis.return_value.detect_univariate_drift.return_value = MagicMock()
+    mock_model_analysis.return_value.detect_multivariate_drift.side_effect = Exception(
+        'Multivariate drift error'
+    )
+
+    reference_data = DataFrame(
+        {'timestamp': ['2023-05-26 11:12:27'], 'target': [1.0], 'feature1': [1.0]}
+    )
+    target_data = DataFrame(
+        {'timestamp': ['2023-05-26 11:12:27'], 'target': [1.0], 'feature1': [1.0]}
+    )
+    reference_columns = reference_data.drop(
+        columns=['target', 'timestamp'], errors='ignore'
+    ).columns
+
+    try:
+        model_metrics_activity.get_drift_metrics(
+            reference_data=reference_data,
+            target_data=target_data,
+            target_name='target',
+            reference_columns=reference_columns,
+            drift_metrics=['ks_test'],
+            chunk_period='min',
+            metadata=metadata['metadata'],
+        )
+    except Exception as e:
+        assert str(e) == 'Multivariate drift error'
+        model_metrics_activity.error.assert_called_once_with(
+            'Error detecting multivariate drift: Multivariate drift error', metadata['metadata']
+        )
+        model_metrics_activity.emit_metric_sync.assert_called_with(
+            metric_object=mock_metrics.MODEL_ANALYZE_ERROR_COUNT, tags=ANY
+        )
+    else:
+        raise AssertionError('Expected Exception')
+
+
+@patch('laborious.activities.model_metrics.to_datetime')
+@patch('laborious.activities.model_metrics.time.time')
+@patch('laborious.activities.model_metrics.DriftAnalysis')
+@patch('laborious.activities.model_metrics.metrics')
+def test_get_drift_metrics_dataframe_error(
+    mock_metrics, mock_model_analysis, mock_time, mock_to_datetime, model_metrics_activity
+):
+    mock_time.return_value = 1000.0
+
+    mock_model_analysis.return_value.detect_univariate_drift.return_value = MagicMock()
+    mock_model_analysis.return_value.detect_multivariate_drift.return_value = MagicMock()
+    mock_model_analysis.return_value.get_drift_metrics_dataframe.side_effect = Exception(
+        'Dataframe error'
+    )
+
+    reference_data = DataFrame(
+        {'timestamp': ['2023-05-26 11:12:27'], 'target': [1.0], 'feature1': [1.0]}
+    )
+    target_data = DataFrame(
+        {'timestamp': ['2023-05-26 11:12:27'], 'target': [1.0], 'feature1': [1.0]}
+    )
+    reference_columns = reference_data.drop(
+        columns=['target', 'timestamp'], errors='ignore'
+    ).columns
+
+    try:
+        model_metrics_activity.get_drift_metrics(
+            reference_data=reference_data,
+            target_data=target_data,
+            target_name='target',
+            reference_columns=reference_columns,
+            drift_metrics=['ks_test'],
+            chunk_period='min',
+            metadata=metadata['metadata'],
+        )
+    except Exception as e:
+        assert str(e) == 'Dataframe error'
+        model_metrics_activity.error.assert_called_once_with(
+            'Error building drift metrics dataframe: Dataframe error', metadata['metadata']
+        )
+        model_metrics_activity.emit_metric_sync.assert_called_with(
+            metric_object=mock_metrics.MODEL_ANALYZE_ERROR_COUNT, tags=ANY
+        )
+    else:
+        raise AssertionError('Expected Exception')
+
+
+def test_calculate_simple_metrics_success_all_metrics(model_metrics_activity):
+    # Arrange
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28', '2023-05-26 11:12:29'],
+            'target': [1.0, 2.0, 3.0],
+            'prediction': [1.1, 2.1, 2.9],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['rmse', 'mse', 'mae', 'r2'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 4
+    assert 'rmse' in result['metric'].values
+    assert 'mse' in result['metric'].values
+    assert 'mae' in result['metric'].values
+    assert 'r2' in result['metric'].values
+    assert all(model_id == 'test_model_id' for model_id in result['model_id'].values)
+    assert all(timestamp == '2023-05-26 11:12:29' for timestamp in result['timestamp'].values)
+    assert all(data_size == 3 for data_size in result['data_size'].values)
+    assert all(interval_minutes == 5 for interval_minutes in result['interval_minutes'].values)
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['rmse', 'mse', 'mae', 'r2']",
+        metadata['metadata'],
+    )
+    model_metrics_activity.debug.assert_called_once()
+
+
+def test_calculate_simple_metrics_success_rmse_only(model_metrics_activity):
+    # Arrange
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28'],
+            'target': [1.0, 2.0],
+            'prediction': [1.1, 2.1],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['rmse'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 1
+    assert result['metric'].values[0] == 'rmse'
+    assert result['model_id'].values[0] == 'test_model_id'
+    assert result['timestamp'].values[0] == '2023-05-26 11:12:28'
+    assert result['data_size'].values[0] == 2
+    assert result['interval_minutes'].values[0] == 5
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['rmse']", metadata['metadata']
+    )
+
+
+def test_calculate_simple_metrics_success_mse_only(model_metrics_activity):
+    # Arrange
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28'],
+            'target': [1.0, 2.0],
+            'prediction': [1.1, 2.1],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['mse'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 1
+    assert result['metric'].values[0] == 'mse'
+    assert result['model_id'].values[0] == 'test_model_id'
+    assert result['timestamp'].values[0] == '2023-05-26 11:12:28'
+    assert result['data_size'].values[0] == 2
+    assert result['interval_minutes'].values[0] == 5
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['mse']", metadata['metadata']
+    )
+
+
+def test_calculate_simple_metrics_success_mae_only(model_metrics_activity):
+    # Arrange
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28'],
+            'target': [1.0, 2.0],
+            'prediction': [1.1, 2.1],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['mae'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 1
+    assert result['metric'].values[0] == 'mae'
+    assert result['model_id'].values[0] == 'test_model_id'
+    assert result['timestamp'].values[0] == '2023-05-26 11:12:28'
+    assert result['data_size'].values[0] == 2
+    assert result['interval_minutes'].values[0] == 5
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['mae']", metadata['metadata']
+    )
+
+
+def test_calculate_simple_metrics_success_r2_only(model_metrics_activity):
+    # Arrange
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28'],
+            'target': [1.0, 2.0],
+            'prediction': [1.1, 2.1],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['r2'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 1
+    assert result['metric'].values[0] == 'r2'
+    assert result['model_id'].values[0] == 'test_model_id'
+    assert result['timestamp'].values[0] == '2023-05-26 11:12:28'
+    assert result['data_size'].values[0] == 2
+    assert result['interval_minutes'].values[0] == 5
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['r2']", metadata['metadata']
+    )
+
+
+def test_calculate_simple_metrics_r2_zero_ss_tot(model_metrics_activity):
+    # Arrange
+    # All target values are the same, so ss_tot will be 0
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28'],
+            'target': [1.0, 1.0],
+            'prediction': [1.1, 1.1],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['r2'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 1
+    assert result['metric'].values[0] == 'r2'
+    assert result['value'].values[0] == 0.0  # Should return 0.0 when ss_tot == 0
+    assert result['model_id'].values[0] == 'test_model_id'
+    assert result['timestamp'].values[0] == '2023-05-26 11:12:28'
+    assert result['data_size'].values[0] == 2
+    assert result['interval_minutes'].values[0] == 5
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['r2']", metadata['metadata']
+    )
+
+
+def test_calculate_simple_metrics_success_multiple_metrics_subset(model_metrics_activity):
+    # Arrange
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28', '2023-05-26 11:12:29'],
+            'target': [1.0, 2.0, 3.0],
+            'prediction': [1.1, 2.1, 2.9],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['rmse', 'mae'],
+        'interval_minutes': 5,
+    }
+
+    # Act
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    # Assert
+    assert len(result['metric']) == 2
+    assert 'rmse' in result['metric'].values
+    assert 'mae' in result['metric'].values
+    assert all(model_id == 'test_model_id' for model_id in result['model_id'].values)
+    assert all(timestamp == '2023-05-26 11:12:29' for timestamp in result['timestamp'].values)
+    assert all(data_size == 3 for data_size in result['data_size'].values)
+    assert all(interval_minutes == 5 for interval_minutes in result['interval_minutes'].values)
+    model_metrics_activity.info.assert_called_once_with(
+        "Calculating simple metrics for model test_model_id: ['rmse', 'mae']", metadata['metadata']
+    )
+
+
+def test_calculate_simple_metrics_unknown_metric_ignored(model_metrics_activity):
+    target_data = DataFrame(
+        {
+            'timestamp': ['2023-05-26 11:12:27', '2023-05-26 11:12:28'],
+            'target': [1.0, 2.0],
+            'prediction': [1.1, 2.1],
+        }
+    )
+
+    input_data = {
+        **metadata,
+        'model_id': 'test_model_id',
+        'target_data': target_data.to_dict(),
+        'metrics': ['unknown_metric', 'rmse'],
+        'interval_minutes': 5,
+    }
+
+    result = DataFrame(model_metrics_activity.calculate_simple_metrics(input_data))
+
+    assert len(result['metric']) == 1
+    assert result['metric'].values[0] == 'rmse'

tests/laborious/activities/test_opc.py

@@ -0,0 +1,720 @@
+from unittest.mock import ANY, MagicMock, call, patch
+
+import pytest
+from pandas import DataFrame
+from pytest import mark
+from sientia_do.notifications.models import NotificationLevel
+
+from laborious.activities.opc import (
+    OPC,
+    OPC_COMMENT_SEPARATOR,
+    OPC_RECONNECT_IN_PROGRESS_COMMENT,
+    OPC_SESSION_BAD_COMMENT_PREFIX,
+    OPC_SESSION_BAD_CONFIDENCE,
+    OPC_WRITTING_ERROR_CONFIDENCE,
+    OPC_WRITTING_ERROR_MESSAGE,
+    _apply_opc_write_error,
+)
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schema_name': 'test_schedule',
+    },
+}
+
+
+def test__init__():
+    servers = {'server1': {'id': 'server1'}}
+    opc = OPC(
+        opc_servers=servers,
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+
+    assert opc.opc_servers == servers
+    assert opc.opc_repository == {}
+
+
+@patch('laborious.activities.opc.OpcRepository')
+@patch('laborious.activities.opc.OPC.send_notification')
+def test_init_opc(mock_send_notification, mock_opc_repository):
+    mock_logger = MagicMock()
+    mock_metrics_controller = MagicMock()
+    server1 = MagicMock(
+        connect=MagicMock(return_value=(True, {})), write_data=MagicMock(return_value=(True, {}))
+    )
+    server2 = MagicMock(
+        connect=MagicMock(return_value=(True, {})), write_data=MagicMock(return_value=(True, {}))
+    )
+    server3 = MagicMock(
+        connect=MagicMock(
+            return_value=(
+                False,
+                {
+                    'notification_id': 'OPC_CONNECTION_ERROR_server3',
+                    'message': 'Failed to connect to OPC server: Test error',
+                    'block': 'opc_repository',
+                    'level': NotificationLevel.ERROR,
+                    'attachment_content': 'Test error',
+                },
+            )
+        ),
+        write_data=MagicMock(return_value=(True, {})),
+    )
+    mock_opc_repository.side_effect = [server1, server2, server3]
+    mock_notification_handler = MagicMock()
+    servers = {
+        'server1': {
+            'server_name': 'server1',
+            'id': 'server1',
+            'url': 'http://localhost:8080',
+            'server_uri': 'opc.tcp://localhost:4840',
+            'cert_path': '',
+            'private_key_path': '',
+            'server_cert_path': '',
+            'reconnection_interval': 60,
+        },
+        'server2': {
+            'server_name': 'server2',
+            'id': 'server2',
+            'url': 'http://localhost:8080',
+            'server_uri': 'opc.tcp://localhost:4840',
+            'cert_path': '',
+            'private_key_path': '',
+            'server_cert_path': '',
+            'reconnection_interval': 60,
+        },
+        'server3': {
+            'server_name': 'server3',
+            'id': 'server3',
+            'url': 'http://localhost:8080',
+            'server_uri': 'opc.tcp://localhost:4840',
+            'cert_path': '',
+            'private_key_path': '',
+            'server_cert_path': '',
+            'reconnection_interval': 60,
+        },
+    }
+    opc = OPC(
+        opc_servers=servers,
+        logger=mock_logger,
+        notification_handler=mock_notification_handler,
+        metrics_controller=mock_metrics_controller,
+    )
+    opc.init_opc()
+
+    assert opc.opc_servers == servers
+    assert opc.logger == mock_logger
+    assert opc.notification_handler == mock_notification_handler
+    assert opc.opc_repository['server1'] == server1
+    assert opc.opc_repository['server2'] == server2
+
+    mock_opc_repository.assert_has_calls(
+        [
+            call(
+                opc_id='server1',
+                server_name='server1',
+                url='http://localhost:8080',
+                logger=mock_logger,
+                server_uri='opc.tcp://localhost:4840',
+                cert_path='',
+                private_key_path='',
+                server_cert_path='',
+                notification_handler=mock_notification_handler,
+                reconnection_interval=60,
+                metrics_controller=mock_metrics_controller,
+            ),
+        ]
+    )
+    mock_opc_repository.assert_has_calls(
+        [
+            call(
+                opc_id='server2',
+                server_name='server2',
+                url='http://localhost:8080',
+                logger=mock_logger,
+                server_uri='opc.tcp://localhost:4840',
+                cert_path='',
+                private_key_path='',
+                server_cert_path='',
+                notification_handler=mock_notification_handler,
+                reconnection_interval=60,
+                metrics_controller=mock_metrics_controller,
+            )
+        ]
+    )
+
+    server1.connect.assert_called_once()
+    server2.connect.assert_called_once()
+
+    mock_send_notification.assert_has_calls(
+        [
+            call(
+                metadata={
+                    'model_id': '-',
+                    'model_name': '-',
+                    'workflow_name': '-',
+                    'schedule_name': 'INITIALIZATION',
+                },
+                notification_id='OPC_CONNECTION_ERROR_server3',
+                message='Failed to connect to OPC server: Test error',
+                block='opc_repository',
+                level=NotificationLevel.ERROR,
+                attachment_content=ANY,
+            )
+        ]
+    )
+
+
+@pytest.fixture
+@patch('laborious.activities.opc.OpcRepository')
+def opc(mock_opc_repository):
+    servers = {
+        'server1': {
+            'id': 'server1',
+            'server_name': 'server1',
+            'url': 'http://localhost:8080',
+            'server_uri': 'opc.tcp://localhost:4840',
+            'cert_path': '',
+            'private_key_path': '',
+            'server_cert_path': '',
+            'reconnection_interval': 60,
+        }
+    }
+
+    mock_opc_repository.return_value.write_data = MagicMock(return_value=(True, {}))
+    mock_opc_repository.return_value.connect = MagicMock(return_value=(True, {}))
+    opc = OPC(
+        opc_servers=servers,
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+    opc.init_opc()
+    opc.send_notification = MagicMock()
+    opc.emit_metric_sync = MagicMock()
+    return opc
+
+
+WRITE_DATA_CASES = [
+    ('tag1', 'int', 50),
+    ('tag2', 'float', 50.5),
+    ('tag3', 'bool', True),
+    ('tag4', 'string', 'test'),
+]
+
+
+@mark.parametrize('tag,data_type,data', WRITE_DATA_CASES)
+def test_write_data_success(opc, tag, data_type, data):
+    opc.opc_repository['server1'].write_data.return_value = (True, {'response_time': 0.1})
+
+    response_time, error_info = opc.write_data(
+        server_id='server1',
+        tag=tag,
+        data=data,
+        data_type=data_type,
+        tag_type='prediction',
+        metadata=metadata,
+    )
+    assert response_time == 0.1
+    assert error_info is None
+    opc.opc_repository['server1'].write_data.assert_called_once_with(tag, data, data_type, metadata)
+
+
+def test_write_data_failed(opc):
+    opc.opc_repository['server1'].write_data.return_value = (
+        False,
+        {
+            'notification_id': 'OPC_WRITE_DATA_ERROR_server1',
+            'message': 'Failed to write data to OPC server: Test error',
+            'block': 'opc_repository',
+            'level': NotificationLevel.ERROR,
+            'attachment_content': 'Test error',
+        },
+    )
+
+    response_time, error_info = opc.write_data(
+        server_id='server1',
+        tag='tag1',
+        data=50,
+        data_type='int',
+        tag_type='prediction',
+        metadata=metadata,
+    )
+    assert response_time is None
+    assert error_info is not None
+
+    opc.send_notification.assert_called_once_with(
+        metadata=metadata,
+        notification_id='OPC_WRITE_DATA_ERROR_server1',
+        message='Failed to write data to OPC server: Test error',
+        block='opc_repository',
+        level=NotificationLevel.ERROR,
+        attachment_content=ANY,
+    )
+
+
+def test_write_data_exception(opc):
+    opc.opc_repository['server1'].write_data.side_effect = Exception('Test error')
+
+    try:
+        opc.write_data(
+            server_id='server1',
+            tag='tag1',
+            data=50,
+            data_type='int',
+            tag_type='prediction',
+            metadata=metadata,
+        )
+
+    except Exception:
+        opc.send_notification.assert_called_once_with(
+            metadata=metadata,
+            notification_id='WRITE_OPC_PREDICTION_ERROR',
+            message='Error writing data to OPC server: Test error',
+            block='write_opc_data',
+            level=NotificationLevel.ERROR,
+            attachment_content=ANY,
+        )
+
+    else:
+        raise AssertionError('Expected an exception to be raised')
+
+
+@mark.parametrize(
+    'error_info,initial_seen,initial_status,initial_reconnect,expected',
+    [
+        (None, False, None, False, (False, None, False)),
+        ({}, False, None, False, (False, None, False)),
+        (
+            {'opc_error_kind': 'session_bad', 'opc_status': 'BadSessionIdInvalid'},
+            False,
+            None,
+            False,
+            (True, 'BadSessionIdInvalid', False),
+        ),
+        (
+            {'opc_error_kind': 'session_bad', 'opc_status': 'NewStatus'},
+            True,
+            'OldStatus',
+            False,
+            (True, 'NewStatus', False),
+        ),
+        (
+            {'opc_error_kind': 'session_bad'},
+            True,
+            'KeptStatus',
+            False,
+            (True, 'KeptStatus', False),
+        ),
+        (
+            {'opc_error_kind': 'reconnect_in_progress'},
+            False,
+            None,
+            False,
+            (False, None, True),
+        ),
+        (
+            {'opc_error_kind': 'other'},
+            True,
+            'Status',
+            True,
+            (True, 'Status', True),
+        ),
+    ],
+)
+def test_apply_opc_write_error(
+    error_info, initial_seen, initial_status, initial_reconnect, expected
+):
+    result = _apply_opc_write_error(
+        error_info,
+        initial_seen,
+        initial_status,
+        initial_reconnect,
+    )
+    assert result == expected
+
+
+def test_write_tags_from_config_prediction_success(opc):
+    opc.write_data = MagicMock(return_value=(0.1, None))
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+    tags_config = {'tag1': {'data_type': 'float'}}
+
+    response_times, session_bad, opc_status, reconnect = opc._write_tags_from_config(
+        server_id='server1',
+        tags_config=tags_config,
+        data=data,
+        data_column='prediction',
+        tag_type='prediction',
+        log_label='Prediction data',
+        metadata=metadata['metadata'],
+    )
+
+    assert response_times == {'tag1': 0.1}
+    assert session_bad is False
+    assert opc_status is None
+    assert reconnect is False
+    opc.write_data.assert_called_once_with(
+        server_id='server1',
+        tag='tag1',
+        data=0.75,
+        data_type='float',
+        tag_type='prediction',
+        metadata=metadata['metadata'],
+    )
+
+
+def test_write_tags_from_config_confidence_success(opc):
+    opc.write_data = MagicMock(return_value=(0.2, None))
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+    tags_config = {'tag2': {'data_type': 'float'}}
+
+    response_times, session_bad, opc_status, reconnect = opc._write_tags_from_config(
+        server_id='server1',
+        tags_config=tags_config,
+        data=data,
+        data_column='prediction_confidence',
+        tag_type='confidence',
+        log_label='Confidence data',
+        metadata=metadata['metadata'],
+    )
+
+    assert response_times == {'tag2': 0.2}
+    assert session_bad is False
+    assert opc_status is None
+    assert reconnect is False
+    opc.write_data.assert_called_once_with(
+        server_id='server1',
+        tag='tag2',
+        data=0.95,
+        data_type='float',
+        tag_type='confidence',
+        metadata=metadata['metadata'],
+    )
+
+
+def test_write_tags_from_config_write_failure(opc):
+    opc.write_data = MagicMock(return_value=(None, {}))
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+
+    response_times, session_bad, opc_status, reconnect = opc._write_tags_from_config(
+        server_id='server1',
+        tags_config={'tag1': {'data_type': 'float'}},
+        data=data,
+        data_column='prediction',
+        tag_type='prediction',
+        log_label='Prediction data',
+        metadata=metadata['metadata'],
+    )
+
+    assert response_times == {'tag1': None}
+    assert session_bad is False
+    assert opc_status is None
+    assert reconnect is False
+
+
+def test_write_tags_from_config_session_bad(opc):
+    opc.write_data = MagicMock(
+        return_value=(
+            None,
+            {
+                'opc_error_kind': 'session_bad',
+                'opc_status': 'BadSessionIdInvalid',
+            },
+        )
+    )
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+
+    response_times, session_bad, opc_status, reconnect = opc._write_tags_from_config(
+        server_id='server1',
+        tags_config={'tag1': {'data_type': 'float'}},
+        data=data,
+        data_column='prediction',
+        tag_type='prediction',
+        log_label='Prediction data',
+        metadata=metadata['metadata'],
+    )
+
+    assert response_times == {'tag1': None}
+    assert session_bad is True
+    assert opc_status == 'BadSessionIdInvalid'
+    assert reconnect is False
+
+
+def test_write_tags_from_config_reconnect_in_progress(opc):
+    opc.write_data = MagicMock(
+        return_value=(
+            None,
+            {'opc_error_kind': 'reconnect_in_progress'},
+        )
+    )
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+
+    response_times, session_bad, opc_status, reconnect = opc._write_tags_from_config(
+        server_id='server1',
+        tags_config={'tag1': {'data_type': 'float'}},
+        data=data,
+        data_column='prediction',
+        tag_type='prediction',
+        log_label='Prediction data',
+        metadata=metadata['metadata'],
+    )
+
+    assert response_times == {'tag1': None}
+    assert session_bad is False
+    assert opc_status is None
+    assert reconnect is True
+
+
+def test_manage_output_tags_success(opc):
+    opc._write_tags_from_config = MagicMock(
+        side_effect=[
+            ({'tag1': 0.1}, False, None, False),
+            ({'tag2': 0.1}, False, None, False),
+        ]
+    )
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+    config = {
+        'prediction_tags': {'tag1': {'data_type': 'float'}},
+        'confidence_tags': {'tag2': {'data_type': 'float'}},
+    }
+
+    output_data, opc_metrics, session_bad, opc_status, reconnect = opc.manage_output_tags(
+        server_id='server1',
+        config=config,
+        data=data,
+        metadata=metadata['metadata'],
+    )
+
+    assert output_data is True
+    assert opc_metrics == {'tag1': 0.1, 'tag2': 0.1}
+    assert session_bad is False
+    assert opc_status is None
+    assert reconnect is False
+    assert opc._write_tags_from_config.call_count == 2
+
+
+def test_manage_output_tags_failed(opc):
+    opc._write_tags_from_config = MagicMock(
+        side_effect=[
+            ({'tag1': 0.1}, False, None, False),
+            ({'tag2': None}, False, None, False),
+        ]
+    )
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+    config = {
+        'prediction_tags': {'tag1': {'data_type': 'float'}},
+        'confidence_tags': {'tag2': {'data_type': 'float'}},
+    }
+
+    output_data, opc_metrics, _, _, _ = opc.manage_output_tags(
+        server_id='server1',
+        config=config,
+        data=data,
+        metadata=metadata['metadata'],
+    )
+
+    assert output_data is False
+    assert opc_metrics == {'tag1': 0.1, 'tag2': None}
+
+
+def test_manage_output_tags_do_nothing(opc):
+    opc._write_tags_from_config = MagicMock()
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+    config = {'_invalid_key': {'tag1': {'data_type': 'float'}}}
+
+    output_data, opc_metrics, _, _, _ = opc.manage_output_tags(
+        server_id='server1',
+        config=config,
+        data=data,
+        metadata=metadata['metadata'],
+    )
+
+    assert output_data is True
+    assert opc_metrics == {}
+    opc._write_tags_from_config.assert_not_called()
+
+
+@patch('laborious.activities.opc.DataFrame')
+def test_write_opc_data_success(mock_dataframe, opc):
+    # Arrange
+    input_data = {
+        **metadata,
+        'data': {'prediction': [0.75], 'prediction_confidence': [0.95]},
+        'opc_output_config': {
+            'server1': {
+                'prediction_tags': {'tag1': {'data_type': 'float'}},
+                'confidence_tags': {'tag2': {'data_type': 'float'}},
+            }
+        },
+    }
+
+    # Act
+    opc.manage_output_tags = MagicMock(
+        return_value=(True, {'tag1': 0.1, 'tag2': 0.2}, False, None, False)
+    )
+
+    opc.process_confidence = MagicMock(return_value={'data': 'data'})
+    output_data, opc_metrics = opc.write_opc_data(input_data)
+
+    # Assert
+    assert output_data == {'data': 'data'}
+    assert opc_metrics == {'server1': {'tag1': 0.1, 'tag2': 0.2}}
+    opc.manage_output_tags.assert_called_once_with(
+        'server1',
+        input_data['opc_output_config']['server1'],
+        mock_dataframe.return_value,
+        metadata['metadata'],
+    )
+    opc.process_confidence.assert_called_once_with(
+        mock_dataframe.return_value,
+        True,
+        metadata['metadata'],
+        session_bad=False,
+        opc_status=None,
+        reconnect_in_progress=False,
+    )
+
+
+def test_write_opc_data_empty_config(opc):
+    # Arrange
+    input_data = {
+        **metadata,
+        'data': {'prediction': [0.75], 'prediction_confidence': [0.95]},
+        'opc_servers': ['server1'],
+        'opc_output_config': {'server1': {'prediction_tags': {}, 'confidence_tags': {}}},
+    }
+
+    # Act
+    opc.write_opc_data(input_data)
+
+    # Assert
+    opc.opc_repository['server1'].write_data.assert_not_called()
+
+
+def test_write_opc_data_no_validate_server(opc):
+    opc.validate_server = MagicMock(return_value=False)
+    input_data = {
+        **metadata,
+        'data': {'prediction': [0.75], 'prediction_confidence': [0.95]},
+        'opc_output_config': {
+            'server1': {
+                'prediction_tags': {'tag1': {'data_type': 'float'}},
+                'confidence_tags': {'tag2': {'data_type': 'float'}},
+            }
+        },
+    }
+
+    # Act
+    opc.write_opc_data(input_data)
+
+    # Assert
+    opc.opc_repository['server1'].write_data.assert_not_called()
+
+
+@mark.parametrize(
+    'data,success,expected',
+    [
+        (DataFrame({'prediction_confidence': [0]}), True, 0),
+        (DataFrame({'prediction_confidence': [0]}), False, 12),
+    ],
+)
+def test_process_confidence(opc, data, success, expected):
+    result = opc.process_confidence(data, success, metadata['metadata'])
+    assert result['prediction_confidence'][0] == expected
+
+
+def test_process_confidence_session_bad(opc):
+    data = DataFrame({'prediction_confidence': [0.9]})
+    result = opc.process_confidence(
+        data,
+        False,
+        metadata['metadata'],
+        session_bad=True,
+        opc_status='BadSessionIdInvalid',
+    )
+    assert result['prediction_confidence'][0] == OPC_SESSION_BAD_CONFIDENCE
+    assert result['comments'][0].startswith(OPC_SESSION_BAD_COMMENT_PREFIX)
+    assert 'BadSessionIdInvalid' in result['comments'][0]
+
+
+def test_process_confidence_generic_failure(opc):
+    data = DataFrame({'prediction_confidence': [0.9]})
+    result = opc.process_confidence(data, False, metadata['metadata'])
+    assert result['prediction_confidence'][0] == OPC_WRITTING_ERROR_CONFIDENCE
+    assert result['comments'][0] == OPC_WRITTING_ERROR_MESSAGE
+
+
+def test_manage_output_tags_merges_error_flags(opc):
+    opc._write_tags_from_config = MagicMock(
+        side_effect=[
+            ({'tag1': None}, True, 'BadSessionIdInvalid', False),
+            ({'tag2': 0.2}, False, None, True),
+        ]
+    )
+    data = DataFrame({'prediction': [0.75], 'prediction_confidence': [0.95]})
+    config = {
+        'prediction_tags': {'tag1': {'data_type': 'float'}},
+        'confidence_tags': {'tag2': {'data_type': 'float'}},
+    }
+
+    (
+        success,
+        metrics,
+        session_bad_seen,
+        opc_status,
+        reconnect_in_progress,
+    ) = opc.manage_output_tags('server1', config, data, metadata['metadata'])
+
+    assert success is False
+    assert session_bad_seen is True
+    assert reconnect_in_progress is True
+    assert opc_status == 'BadSessionIdInvalid'
+    assert metrics == {'tag1': None, 'tag2': 0.2}
+
+
+def test_process_confidence_reconnect_in_progress(opc):
+    data = DataFrame({'prediction_confidence': [0.9]})
+    result = opc.process_confidence(
+        data,
+        False,
+        metadata['metadata'],
+        reconnect_in_progress=True,
+    )
+    assert result['prediction_confidence'][0] == OPC_SESSION_BAD_CONFIDENCE
+    assert result['comments'][0] == OPC_RECONNECT_IN_PROGRESS_COMMENT
+
+
+def test_process_confidence_concatenates_multiple_comments(opc):
+    data = DataFrame({'prediction_confidence': [0.9]})
+    session_comment = f'{OPC_SESSION_BAD_COMMENT_PREFIX} BadSessionIdInvalid'
+
+    result = opc.process_confidence(
+        data,
+        False,
+        metadata['metadata'],
+        session_bad=True,
+        opc_status='BadSessionIdInvalid',
+        reconnect_in_progress=True,
+    )
+
+    assert result['prediction_confidence'][0] == OPC_SESSION_BAD_CONFIDENCE
+    assert result['comments'][0] == OPC_COMMENT_SEPARATOR.join(
+        [session_comment, OPC_RECONNECT_IN_PROGRESS_COMMENT]
+    )
+
+
+def test_validate_server(opc):
+    assert opc.validate_server('server1', metadata) is True
+    assert opc.validate_server('server2', metadata) is False
+
+
+def test_close(opc):
+    repo = opc.opc_repository['server1']
+    repo.disconnect = MagicMock(return_value=True)
+    opc.close()
+    repo.disconnect.assert_called_once()

tests/laborious/activities/test_storage.py

@@ -0,0 +1,324 @@
+import datetime
+import os
+from unittest.mock import ANY, MagicMock, patch
+
+from pytest import fixture, raises
+from sientia_do.notifications.models import NotificationLevel
+from sientia_do.observability.sientia_monitoring import SientiaMonitoring
+from sientia_do.temporal.activities.postgres_sync import Postgres
+
+from laborious.activities.storage import Storage
+
+
+@fixture(autouse=True)
+def _passthrough_from_dict():
+    with patch(
+        'laborious.activities.storage.MinioDataFramePayload.from_dict', side_effect=lambda x: x
+    ):
+        yield
+
+
+@fixture(autouse=True)
+def _patch_monitoring_shutdown():
+    """
+    Avoid running real async SientiaMonitoring.shutdown when Storage.close runs inside tests.
+    """
+    with patch.object(SientiaMonitoring, 'shutdown') as mock_shutdown:
+        yield mock_shutdown
+
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model_id',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schedule_name': 'test_schedule',
+    }
+}
+
+
+@fixture
+@patch('laborious.activities.storage.MinioRepository')
+def storage(mock_minio_repository):
+    return Storage(
+        host='localhost',
+        port=5432,
+        user='postgres',
+        password='postgres',
+        dbname='postgres',
+        min_connections=1,
+        max_connections=10,
+        retention_hours=24,
+        minio_repository=mock_minio_repository.return_value,
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+
+
+@patch('laborious.activities.storage.MinioRepository')
+def test___init___not_hasattr(mock_minio_repository):
+    logger = MagicMock()
+    notification_handler = MagicMock()
+    metrics_controller = MagicMock()
+    minio_repo = mock_minio_repository.return_value
+    storage = Storage(
+        host='localhost',
+        port=5432,
+        user='postgres',
+        password='postgres',
+        dbname='postgres',
+        min_connections=1,
+        max_connections=10,
+        retention_hours=24,
+        minio_repository=minio_repo,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+    )
+    assert isinstance(storage, Postgres)
+
+    assert storage.minio_repository is minio_repo
+    mock_minio_repository.assert_not_called()
+
+
+@patch('laborious.activities.storage.MinioRepository')
+def test___init___none_minio_repository(mock_minio_repository, storage):
+    storage.minio_repository = None
+    logger = MagicMock()
+    notification_handler = MagicMock()
+    metrics_controller = MagicMock()
+    storage.__init__(
+        host='localhost',
+        port=5432,
+        user='postgres',
+        password='postgres',
+        dbname='postgres',
+        min_connections=1,
+        max_connections=10,
+        retention_hours=24,
+        minio_repository=None,
+        logger=logger,
+        notification_handler=notification_handler,
+        metrics_controller=metrics_controller,
+    )
+
+    assert storage.minio_repository is None
+    mock_minio_repository.assert_not_called()
+
+
+@patch('laborious.activities.storage.MinioRepository')
+def test___init___done_repository(mock_minio_repository, storage):
+    storage.__init__(
+        host='localhost',
+        port=5432,
+        user='postgres',
+        password='postgres',
+        dbname='postgres',
+        min_connections=1,
+        max_connections=10,
+        retention_hours=24,
+        minio_repository=mock_minio_repository.return_value,
+        logger=MagicMock(),
+        notification_handler=MagicMock(),
+        metrics_controller=MagicMock(),
+    )
+    mock_minio_repository.assert_not_called()
+    assert storage.minio_repository is not None
+
+
+def test_close(storage, _patch_monitoring_shutdown):
+    storage.minio_repository = MagicMock()
+
+    storage.close()
+
+    assert storage.minio_repository is None
+    _patch_monitoring_shutdown.assert_called_once_with(storage)
+
+
+def test_close_when_minio_repository_already_none(storage, _patch_monitoring_shutdown):
+    """Closing without an initialized MinIO repository skips MinIO teardown."""
+    storage.minio_repository = None
+
+    storage.close()
+
+    assert storage.minio_repository is None
+    _patch_monitoring_shutdown.assert_called_once_with(storage)
+
+
+def test_load_query_with_minio_offload_no_rows(storage):
+    storage.load_custom_query = MagicMock(return_value=None)
+    storage_result = {'success': False}
+    with patch(
+        'laborious.activities.storage.MinioDataFramePayload.from_dataframe',
+        new_callable=MagicMock,
+        return_value=storage_result,
+    ) as mock_from_dataframe:
+        result = storage.load_query_with_minio_offload(
+            {**metadata, 'query': 'SELECT 1', 'model_name': 'm', 'key_prefix': 'predictions/s'}
+        )
+    assert result == storage_result
+    mock_from_dataframe.assert_called_once()
+
+
+def test_load_query_with_minio_offload_inline(storage):
+    storage.load_custom_query = MagicMock(return_value=[{'a': 1}])
+    storage_result = {'success': True, 'data': {'a': [1]}, 'object_key': None}
+    with patch(
+        'laborious.activities.storage.MinioDataFramePayload.from_dataframe',
+        new_callable=MagicMock,
+        return_value=storage_result,
+    ) as mock_from_dataframe:
+        result = storage.load_query_with_minio_offload(
+            {
+                **metadata,
+                'query': 'SELECT 1',
+                'model_name': 'my-model',
+                'key_prefix': 'predictions/s',
+            }
+        )
+    assert result == storage_result
+    mock_from_dataframe.assert_called_once()
+
+
+def test_load_query_with_minio_offload_minio(storage):
+    storage.load_custom_query = MagicMock(return_value=[{'a': 1}])
+    storage_result = {'success': True, 'data': None, 'object_key': 'object-key'}
+    with patch(
+        'laborious.activities.storage.MinioDataFramePayload.from_dataframe',
+        new_callable=MagicMock,
+        return_value=storage_result,
+    ) as mock_from_dataframe:
+        result = storage.load_query_with_minio_offload(
+            {**metadata, 'query': 'SELECT 1', 'model_name': 'm', 'key_prefix': 'predictions/s'}
+        )
+
+    assert result == storage_result
+    mock_from_dataframe.assert_called_once()
+
+
+@patch.dict(os.environ, {'SIENTIA_MINIO_RETENTION_HOURS': '1'})
+@patch('laborious.activities.storage.now')
+def test_cleanup_minio_objects_expired(mock_now, storage):
+    mock_now.return_value = datetime.datetime(2025, 1, 10, 12, 0, 0)
+    storage.minio_repository.list_objects = MagicMock(
+        return_value=[
+            'sientia/streamlit-connectors/training_datasets/m/m-initial-2024-12-01_00-00-00.parquet',
+            'sientia/streamlit-connectors/training_datasets/m/m-initial-2025-01-10_12-00-00.parquet',
+        ]
+    )
+    storage.minio_repository.delete_file = MagicMock()
+    storage.send_notification = MagicMock()
+
+    data_mock = MagicMock()
+    data_mock.cleanup_prefix.return_value = 'training_datasets/m'
+
+    result = storage.cleanup_minio_objects_expired({**metadata, 'data': data_mock})
+
+    assert result['deleted_count'] == 1
+    assert result['failed_count'] == 0
+    deleted_key = (
+        'sientia/streamlit-connectors/training_datasets/m/m-initial-2024-12-01_00-00-00.parquet'
+    )
+    assert deleted_key in result['deleted']
+    assert result['deleted'][deleted_key]['success'] is True
+    storage.minio_repository.list_objects.assert_called_once_with(
+        prefix='training_datasets/m',
+        recursive=True,
+        metadata=metadata['metadata'],
+    )
+    storage.minio_repository.delete_file.assert_called_once_with(
+        object_name='sientia/streamlit-connectors/training_datasets/m/m-initial-2024-12-01_00-00-00.parquet',
+        metadata=metadata['metadata'],
+    )
+
+
+def test_load_query_with_minio_offload_minio_not_initialized(storage):
+    storage.minio_repository = None
+
+    with raises(ValueError, match='Minio repository not initialized'):
+        storage.load_query_with_minio_offload({**metadata, 'query': 'SELECT 1', 'model_name': 'm'})
+
+
+def test_export_payload_to_postgres(storage):
+    payload = MagicMock()
+    payload.retrieve = MagicMock(return_value=MagicMock())
+    storage.export_data_to_postgres = MagicMock(return_value={'success': True})
+
+    result = storage.export_payload_to_postgres(
+        {**metadata, 'data': payload, 'schema': 'public', 'table': 't'}
+    )
+
+    payload.retrieve.assert_called_once_with(storage.minio_repository, metadata['metadata'])
+    storage.export_data_to_postgres.assert_called_once()
+    assert result == {'success': True}
+
+
+def test_cleanup_minio_objects_expired_minio_not_initialized(storage):
+    storage.minio_repository = None
+
+    data_mock = MagicMock()
+    data_mock.cleanup_prefix.return_value = 'test'
+    with raises(ValueError, match='Minio repository not initialized'):
+        storage.cleanup_minio_objects_expired({**metadata, 'data': data_mock})
+
+
+@patch('laborious.activities.storage.now')
+def test_cleanup_minio_objects_expired_unparseable_key(mock_now, storage):
+    mock_now.return_value = datetime.datetime(2025, 1, 10, 12, 0, 0)
+    storage.minio_repository.list_objects = MagicMock(
+        return_value=['some/random/key-without-timestamp.parquet']
+    )
+    storage.minio_repository.delete_file = MagicMock()
+    storage.send_notification = MagicMock()
+
+    data_mock = MagicMock()
+    data_mock.cleanup_prefix.return_value = 'test'
+    result = storage.cleanup_minio_objects_expired({**metadata, 'data': data_mock})
+
+    assert result['deleted_count'] == 0
+    assert result['failed_count'] == 0
+    storage.minio_repository.delete_file.assert_not_called()
+
+
+@patch('laborious.activities.storage.now')
+def test_cleanup_minio_objects_expired_delete_fails(mock_now, storage):
+    mock_now.return_value = datetime.datetime(2025, 1, 10, 12, 0, 0)
+    old_key = 'training_datasets/m/m-initial-2024-12-01_00-00-00.parquet'
+    storage.minio_repository.list_objects = MagicMock(return_value=[old_key])
+    storage.minio_repository.delete_file = MagicMock(side_effect=Exception('delete error'))
+    storage.send_notification = MagicMock()
+
+    data_mock = MagicMock()
+    data_mock.cleanup_prefix.return_value = 'training_datasets/m'
+    result = storage.cleanup_minio_objects_expired({**metadata, 'data': data_mock})
+
+    assert result['deleted_count'] == 0
+    assert result['failed_count'] == 1
+    assert old_key in result['failed']
+    assert result['failed'][old_key]['success'] is False
+    assert result['failed'][old_key]['message'] == 'delete error'
+
+
+@patch('laborious.activities.storage.now')
+def test_cleanup_minio_objects_expired_list_objects_error(mock_now, storage):
+    mock_now.return_value = datetime.datetime(2025, 1, 10, 12, 0, 0)
+    storage.minio_repository.list_objects = MagicMock(side_effect=Exception('list error'))
+    storage.send_notification = MagicMock()
+    storage.error = MagicMock()
+
+    data_mock = MagicMock()
+    data_mock.cleanup_prefix.return_value = 'training_datasets/m'
+    result = storage.cleanup_minio_objects_expired({**metadata, 'data': data_mock})
+
+    assert result['deleted_count'] == 0
+    assert result['failed_count'] == 0
+    storage.send_notification.assert_called_once_with(
+        metadata=metadata['metadata'],
+        notification_id='ERROR_CLEANUP_MINIO_OBJECTS_EXPIRED',
+        message='Error cleaning up MinIO objects: list error',
+        block='cleanup_minio_objects_expired',
+        level=NotificationLevel.ERROR,
+        attachment_content=ANY,
+    )
+    storage.error.assert_called_once()

tests/laborious/utils/filters/test_conditional_filters.py

@@ -0,0 +1,44 @@
+from pandas import DataFrame
+
+from laborious.utils.filters.conditional_filters import (
+    filter_empty_data,
+    filter_specific_variables_null_values,
+)
+
+
+def test_filter_specific_variables_null_values():
+    assert (
+        filter_specific_variables_null_values(
+            DataFrame({'variable': ['variable1', 'variable2'], 'value': [1, 2]}),
+            config={'variables': ['variable2']},
+        )
+        is False
+    )
+
+
+def test_filter_specific_variables_null_values_with_empty_data():
+    assert (
+        filter_specific_variables_null_values(DataFrame(), config={'variables': ['variable2']})
+        is False
+    )
+
+
+def test_filter_specific_variables_null_values_with_null_values():
+    assert (
+        filter_specific_variables_null_values(
+            DataFrame({'variable': ['variable1', 'variable2'], 'value': [1, None]}),
+            config={'variables': ['variable2']},
+        )
+        is True
+    )
+
+
+def test_filter_empty_data():
+    assert filter_empty_data(DataFrame(), {}) is True
+
+
+def test_filter_empty_data_with_data():
+    assert (
+        filter_empty_data(DataFrame({'variable': ['variable1', 'variable2'], 'value': [1, 2]}), {})
+        is False
+    )

tests/laborious/utils/filters/test_mlflow_filters.py

@@ -0,0 +1,23 @@
+from pandas import DataFrame
+
+from laborious.utils.filters.mlflow_filters import api_error_filter, nan_values_filter
+
+
+def test_api_error_filter_invalid_response():
+    assert api_error_filter(None, {}) is True  # NOSONAR
+
+
+def test_api_error_filter_valid_response_fail():
+    assert api_error_filter({'success': False}, {}) is True
+
+
+def test_api_error_filter_valid_response_success():
+    assert api_error_filter({'success': True}, {}) is False
+
+
+def test_nan_values_filter_all_nan_values():
+    assert nan_values_filter(DataFrame({'variable': [None, None]}), {}) is True
+
+
+def test_nan_values_filter_no_nan_values():
+    assert nan_values_filter(DataFrame({'variable': [1, 2]}), {}) is False

tests/laborious/utils/models/test_minio_dataframe_payload.py

@@ -0,0 +1,278 @@
+from datetime import datetime
+from io import BytesIO
+from unittest.mock import MagicMock, patch
+
+from pandas import DataFrame
+
+from laborious.utils.models.minio_dataframe_payload import (
+    MinioDataFramePayload,
+    _build_object_key,
+)
+
+
+def test_parse_object_timestamp_hyphenated_model():
+    key = 'predictions/sched/my-long-model-initial-2024-06-15_10-30-45.parquet'
+    ts = MinioDataFramePayload.parse_object_timestamp(key)
+    assert ts == datetime(2024, 6, 15, 10, 30, 45)
+
+
+def test_parse_object_timestamp_transform():
+    key = 'p/m-transform-2024-01-02_03-04-05.parquet'
+    ts = MinioDataFramePayload.parse_object_timestamp(key)
+    assert ts == datetime(2024, 1, 2, 3, 4, 5)
+
+
+def test_parse_object_timestamp_invalid():
+    assert MinioDataFramePayload.parse_object_timestamp('bad.parquet') is None
+
+
+def test_estimate_size_bytes_returns_positive_for_nonempty_frame():
+    df = DataFrame({'a': [1, 2]})
+    size = MinioDataFramePayload.estimate_size_bytes(df)
+    assert isinstance(size, int)
+    assert size > 0
+
+
+def test_cleanup_prefix_when_offloaded_returns_object_prefix():
+    payload = MinioDataFramePayload(
+        last_timestamp='t',
+        data=None,
+        object_key='training_datasets/m/m-initial-2024-01-01_00-00-00.parquet',
+        object_prefix='training_datasets/m',
+    )
+    assert MinioDataFramePayload.cleanup_prefix(payload) == 'training_datasets/m'
+
+
+def test_cleanup_prefix_when_inline_returns_none():
+    payload = MinioDataFramePayload(last_timestamp='t', data={'x': [1]}, object_key=None)
+    assert MinioDataFramePayload.cleanup_prefix(payload) is None
+
+
+def test_has_data_true_when_object_key_set():
+    payload = MinioDataFramePayload(last_timestamp='t', data=None, object_key='k')
+    assert payload.has_data() is True
+
+
+def test_retrieve_inline_dict_as_dataframe():
+    payload = MinioDataFramePayload(last_timestamp='t', data={'a': [1, 2]})
+    minio = MagicMock()
+    out = payload.retrieve(minio, {'metadata': {}})
+    assert list(out.columns) == ['a']
+    minio.download_file.assert_not_called()
+
+
+def test_retrieve_downloads_parquet_when_offloaded():
+    source = DataFrame({'a': [1, 2]})
+    buf = BytesIO()
+    source.to_parquet(buf, engine='pyarrow', index=True)
+    file_bytes = buf.getvalue()
+
+    payload = MinioDataFramePayload(
+        last_timestamp='t',
+        data=None,
+        object_key='training_datasets/m/f.parquet',
+        object_prefix='training_datasets/m',
+    )
+    minio = MagicMock()
+    minio.download_file = MagicMock(return_value=file_bytes)
+
+    out = payload.retrieve(minio, {'metadata': {}})
+
+    minio.download_file.assert_called_once_with(
+        object_name='training_datasets/m/f.parquet',
+        metadata={'metadata': {}},
+    )
+    assert list(out.columns) == ['a']
+
+
+def test_build_object_key():
+    key, prefix = _build_object_key('my-model', 'initial', '2024-01-01_00-00-00')
+    assert key == 'prediction_datasets/my-model/my-model-initial-2024-01-01_00-00-00.parquet'
+    assert prefix == 'prediction_datasets/my-model'
+
+
+def test_build_object_key_strips_slashes():
+    key, prefix = _build_object_key(' /my-model/ ', 'transform', '2024-06-15_10-30-45')
+    assert prefix == 'prediction_datasets/my-model'
+    assert key.startswith('prediction_datasets/my-model/')
+
+
+def test_estimate_size_bytes_fallback():
+    df = DataFrame({'a': [1, 2]})
+    with patch.object(df, 'to_dict', side_effect=RuntimeError('to_dict failed')):
+        size = MinioDataFramePayload.estimate_size_bytes(df)
+    assert isinstance(size, int)
+    assert size > 0
+
+
+def test_parse_object_timestamp_bad_datetime():
+    key = 'p/m-initial-9999-99-99_99-99-99.parquet'
+    assert MinioDataFramePayload.parse_object_timestamp(key) is None
+
+
+def test_retrieve_empty_when_no_data():
+    payload = MinioDataFramePayload(last_timestamp='t', data=None, object_key=None)
+    minio = MagicMock()
+    out = payload.retrieve(minio, {})
+    assert out.empty
+    minio.download_file.assert_not_called()
+
+
+@patch('laborious.utils.models.minio_dataframe_payload.now')
+def test_from_dataframe_none(mock_now):
+    mock_now.return_value = datetime(2024, 1, 1, 0, 0, 0)
+    minio = MagicMock()
+    result = MinioDataFramePayload.from_dataframe(
+        dataframe=None,
+        minio_repo=minio,
+        model_name='m',
+        operation='initial',
+        status={'success': False, 'message': 'no data'},
+    )
+    assert result.data is None
+    assert result.status == {'success': False, 'message': 'no data'}
+    assert result.object_key is None
+
+
+@patch('laborious.utils.models.minio_dataframe_payload.now')
+def test_from_dataframe_empty(mock_now):
+    mock_now.return_value = datetime(2024, 1, 1, 0, 0, 0)
+    minio = MagicMock()
+    mock_df = MagicMock()
+    mock_df.__bool__ = MagicMock(return_value=True)
+    mock_df.empty = True
+    result = MinioDataFramePayload.from_dataframe(
+        dataframe=mock_df,
+        minio_repo=minio,
+        model_name='m',
+        operation='initial',
+    )
+    assert result.data is None
+    assert result.object_key is None
+
+
+def _mock_dataframe(data_dict, timestamp_values=None):
+    """Build a MagicMock that behaves enough like a DataFrame for from_dataframe."""
+    mock_df = MagicMock()
+    mock_df.__bool__ = MagicMock(return_value=True)
+    mock_df.empty = False
+    if timestamp_values is None:
+        timestamp_values = data_dict.get('timestamp', ['2024-01-01'])
+    ts_col = MagicMock()
+    ts_col.values.tolist.return_value = timestamp_values
+    mock_df.__getitem__ = MagicMock(return_value=ts_col)
+    mock_df.to_dict.return_value = data_dict
+    buf = BytesIO()
+    DataFrame(data_dict).to_parquet(buf, engine='pyarrow', index=True)
+    mock_df.to_parquet = MagicMock(side_effect=lambda b, **kw: b.write(buf.getvalue()))
+    return mock_df
+
+
+@patch('laborious.utils.models.minio_dataframe_payload.OFFLOAD_THRESHOLD_BYTES', 10**9)
+def test_from_dataframe_inline():
+    minio = MagicMock()
+    df = _mock_dataframe({'timestamp': ['2024-01-01'], 'value': [42]})
+    result = MinioDataFramePayload.from_dataframe(
+        dataframe=df,
+        minio_repo=minio,
+        model_name='m',
+        operation='initial',
+    )
+    assert result.data is not None
+    assert result.object_key is None
+    assert result.last_timestamp == '2024-01-01'
+
+
+@patch('laborious.utils.models.minio_dataframe_payload.OFFLOAD_THRESHOLD_BYTES', 10**9)
+def test_from_dataframe_inline_uses_provided_last_timestamp():
+    minio = MagicMock()
+    df = _mock_dataframe({'timestamp': ['2024-01-01'], 'value': [42]})
+    result = MinioDataFramePayload.from_dataframe(
+        dataframe=df,
+        minio_repo=minio,
+        model_name='m',
+        operation='initial',
+        last_timestamp='2024-01-02',
+    )
+    assert result.last_timestamp == '2024-01-02'
+
+
+@patch('laborious.utils.models.minio_dataframe_payload.now')
+@patch('laborious.utils.models.minio_dataframe_payload.OFFLOAD_THRESHOLD_BYTES', 0)
+def test_from_dataframe_offloaded(mock_now):
+    mock_now.return_value = datetime(2024, 1, 1, 0, 0, 0)
+    minio = MagicMock()
+    minio.upload_file = MagicMock(return_value={'minio_object_name': 'full/key.parquet'})
+    minio.bucket = 'test-bucket'
+
+    df = _mock_dataframe({'timestamp': ['2024-01-01'], 'value': [42]})
+    result = MinioDataFramePayload.from_dataframe(
+        dataframe=df,
+        minio_repo=minio,
+        model_name='m',
+        operation='initial',
+        workflow_metadata={'wf': 'data'},
+    )
+    assert result.data is None
+    assert result.object_key == 'full/key.parquet'
+    assert result.bucket == 'test-bucket'
+    assert result.uri == 's3://test-bucket/full/key.parquet'
+    minio.upload_file.assert_called_once()
+
+
+def test_from_dict_inline():
+    raw = {
+        'last_timestamp': '2024-01-01T00:00:00+00:00',
+        'status': None,
+        'data': {'col1': {0: 'val1'}},
+        'bucket': None,
+        'object_key': None,
+        'object_prefix': None,
+        'uri': None,
+    }
+    payload = MinioDataFramePayload.from_dict(raw)
+    assert isinstance(payload, MinioDataFramePayload)
+    assert payload.last_timestamp == '2024-01-01T00:00:00+00:00'
+    assert payload.data == {'col1': {0: 'val1'}}
+    assert payload.object_key is None
+
+
+def test_from_dict_offloaded():
+    raw = {
+        'last_timestamp': '2024-06-15T10:30:45+00:00',
+        'status': {'success': True},
+        'data': None,
+        'bucket': 'my-bucket',
+        'object_key': 'training_datasets/model/model-initial-2024-06-15_10-30-45.parquet',
+        'object_prefix': 'training_datasets/model',
+        'uri': 's3://my-bucket/training_datasets/model/model-initial-2024-06-15_10-30-45.parquet',
+    }
+    payload = MinioDataFramePayload.from_dict(raw)
+    assert isinstance(payload, MinioDataFramePayload)
+    assert payload.data is None
+    assert payload.bucket == 'my-bucket'
+    assert payload.object_key == raw['object_key']
+    assert payload.object_prefix == 'training_datasets/model'
+    assert payload.uri == raw['uri']
+    assert payload.status == {'success': True}
+
+
+def test_from_dict_minimal_keys():
+    raw = {'last_timestamp': '2024-01-01'}
+    payload = MinioDataFramePayload.from_dict(raw)
+    assert payload.last_timestamp == '2024-01-01'
+    assert payload.data is None
+    assert payload.bucket is None
+    assert payload.object_key is None
+
+
+def test_from_dict_passthrough_existing_instance():
+    original = MinioDataFramePayload(last_timestamp='2024-01-01', data={'a': 1}, bucket='b')
+    result = MinioDataFramePayload.from_dict(original)
+    assert result is original
+
+
+def test_debug_with_logger_calls_custom_debug():
+    logger = MagicMock()
+    MinioDataFramePayload._debug(logger, 'msg', {'a': 1})
+    logger.custom_debug.assert_called_once_with('msg', {'a': 1})

tests/laborious/utils/repository/test_opc_repository.py

@@ -0,0 +1,584 @@
+import concurrent.futures
+import json
+from datetime import datetime
+from unittest.mock import ANY, MagicMock, Mock, patch
+
+import pytest
+from asyncua.crypto import security_policies
+from asyncua.ua.uaerrors import BadNodeIdUnknown, BadSessionIdInvalid
+from sientia_do.notifications.models import NotificationLevel
+
+from laborious.utils.repository.opc_repository import (
+    OpcClientAlreadyExistsError,
+    OpcClientNotInitializedError,
+    OpcRepository,
+    OpcSessionAlreadyConnectedError,
+    is_reconnectable_opcua_bad,
+)
+
+
+@pytest.fixture
+def mock_logger():
+    return Mock()
+
+
+@pytest.fixture
+def opc_repository(mock_logger):
+    repository = OpcRepository(
+        opc_id='test_repo',
+        server_name='test_server',
+        url='opc.tcp://localhost:4840',
+        logger=mock_logger,
+        notification_handler=Mock(),
+        reconnection_interval=60,
+        server_uri='urn:test:server',
+        cert_path='/path/to/cert.pem',
+        private_key_path='/path/to/key.pem',
+        server_cert_path='/path/to/server_cert.pem',
+        metrics_controller=MagicMock(),
+    )
+    repository.disconnection_interval = 0.1
+    repository.send_notification = MagicMock()
+    repository.send_notification = MagicMock()
+    repository.emit_metric_sync = MagicMock()
+    repository.info = MagicMock()
+    repository.error = MagicMock()
+    repository.warning = MagicMock()
+    repository.debug = MagicMock()
+    repository._session_ready.set()
+    return repository
+
+
+@pytest.fixture
+def mock_client():
+    with patch('laborious.utils.repository.opc_repository.Client') as mock:
+        client_instance = MagicMock()
+        aio = MagicMock()
+        client_instance.aio_obj = aio
+        aio.uaclient = MagicMock()
+        aio.uaclient.protocol = MagicMock(state='closed')
+        aio.session_timeout = 600_000
+        aio.secure_channel_timeout = 600_000
+        mock.return_value = client_instance
+        yield client_instance
+
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schema_name': 'test_schedule',
+    },
+}
+
+
+def test_init(opc_repository):
+    assert opc_repository.id == 'test_repo'
+    assert opc_repository.server_name == 'test_server'
+    assert opc_repository.url == 'opc.tcp://localhost:4840'
+    assert opc_repository.server_uri == 'urn:test:server'
+    assert opc_repository.cert_path == '/path/to/cert.pem'
+    assert opc_repository.private_key_path == '/path/to/key.pem'
+    assert opc_repository.server_cert_path == '/path/to/server_cert.pem'
+    assert opc_repository.reconnection_interval == 60
+    assert opc_repository.client is None
+    assert opc_repository.last_reconnection_time is None
+
+
+def test_set_security(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    opc_repository.set_security()
+
+    mock_client.application_uri = 'urn:test:server'
+    mock_client.set_security.assert_called_once_with(
+        security_policies.SecurityPolicyBasic256,
+        '/path/to/cert.pem',
+        '/path/to/key.pem',
+        None,
+        '/path/to/server_cert.pem',
+    )
+    assert mock_client.aio_obj.secure_channel_timeout == 600_000
+    assert mock_client.aio_obj.session_timeout == 600_000
+
+
+def test_set_security_missing_certificates(opc_repository):
+    opc_repository.cert_path = None
+    opc_repository.private_key_path = None
+
+    try:
+        opc_repository.set_security()
+    except ValueError as e:
+        assert str(e) == 'Certificate and private key paths must be provided for secure connection.'
+
+
+def test_set_security_missing_client(opc_repository):
+    opc_repository.client = None
+    try:
+        opc_repository.set_security()
+    except ValueError as e:
+        assert str(e) == 'Client must be initialized before setting security'
+
+
+def test_connect_with_security(opc_repository, mock_client):
+    opc_repository._create_client = MagicMock()
+    opc_repository._open_session = MagicMock(return_value=(True, {}))
+    result = opc_repository.connect()
+
+    opc_repository._create_client.assert_called_once()
+    opc_repository._open_session.assert_called_once()
+    assert result == (True, {})
+
+
+def test_connect_without_security(opc_repository, mock_client):
+    opc_repository.cert_path = None
+    opc_repository._create_client = MagicMock()
+    opc_repository._open_session = MagicMock(return_value=(True, {}))
+    opc_repository.set_security = MagicMock()
+    result = opc_repository.connect()
+
+    opc_repository._create_client.assert_called_once()
+    opc_repository._open_session.assert_called_once()
+    opc_repository.set_security.assert_not_called()
+    assert result == (True, {})
+
+
+def test_connect_raises_when_session_already_open(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    proto = MagicMock()
+    proto.state = 'open'
+    mock_client.aio_obj.uaclient.protocol = proto
+
+    with pytest.raises(OpcSessionAlreadyConnectedError, match='disconnect'):
+        opc_repository.connect()
+
+
+def test_create_client_raises_when_client_exists(opc_repository, mock_client):
+    opc_repository.client = mock_client
+
+    with pytest.raises(OpcClientAlreadyExistsError, match='already exists'):
+        opc_repository._create_client()
+
+
+def test_open_session_success(opc_repository):
+    closed_proto = MagicMock()
+    closed_proto.state = 'closed'
+    opc_repository.client = MagicMock()
+    aio = MagicMock()
+    opc_repository.client.aio_obj = aio
+    aio.uaclient = MagicMock(protocol=closed_proto)
+    aio.session_timeout = 600_000
+    aio.secure_channel_timeout = 600_000
+
+    open_proto = MagicMock()
+    open_proto.state = 'open'
+    open_proto.authentication_token = 'tok'
+
+    def connect_side_effect():
+        aio.uaclient.protocol = open_proto
+
+    opc_repository.client.connect = MagicMock(side_effect=connect_side_effect)
+
+    result = opc_repository._open_session()
+
+    opc_repository.client.connect.assert_called_once()
+    assert result == (True, {})
+    assert opc_repository._session_ready.is_set()
+
+
+def test_open_session_raises_when_already_connected(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    proto = MagicMock()
+    proto.state = 'open'
+    mock_client.aio_obj.uaclient.protocol = proto
+
+    with pytest.raises(OpcSessionAlreadyConnectedError, match='disconnect'):
+        opc_repository._open_session()
+
+
+def test_open_session_fail(opc_repository):
+    opc_repository._disconnect_locked = MagicMock()
+    opc_repository.client = MagicMock()
+    aio = MagicMock()
+    opc_repository.client.aio_obj = aio
+    aio.uaclient = MagicMock(protocol=MagicMock(state='closed'))
+    opc_repository.client.connect = MagicMock(side_effect=Exception('Test error'))
+
+    is_connected, error_data = opc_repository._open_session()
+
+    opc_repository._disconnect_locked.assert_called_once()
+    opc_repository.client.connect.assert_called_once()
+    assert is_connected is False
+    assert error_data['notification_id'] == f'OPC_CONNECTION_ERROR_{opc_repository.id}'
+    assert error_data['message'] == 'Failed to connect to OPC server: Test error'
+    assert error_data['block'] == 'opc_repository'
+    assert error_data['level'] == NotificationLevel.ERROR
+    assert error_data['attachment_content'] is not None
+
+
+def test_open_session_raises_when_no_client(opc_repository):
+    opc_repository.client = None
+
+    with pytest.raises(OpcClientNotInitializedError, match='not initialized'):
+        opc_repository._open_session()
+
+
+def test_disconnection_fallback_success(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    mock_client.disconnect.return_value = True
+    result = opc_repository._disconnection_fallback()
+
+    mock_client.disconnect.assert_called_once()
+    assert result == []
+
+
+def test_disconnection_fallback_fail(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    mock_client.disconnect.side_effect = Exception('Test error')
+    result = opc_repository._disconnection_fallback()
+    assert result == [
+        {'attempt': 1, 'error': 'Test error', 'traceback': ANY},
+        {'attempt': 2, 'error': 'Test error', 'traceback': ANY},
+        {'attempt': 3, 'error': 'Test error', 'traceback': ANY},
+        {'attempt': 4, 'error': 'Test error', 'traceback': ANY},
+        {'attempt': 5, 'error': 'Test error', 'traceback': ANY},
+    ]
+    assert mock_client.disconnect.call_count == 5
+
+
+def test_disconnect(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    opc_repository._disconnection_fallback = MagicMock(return_value=[])
+    opc_repository.disconnect()
+
+    opc_repository._disconnection_fallback.assert_called_once()
+    assert opc_repository.client is None
+    assert opc_repository._allow_reconnect is False
+
+
+def test_disconnect_no_client(opc_repository):
+    opc_repository.client = None
+    assert opc_repository.disconnect() is None
+
+
+def test_disconnect_error(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    opc_repository._disconnection_fallback = MagicMock(
+        return_value=[{'attempt': 1, 'error': 'Test error', 'traceback': 'text'}]
+    )
+    opc_repository.disconnect()
+
+    opc_repository._disconnection_fallback.assert_called_once()
+    opc_repository.send_notification.assert_called_once_with(
+        metadata=opc_repository.metadata,
+        notification_id=f'OPC_DISCONNECTION_ERROR_{opc_repository.id}',
+        message='Failed to disconnect from OPC server in 5 attempts.',
+        block='opc_repository',
+        level=NotificationLevel.ERROR,
+        attachment_content=json.dumps(
+            [{'attempt': 1, 'error': 'Test error', 'traceback': 'text'}], indent=4
+        ),
+    )
+    assert opc_repository.client is None
+
+
+def test_validate_connection_none_client(opc_repository):
+    opc_repository.client = None
+    response = opc_repository.validate_connection()
+    assert response == (False, opc_repository._not_connected_error())
+
+
+def test_validate_connection_session_not_open(opc_repository):
+    opc_repository.client = MagicMock()
+    opc_repository.client.aio_obj.uaclient.protocol = None
+
+    response = opc_repository.validate_connection()
+
+    assert response == (False, opc_repository._not_connected_error())
+    opc_repository.error.assert_called_once()
+
+
+def test_validate_connection_success(opc_repository):
+    opc_repository.client = MagicMock()
+    proto = MagicMock()
+    proto.state = 'open'
+    opc_repository.client.aio_obj.uaclient.protocol = proto
+
+    output = opc_repository.validate_connection()
+    assert output == (True, {})
+
+
+def test_write_data_validate_connection_do_nothing(opc_repository):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = MagicMock(get_node=MagicMock())
+    mock_node = MagicMock()
+    opc_repository.client.get_node.return_value = mock_node
+
+    result = opc_repository.write_data('ns=2;s=TestNode', 42.0, 'float', metadata['metadata'])
+
+    opc_repository.validate_connection.assert_called_once()
+    opc_repository.client.get_node.assert_called_once_with('ns=2;s=TestNode')
+    assert result == (True, {'response_time': ANY})
+
+
+def test_write_data_validate_connection_failed(opc_repository):
+    opc_repository.validate_connection = MagicMock(return_value=(False, {}))
+    opc_repository.client = MagicMock()
+    opc_repository._start_reconnect = MagicMock()
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    opc_repository.validate_connection.assert_called_once()
+    opc_repository.client.get_node.assert_not_called()
+    opc_repository._start_reconnect.assert_called_once()
+    assert opc_repository._start_reconnect.call_args.args[0] == 'ProtocolClosed'
+    assert is_success is False
+    assert error_data['opc_error_kind'] == 'connection_lost'
+    assert error_data['opc_status'] == 'ProtocolClosed'
+
+
+def test_write_data_get_node_failed(opc_repository):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = MagicMock()
+    opc_repository.client.get_node = MagicMock(side_effect=Exception('Test error'))
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    opc_repository.validate_connection.assert_called_once()
+    opc_repository.client.get_node.assert_called_once_with('ns=2;s=TestNode')
+    assert is_success is False
+    assert error_data['notification_id'] == f'OPC_WRITE_GET_NODE_ERROR_{opc_repository.id}'
+    assert (
+        error_data['message']
+        == "Failed to get node from OPC server: Test error | metadata: {'model_id': 'test_model', 'model_name': 'test_model', 'workflow_name': 'test_workflow', 'schema_name': 'test_schedule'}"
+    )
+    assert error_data['block'] == 'opc_repository'
+    assert error_data['level'] == NotificationLevel.ERROR
+    assert error_data['attachment_content'] is not None
+
+
+def test_write_data_invalid_data_type(opc_repository, mock_client):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = mock_client
+    mock_node = MagicMock()
+    mock_client.get_node = MagicMock(return_value=mock_node)
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'invalid_type', metadata['metadata']
+    )
+
+    opc_repository.validate_connection.assert_called_once()
+    mock_client.get_node.assert_called_once_with('ns=2;s=TestNode')
+
+    assert is_success is False
+    assert error_data['notification_id'] == f'OPC_WRITE_DATA_TYPE_ERROR_{opc_repository.id}'
+    assert (
+        error_data['message']
+        == "Unsupported data type: invalid_type | metadata: {'model_id': 'test_model', 'model_name': 'test_model', 'workflow_name': 'test_workflow', 'schema_name': 'test_schedule'}"
+    )
+    assert error_data['block'] == 'opc_repository'
+    assert error_data['level'] == NotificationLevel.ERROR
+    assert error_data.get('attachment_content') is None
+
+
+def test_write_data(opc_repository, mock_client):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = mock_client
+    mock_node = MagicMock()
+    mock_client.get_node = MagicMock(return_value=mock_node)
+
+    result = opc_repository.write_data('ns=2;s=TestNode', 42.0, 'float', metadata['metadata'])
+
+    mock_client.get_node.assert_called_once_with('ns=2;s=TestNode')
+    mock_node.write_value.assert_called_once()
+    assert result == (True, {'response_time': ANY})
+
+
+def test_write_data_write_value_failed(opc_repository, mock_client):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = mock_client
+    mock_node = MagicMock()
+    mock_client.get_node = MagicMock(return_value=mock_node)
+    mock_node.write_value.side_effect = Exception('Test error')
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    opc_repository.validate_connection.assert_called_once()
+    mock_client.get_node.assert_called_once_with('ns=2;s=TestNode')
+    mock_node.write_value.assert_called_once()
+    assert is_success is False
+    assert error_data['notification_id'] == f'OPC_WRITE_DATA_ERROR_{opc_repository.id}'
+    assert (
+        error_data['message']
+        == "Failed to write data to OPC server: Test error | metadata: {'model_id': 'test_model', 'model_name': 'test_model', 'workflow_name': 'test_workflow', 'schema_name': 'test_schedule'}"
+    )
+    assert error_data['block'] == 'opc_repository'
+    assert error_data['level'] == NotificationLevel.ERROR
+    assert error_data['attachment_content'] is not None
+
+
+def test_is_reconnectable_opcua_bad():
+    assert is_reconnectable_opcua_bad(BadSessionIdInvalid()) is True
+    assert is_reconnectable_opcua_bad(BadNodeIdUnknown()) is False
+    assert is_reconnectable_opcua_bad(Exception('other')) is False
+
+
+def test_write_data_bad_session_id_invalid_schedules_reconnect(opc_repository, mock_client):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = mock_client
+    opc_repository._start_reconnect = MagicMock()
+    mock_node = MagicMock()
+    mock_client.get_node = MagicMock(return_value=mock_node)
+    mock_node.write_value.side_effect = BadSessionIdInvalid()
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    mock_node.write_value.assert_called_once()
+    opc_repository._start_reconnect.assert_called_once()
+    assert is_success is False
+    assert error_data['opc_error_kind'] == 'session_bad'
+    assert error_data['opc_status'] == 'BadSessionIdInvalid'
+
+
+def test_write_data_reconnect_in_progress_immediate(opc_repository):
+    opc_repository._session_ready.clear()
+    opc_repository._reconnect_thread = MagicMock()
+    opc_repository._reconnect_thread.is_alive.return_value = True
+    opc_repository.validate_connection = MagicMock()
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    opc_repository.validate_connection.assert_not_called()
+    assert is_success is False
+    assert error_data['opc_error_kind'] == 'reconnect_in_progress'
+
+
+def test_start_reconnect_skips_within_interval(opc_repository):
+    opc_repository.last_reconnection_time = datetime.now()
+    opc_repository.reconnection_interval = 3600
+
+    opc_repository._start_reconnect('BadSessionIdInvalid', 'tok')
+
+    assert opc_repository._reconnect_thread is None
+
+
+def test_write_data_protocol_closed_schedules_reconnect(opc_repository):
+    opc_repository.client = MagicMock()
+    opc_repository.client.aio_obj.uaclient.protocol = MagicMock(state='closed')
+    opc_repository._start_reconnect = MagicMock()
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    opc_repository._start_reconnect.assert_called_once()
+    assert opc_repository._start_reconnect.call_args.args[0] == 'ProtocolClosed'
+    assert is_success is False
+    assert error_data['opc_error_kind'] == 'connection_lost'
+    assert error_data['opc_status'] == 'ProtocolClosed'
+
+
+def test_write_data_protocol_closed_skips_reconnect_within_interval(opc_repository):
+    opc_repository.client = MagicMock()
+    opc_repository.client.aio_obj.uaclient.protocol = MagicMock(state='closed')
+    opc_repository.last_reconnection_time = datetime.now()
+    opc_repository.reconnection_interval = 3600
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    assert opc_repository._reconnect_thread is None
+    assert is_success is False
+    assert error_data['opc_error_kind'] == 'connection_lost'
+
+
+def test_write_data_after_failed_reconnect_schedules_again(opc_repository):
+    opc_repository._session_ready.clear()
+    opc_repository.reconnection_interval = 0
+    opc_repository.last_reconnection_time = None
+    opc_repository._reconnect_locked = MagicMock(
+        return_value=(False, {'message': 'connect failed'})
+    )
+
+    opc_repository.write_data('ns=2;s=TestNode', 42.0, 'float', metadata['metadata'])
+    if opc_repository._reconnect_thread is not None:
+        opc_repository._reconnect_thread.join(timeout=2)
+    assert opc_repository._reconnect_locked.call_count == 1
+    assert not opc_repository._reconnect_thread_in_progress()
+
+    opc_repository.write_data('ns=2;s=TestNode', 42.0, 'float', metadata['metadata'])
+    if opc_repository._reconnect_thread is not None:
+        opc_repository._reconnect_thread.join(timeout=2)
+    assert opc_repository._reconnect_locked.call_count == 2
+
+
+def test_write_data_after_disconnect_does_not_schedule_reconnect(opc_repository, mock_client):
+    opc_repository.client = mock_client
+    proto = MagicMock()
+    proto.state = 'closed'
+    mock_client.aio_obj.uaclient.protocol = proto
+    opc_repository._disconnection_fallback = MagicMock(return_value=[])
+    opc_repository.disconnect()
+
+    is_success, error_data = opc_repository.write_data(
+        'ns=2;s=TestNode', 42.0, 'float', metadata['metadata']
+    )
+
+    assert opc_repository._reconnect_thread is None
+    assert is_success is False
+    assert error_data['opc_error_kind'] == 'connection_lost'
+
+
+def test_parallel_bad_writes_single_reconnect_task(opc_repository, mock_client):
+    opc_repository.validate_connection = MagicMock(return_value=(True, {}))
+    opc_repository.client = mock_client
+    opc_repository.reconnection_interval = 0
+    opc_repository.last_reconnection_time = None
+    mock_node = MagicMock()
+    mock_client.get_node = MagicMock(return_value=mock_node)
+    mock_node.write_value.side_effect = BadSessionIdInvalid()
+    opc_repository._start_reconnect = MagicMock()
+
+    with concurrent.futures.ThreadPoolExecutor(max_workers=2) as executor:
+        futures = [
+            executor.submit(
+                opc_repository.write_data,
+                node,
+                value,
+                'float',
+                metadata['metadata'],
+            )
+            for node, value in (('ns=2;s=TestNode', 1.0), ('ns=2;s=TestNode2', 2.0))
+        ]
+        results = [future.result() for future in futures]
+
+    assert 1 <= opc_repository._start_reconnect.call_count <= 2
+    assert mock_node.write_value.call_count == 2
+    error_kinds = [r[1].get('opc_error_kind') for r in results]
+    assert error_kinds.count('session_bad') >= 1
+    assert all(k in ('session_bad', 'reconnect_in_progress') for k in error_kinds)
+
+
+@patch('laborious.utils.repository.opc_repository.datetime')
+def test_reconnect_locked_sets_last_reconnection_time(mock_datetime, opc_repository):
+    mock_datetime.now = MagicMock(return_value=datetime(2025, 1, 1, 12, 0, 0))
+    opc_repository._disconnect_locked = MagicMock()
+    opc_repository._connect_locked = MagicMock(return_value=(True, {}))
+
+    result = opc_repository._reconnect_locked()
+
+    opc_repository._disconnect_locked.assert_called_once()
+    opc_repository._connect_locked.assert_called_once()
+    assert result == (True, {})
+    assert opc_repository.last_reconnection_time == datetime(2025, 1, 1, 12, 0, 0)

tests/laborious/utils/test_connectors_config.py

@@ -0,0 +1,152 @@
+from os import environ
+
+from laborious.utils.connectors_config import (
+    build_minio_config,
+    build_mlflow_config,
+    build_opc_config,
+    build_plugin_store_config,
+)
+
+
+def test_build_mlflow_config_with_env_vars():
+    # Arrange
+    environ['MLFLOW_URL'] = 'http://test-host:8080'
+    environ['MLFLOW_USERNAME'] = 'test-user'
+    environ['MLFLOW_PASSWORD'] = 'test-pass'
+
+    # Act
+    config = build_mlflow_config()
+
+    # Assert
+    assert config['url'] == 'http://test-host:8080'
+    assert config['username'] == 'test-user'
+    assert config['password'] == 'test-pass'
+
+
+def test_build_mlflow_config_host_already_has_port():
+    environ['MLFLOW_URL'] = 'http://tracker.example.com:443'
+    environ['MLFLOW_USERNAME'] = 'u'
+    environ['MLFLOW_PASSWORD'] = 'p'
+
+    config = build_mlflow_config()
+
+    assert config['url'] == 'http://tracker.example.com:443'
+
+
+def test_build_mlflow_config_with_defaults():
+    # Arrange
+    # Clear any existing env vars
+    environ.pop('MLFLOW_URL', None)
+    environ.pop('MLFLOW_USERNAME', None)
+    environ.pop('MLFLOW_PASSWORD', None)
+
+    # Act
+    config = build_mlflow_config()
+
+    # Assert
+    assert config['url'] == 'http://localhost:5080'
+    assert config['username'] == 'aignosi'
+    assert config['password'] == 'aignosi'
+
+
+def test_build_plugin_store_config_defaults():
+    environ.pop('STORE_BASE_URL', None)
+    environ.pop('STORE_OWNER', None)
+    environ.pop('STORE_REPO', None)
+    environ.pop('STORE_BRANCH', None)
+    environ.pop('STORE_USERNAME', None)
+    environ.pop('STORE_PASSWORD', None)
+    environ.pop('STORE_CACHE_TTL_SECONDS', None)
+    environ.pop('PYPI_SERVER', None)
+    environ.pop('PYPI_USERNAME', None)
+    environ.pop('PYPI_PASSWORD', None)
+
+    cfg = build_plugin_store_config()
+    assert cfg['base_url'] == 'http://localhost:3000'
+    assert cfg['owner'] == 'sientia'
+    assert cfg['repo'] == 'model-library-store'
+    assert cfg['pypi_index_url'] == 'http://localhost:5000'
+
+
+def test_build_opc_config_with_env_vars():
+    # Arrange
+    environ['OPC_CONFIG'] = '{"opc": {"name": "test-opc", "url": "opc.tcp://test:4840"}}'
+
+    # Act
+    config = build_opc_config()
+
+    # Assert
+    assert config['opc']['name'] == 'test-opc'
+    assert config['opc']['url'] == 'opc.tcp://test:4840'
+
+
+def test_build_opc_config_with_individual_env_vars():
+    # Arrange
+    environ.pop('OPC_CONFIG', None)
+    environ['OPC_ID'] = '1'
+    environ['OPC_URL'] = 'opc.tcp://test:4840'
+    environ['OPC_SERVER_URI'] = 'opc.tcp://test:4840'
+    environ['OPC_RECONNECTION_INTERVAL'] = '300'
+
+    # Act
+    config = build_opc_config()
+
+    # Assert
+    assert config['1']['id'] == '1'
+    assert config['1']['url'] == 'opc.tcp://test:4840'
+    assert config['1']['server_uri'] == 'opc.tcp://test:4840'
+    assert config['1']['reconnection_interval'] == 300
+
+
+def test_build_opc_config_with_defaults():
+    # Arrange
+    environ.pop('OPC_CONFIG', None)
+    environ.pop('OPC_ID', None)
+    environ.pop('OPC_URL', None)
+    environ.pop('OPC_SERVER_URI', None)
+    environ.pop('OPC_RECONNECTION_INTERVAL', None)
+
+    # Act
+    config = build_opc_config()
+
+    # Assert
+    assert config['1']['id'] == '1'
+    assert config['1']['url'] == 'opc.tcp://localhost:4840'
+    assert config['1']['server_uri'] == 'opc.tcp://localhost:4840'
+    assert config['1']['reconnection_interval'] == 120
+
+
+def test_build_minio_config_with_env_vars():
+    environ['MINIO_ENDPOINT_URL'] = 'http://test-host'
+    environ['MINIO_ACCESS_KEY'] = 'test-key'
+    environ['MINIO_SECRET_KEY'] = 'test-secret'
+    environ['MINIO_REGION_NAME'] = 'test-region'
+    environ['MINIO_DEFAULT_BUCKET'] = 'test-bucket'
+    # Isolate from IDE/CI env (e.g. VS Code may export MINIO_SECURE=true).
+    environ['MINIO_SECURE'] = 'false'
+    assert build_minio_config() == {
+        'endpoint_url': 'http://test-host',
+        'access_key': 'test-key',
+        'secret_key': 'test-secret',
+        'default_bucket': 'test-bucket',
+        'retention_hours': 24,
+        'secure': False,
+    }
+
+
+def test_build_minio_config_with_defaults():
+    environ.pop('MINIO_ENDPOINT_URL', None)
+    environ.pop('MINIO_ACCESS_KEY', None)
+    environ.pop('MINIO_SECRET_KEY', None)
+    environ.pop('MINIO_REGION_NAME', None)
+    environ.pop('MINIO_DEFAULT_BUCKET', None)
+    environ.pop('MINIO_SECURE', None)
+    environ.pop('MINIO_RETENTION_HOURS', None)
+    assert build_minio_config() == {
+        'endpoint_url': 'http://localhost:9000',
+        'access_key': 'minioadmin',
+        'secret_key': 'minioadmin',
+        'default_bucket': 'laborious',
+        'retention_hours': 24,
+        'secure': False,
+    }

tests/laborious/utils/test_dataframe_debug.py

@@ -0,0 +1,12 @@
+from pandas import DataFrame
+
+from laborious.utils.dataframe_debug import build_dataframe_debug_message
+
+
+def test_build_dataframe_debug_message_skips_large_dataframe():
+    df = DataFrame({'a': [1, 2, 3]})
+
+    msg = build_dataframe_debug_message('payload', df, max_rows=1)
+
+    assert 'skipped because dataframe has 3 rows' in msg
+    assert '(max: 1)' in msg

tests/laborious/worker/test_worker.py

@@ -0,0 +1,243 @@
+from unittest.mock import AsyncMock, MagicMock, patch
+
+from pytest import mark, raises
+
+from laborious.worker import worker
+
+
+def _build_fake_activities():
+    inst = MagicMock()
+    inst.init_opc = MagicMock()
+    inst.shutdown = MagicMock()
+    inst.load_query_with_minio_offload = MagicMock()
+    inst.retrain_model = MagicMock()
+    inst.update_production_model = MagicMock()
+    inst.format_retrain_report = MagicMock()
+    inst.export_data_to_postgres = MagicMock()
+    inst.load_custom_query = MagicMock()
+    inst.calculate_simple_metrics = MagicMock()
+    inst.get_reference_data = MagicMock()
+    inst.calculate_drift = MagicMock()
+    inst.request_predict = MagicMock()
+    inst.request_transform = MagicMock()
+    inst.input_gate = MagicMock()
+    inst.mlflow_response_gate = MagicMock()
+    inst.mlflow_content_gate = MagicMock()
+    inst.format_transformed_data = MagicMock()
+    inst.format_prediction = MagicMock()
+    inst.format_default_prediction = MagicMock()
+    inst.write_opc_data = MagicMock()
+    inst.cleanup_minio_objects_expired = MagicMock()
+    inst.repeat_last_prediction = MagicMock()
+    inst.write_metrics = MagicMock()
+    inst.write_pi_web_api_data = MagicMock()
+    return inst
+
+
+def _build_fake_worker(async_result=None, async_error: Exception | None = None):
+    w = MagicMock()
+
+    async def _run():
+        if async_error is not None:
+            raise async_error
+        return async_result
+
+    w.run = MagicMock(side_effect=_run)
+    return w
+
+
+@patch('laborious.worker.worker.start_http_server')
+def test_start_prometheus_server_success(mock_start_http):
+    with patch.object(worker.metrics.APP_UP, 'labels') as labels:
+        gauge = MagicMock()
+        labels.return_value = gauge
+        with patch('laborious.worker.worker.os.getenv', return_value='9090'):
+            worker.start_prometheus_server()
+    mock_start_http.assert_called_once_with(9090)
+    gauge.set.assert_called_once_with(1)
+
+
+@patch('laborious.worker.worker.start_http_server', side_effect=RuntimeError('nope'))
+def test_start_prometheus_server_error_exits(_mock_start_http):
+    with patch('laborious.worker.worker.os._exit', side_effect=SystemExit(1)) as m_exit:
+        with raises(SystemExit):
+            worker.start_prometheus_server()
+    m_exit.assert_called_once_with(1)
+
+
+@mark.asyncio
+async def test_main_missing_runtime_exits_fast(monkeypatch):
+    monkeypatch.setenv('RUNTIME', '')
+    with (
+        patch('laborious.worker.worker.start_prometheus_server'),
+        patch('laborious.worker.worker.get_logger') as m_logger,
+        patch('laborious.worker.worker.NotificationHandler'),
+        patch('laborious.worker.worker.MetricsController'),
+        patch.object(worker.metrics.APP_UP, 'labels') as labels,
+        patch('laborious.worker.worker.sys.exit', side_effect=SystemExit(1)),
+    ):
+        labels.return_value = MagicMock()
+        with raises(SystemExit):
+            await worker.main()
+    assert m_logger.return_value.custom_critical.called
+
+
+@mark.asyncio
+async def test_main_plugin_install_failure(monkeypatch):
+    monkeypatch.setenv('RUNTIME', 'single')
+    fake_activities = _build_fake_activities()
+    fake_plugin = MagicMock()
+    fake_plugin.install_runtime = AsyncMock(side_effect=RuntimeError('install failed'))
+    with (
+        patch('laborious.worker.worker.start_prometheus_server'),
+        patch('laborious.worker.worker.get_logger'),
+        patch(
+            'laborious.worker.worker.build_mongodb_config',
+            return_value={'connection_string': 'cs', 'database_name': 'db'},
+        ),
+        patch('laborious.worker.worker.NotificationHandler'),
+        patch('laborious.worker.worker.MetricsController'),
+        patch(
+            'laborious.worker.worker.build_plugin_store_config',
+            return_value={
+                'base_url': '',
+                'owner': '',
+                'repo': '',
+                'username': None,
+                'password': None,
+                'branch': None,
+                'cache_ttl_seconds': None,
+                'pypi_index_url': '',
+                'pypi_username': None,
+                'pypi_password': None,
+            },
+        ),
+        patch('laborious.worker.worker.PluginStore', return_value=fake_plugin),
+        patch('laborious.worker.worker.Activities', return_value=fake_activities),
+        patch.object(worker.metrics.APP_UP, 'labels') as labels,
+        patch('laborious.worker.worker.sys.exit', side_effect=SystemExit(1)),
+    ):
+        labels.return_value = MagicMock()
+        with raises(SystemExit):
+            await worker.main()
+
+
+@mark.asyncio
+async def test_main_success_exit_zero(monkeypatch):
+    monkeypatch.setenv('RUNTIME', 'single')
+    fake_activities = _build_fake_activities()
+    fake_plugin = MagicMock()
+    fake_plugin.install_runtime = AsyncMock(return_value=None)
+    fake_workers = [_build_fake_worker() for _ in range(4)]
+
+    with (
+        patch('laborious.worker.worker.start_prometheus_server'),
+        patch('laborious.worker.worker.get_logger'),
+        patch(
+            'laborious.worker.worker.build_mongodb_config',
+            return_value={'connection_string': 'cs', 'database_name': 'db'},
+        ),
+        patch('laborious.worker.worker.NotificationHandler') as m_notif_cls,
+        patch('laborious.worker.worker.MetricsController'),
+        patch(
+            'laborious.worker.worker.build_plugin_store_config',
+            return_value={
+                'base_url': '',
+                'owner': '',
+                'repo': '',
+                'username': None,
+                'password': None,
+                'branch': None,
+                'cache_ttl_seconds': None,
+                'pypi_index_url': '',
+                'pypi_username': None,
+                'pypi_password': None,
+            },
+        ),
+        patch('laborious.worker.worker.PluginStore', return_value=fake_plugin),
+        patch('laborious.worker.worker.Activities', return_value=fake_activities),
+        patch('laborious.worker.worker.build_postgres_config', return_value={}),
+        patch('laborious.worker.worker.build_minio_config', return_value={}),
+        patch('laborious.worker.worker.build_opc_config', return_value={}),
+        patch('laborious.worker.worker.build_api_config', return_value={}),
+        patch('laborious.worker.worker.PrometheusConfig', return_value=MagicMock()),
+        patch('laborious.worker.worker.TelemetryConfig', return_value=MagicMock()),
+        patch('laborious.worker.worker.Runtime', return_value=MagicMock()),
+        patch(
+            'laborious.worker.worker.client.Client.connect', new=AsyncMock(return_value=MagicMock())
+        ),
+        patch('laborious.worker.worker.prepare_worker', side_effect=fake_workers) as m_prepare,
+        patch.object(worker.metrics.APP_UP, 'labels') as labels,
+        patch('laborious.worker.worker.sys.exit', side_effect=SystemExit(0)),
+    ):
+        labels.return_value = MagicMock()
+        with raises(SystemExit):
+            await worker.main()
+
+    notif = m_notif_cls.return_value
+    notif.shutdown.assert_called_once()
+    fake_activities.shutdown.assert_called_once()
+    assert m_prepare.call_count == 4
+    prepare_calls = m_prepare.call_args_list
+    assert prepare_calls[0].kwargs['runtime'] == 'single'
+    assert prepare_calls[3].kwargs['runtime'] == 'single'
+
+
+@mark.asyncio
+async def test_main_worker_gather_error_exits_one(monkeypatch):
+    monkeypatch.setenv('RUNTIME', 'single')
+    fake_activities = _build_fake_activities()
+    fake_plugin = MagicMock()
+    fake_plugin.install_runtime = AsyncMock(return_value=None)
+    fake_workers = [
+        _build_fake_worker(async_error=RuntimeError('boom')),
+        _build_fake_worker(),
+        _build_fake_worker(),
+        _build_fake_worker(),
+    ]
+
+    with (
+        patch('laborious.worker.worker.start_prometheus_server'),
+        patch('laborious.worker.worker.get_logger') as m_logger,
+        patch(
+            'laborious.worker.worker.build_mongodb_config',
+            return_value={'connection_string': 'cs', 'database_name': 'db'},
+        ),
+        patch('laborious.worker.worker.NotificationHandler'),
+        patch('laborious.worker.worker.MetricsController'),
+        patch(
+            'laborious.worker.worker.build_plugin_store_config',
+            return_value={
+                'base_url': '',
+                'owner': '',
+                'repo': '',
+                'username': None,
+                'password': None,
+                'branch': None,
+                'cache_ttl_seconds': None,
+                'pypi_index_url': '',
+                'pypi_username': None,
+                'pypi_password': None,
+            },
+        ),
+        patch('laborious.worker.worker.PluginStore', return_value=fake_plugin),
+        patch('laborious.worker.worker.Activities', return_value=fake_activities),
+        patch('laborious.worker.worker.build_postgres_config', return_value={}),
+        patch('laborious.worker.worker.build_minio_config', return_value={}),
+        patch('laborious.worker.worker.build_opc_config', return_value={}),
+        patch('laborious.worker.worker.build_api_config', return_value={}),
+        patch('laborious.worker.worker.PrometheusConfig', return_value=MagicMock()),
+        patch('laborious.worker.worker.TelemetryConfig', return_value=MagicMock()),
+        patch('laborious.worker.worker.Runtime', return_value=MagicMock()),
+        patch(
+            'laborious.worker.worker.client.Client.connect', new=AsyncMock(return_value=MagicMock())
+        ),
+        patch('laborious.worker.worker.prepare_worker', side_effect=fake_workers),
+        patch.object(worker.metrics.APP_UP, 'labels') as labels,
+        patch('laborious.worker.worker.sys.exit', side_effect=SystemExit(1)),
+    ):
+        labels.return_value = MagicMock()
+        with raises(SystemExit):
+            await worker.main()
+
+    assert m_logger.return_value.custom_error.called

tests/laborious/workflows/subworkflows/test_format_and_export_prediction.py

@@ -0,0 +1,679 @@
+from unittest.mock import ANY, AsyncMock, MagicMock, call, patch
+
+from pytest import fixture, mark
+from sientia_do.temporal.constants import DATETIME_FORMAT_WITH_TZ
+
+from laborious.activities.activities import Activities
+from laborious.workflows.sub_workflows.format_and_export_prediction import FormatAndExportPrediction
+
+
+@fixture
+def format_and_export_prediction():
+    return FormatAndExportPrediction()
+
+
+metadata = {
+    'metadata': {
+        'model_id': 'test_model',
+        'model_name': 'test_model',
+        'workflow_name': 'test_workflow',
+        'schema_name': 'test_schedule',
+    },
+}
+
+
+@mark.asyncio
+@patch(
+    'laborious.workflows.sub_workflows.format_and_export_prediction.workflow',
+    new_callable=AsyncMock,
+)
+async def test_run_none_path_flag(workflow_mock, format_and_export_prediction):
+    input_data = {
+        'metadata': metadata,
+        'path_flag': None,
+        'data': {'test': 'data'},
+        'timestamp': '2021-01-01',
+        'model_id': 1,
+        'model_name': metadata['metadata']['model_name'],
+        'prediction_confidence': 0,
+        'schema': 'test_schema',
+        'table_name': 'test_table',
+        'opc_servers': ['test_server'],
+        'opc_output_config': {'test': 'config'},
+        'prediction_store_policy': 'erl:1',
+    }
+
+    prediction_data = MagicMock()
+    opc_metrics = MagicMock()
+
+    workflow_mock.execute_activity_method.side_effect = [
+        (prediction_data, opc_metrics),
+        MagicMock(),
+        MagicMock(),
+    ]
+
+    await format_and_export_prediction.run(input_data)
+
+    workflow_mock.execute_local_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.format_prediction,
+                {
+                    **metadata,
+                    'data': input_data['data'],
+                    'timestamp': input_data['timestamp'],
+                    'model_id': input_data['model_id'],
+                    'prediction_confidence': input_data['prediction_confidence'],
+                    'prediction_store_policy': input_data['prediction_store_policy'],
+                    'model_name': input_data['model_name'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_opc_data,
+                {
+                    'opc_output_config': input_data['opc_output_config'],
+                    'data': workflow_mock.execute_local_activity_method.return_value,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['table_name'],
+                    'data': prediction_data,
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                    'on_conflict': 'error',
+                    'unique_columns': ['model_id', 'timestamp'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_metrics,
+                {
+                    **metadata,
+                    'prediction': prediction_data,
+                    'opc_metrics': opc_metrics,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    assert workflow_mock.execute_activity_method.call_count == 3
+    assert workflow_mock.execute_local_activity_method.call_count == 1
+
+
+@mark.asyncio
+@patch(
+    'laborious.workflows.sub_workflows.format_and_export_prediction.workflow',
+    new_callable=AsyncMock,
+)
+async def test_run_none_path_flag_with_transformed_data(
+    workflow_mock, format_and_export_prediction
+):
+    # Arrange
+    input_data = {
+        'metadata': metadata,
+        'path_flag': None,
+        'data': {'test': 'data'},
+        'transformed_data': {'transformed': 'data'},
+        'timestamp': '2021-01-01',
+        'model_id': 1,
+        'model_name': metadata['metadata']['model_name'],
+        'prediction_confidence': 0.9,
+        'schema': 'test_schema',
+        'table_name': 'test_table',
+        'transform_table_name': 'test_transform_table',
+        'opc_servers': ['test_server'],
+        'opc_output_config': {'test': 'config'},
+        'prediction_store_policy': 'lts:1',
+    }
+
+    prediction_data = MagicMock()
+    opc_metrics = MagicMock()
+    transformed_data = MagicMock()
+
+    workflow_mock.execute_local_activity_method.side_effect = [
+        prediction_data,  # format_prediction
+        transformed_data,  # format_transformed_data
+    ]
+
+    write_transformed_handler = AsyncMock()
+    workflow_mock.start_activity_method.return_value = write_transformed_handler
+    workflow_mock.execute_activity_method.side_effect = [
+        (prediction_data, opc_metrics),  # write_opc_data
+        MagicMock(),  # export_data_to_postgres (prediction)
+        MagicMock(),  # write_metrics
+    ]
+
+    # Act
+    await format_and_export_prediction.run(input_data)
+
+    # Assert - format_prediction call
+    workflow_mock.execute_local_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.format_prediction,
+                {
+                    **metadata,
+                    'data': input_data['data'],
+                    'timestamp': input_data['timestamp'],
+                    'model_id': input_data['model_id'],
+                    'prediction_confidence': input_data['prediction_confidence'],
+                    'prediction_store_policy': input_data['prediction_store_policy'],
+                    'model_name': input_data['model_name'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            ),
+            call(
+                Activities.format_transformed_data,
+                {
+                    **metadata,
+                    'data': input_data['transformed_data'],
+                    'model_id': input_data['model_id'],
+                    'model_name': input_data['model_name'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            ),
+        ]
+    )
+
+    # Assert - start_activity_method for transformed data export
+    workflow_mock.start_activity_method.assert_called_once_with(
+        Activities.export_payload_to_postgres,
+        {
+            **metadata,
+            'schema': input_data['schema'],
+            'table_name': input_data['transform_table_name'],
+            'data': transformed_data,
+            'timestamp_conversion': {
+                'column': 'timestamp',
+                'format': DATETIME_FORMAT_WITH_TZ,
+            },
+        },
+        retry_policy=ANY,
+        start_to_close_timeout=ANY,
+    )
+
+    # Assert - write_opc_data call
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_opc_data,
+                {
+                    'opc_output_config': input_data['opc_output_config'],
+                    'data': prediction_data,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    # Assert - export_data_to_postgres for prediction call
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['table_name'],
+                    'data': prediction_data,
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                    'on_conflict': 'error',
+                    'unique_columns': ['model_id', 'timestamp'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    # Assert - write_metrics call
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_metrics,
+                {
+                    **metadata,
+                    'prediction': prediction_data,
+                    'opc_metrics': opc_metrics,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    # Assert - verify counts
+    assert workflow_mock.execute_activity_method.call_count == 3
+    assert workflow_mock.execute_local_activity_method.call_count == 2
+    assert workflow_mock.start_activity_method.call_count == 1
+
+
+@mark.asyncio
+@patch(
+    'laborious.workflows.sub_workflows.format_and_export_prediction.workflow',
+    new_callable=AsyncMock,
+)
+async def test_run_default_path_flag(workflow_mock, format_and_export_prediction):
+    input_data = {
+        'metadata': metadata,
+        'path_flag': 'default',
+        'data': {'test': 'data'},
+        'timestamp': '2021-01-01',
+        'model_id': 1,
+        'model_name': metadata['metadata']['model_name'],
+        'prediction_confidence': 0,
+        'schema': 'test_schema',
+        'table_name': 'test_table',
+        'opc_servers': ['test_server'],
+        'opc_output_config': {'test': 'config'},
+        'comment': 'test_comment',
+    }
+
+    prediction_data = MagicMock()
+    opc_metrics = MagicMock()
+
+    workflow_mock.execute_activity_method.side_effect = [
+        (prediction_data, opc_metrics),
+        MagicMock(),
+        MagicMock(),
+    ]
+
+    await format_and_export_prediction.run(input_data)
+
+    workflow_mock.execute_local_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.format_default_prediction,
+                {
+                    **metadata,
+                    'timestamp': input_data['timestamp'],
+                    'model_id': input_data['model_id'],
+                    'prediction_confidence': input_data['prediction_confidence'],
+                    'comment': input_data['comment'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_opc_data,
+                {
+                    'opc_output_config': input_data['opc_output_config'],
+                    'data': workflow_mock.execute_local_activity_method.return_value,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['table_name'],
+                    'data': prediction_data,
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                    'on_conflict': 'error',
+                    'unique_columns': ['model_id', 'timestamp'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_metrics,
+                {
+                    **metadata,
+                    'prediction': prediction_data,
+                    'opc_metrics': opc_metrics,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    assert workflow_mock.execute_activity_method.call_count == 3
+    assert workflow_mock.execute_local_activity_method.call_count == 1
+
+
+@mark.asyncio
+@patch(
+    'laborious.workflows.sub_workflows.format_and_export_prediction.workflow',
+    new_callable=AsyncMock,
+)
+async def test_run_none_path_flag_with_pi_web_api(workflow_mock, format_and_export_prediction):
+    input_data = {
+        'metadata': metadata,
+        'path_flag': None,
+        'data': {'test': 'data'},
+        'timestamp': '2021-01-01',
+        'model_id': 1,
+        'model_name': metadata['metadata']['model_name'],
+        'prediction_confidence': 0,
+        'schema': 'test_schema',
+        'table_name': 'test_table',
+        'pi_web_api_output_config': {
+            'endpoint': 'https://test-pi-server.com',
+            'prediction_tags': {},
+            'confidence_tags': {},
+        },
+        'prediction_store_policy': 'erl:1',
+    }
+
+    pi_web_api_data = MagicMock()
+
+    workflow_mock.execute_activity_method.side_effect = [
+        pi_web_api_data,  # write_pi_web_api_data
+        MagicMock(),  # export_data_to_postgres
+        MagicMock(),  # write_metrics
+    ]
+
+    await format_and_export_prediction.run(input_data)
+
+    workflow_mock.execute_local_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.format_prediction,
+                {
+                    **metadata,
+                    'data': input_data['data'],
+                    'timestamp': input_data['timestamp'],
+                    'model_id': input_data['model_id'],
+                    'prediction_confidence': input_data['prediction_confidence'],
+                    'prediction_store_policy': input_data['prediction_store_policy'],
+                    'model_name': input_data['model_name'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_pi_web_api_data,
+                {
+                    'pi_web_api_output_config': input_data['pi_web_api_output_config'],
+                    'data': workflow_mock.execute_local_activity_method.return_value,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['table_name'],
+                    'data': pi_web_api_data,
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                    'on_conflict': 'error',
+                    'unique_columns': ['model_id', 'timestamp'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_metrics,
+                {
+                    **metadata,
+                    'prediction': pi_web_api_data,
+                    'opc_metrics': {},
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    assert workflow_mock.execute_activity_method.call_count == 3
+    assert workflow_mock.execute_local_activity_method.call_count == 1
+
+
+@mark.asyncio
+@patch(
+    'laborious.workflows.sub_workflows.format_and_export_prediction.workflow',
+    new_callable=AsyncMock,
+)
+async def test_run_none_path_flag_with_pi_web_api_and_opc(
+    workflow_mock, format_and_export_prediction
+):
+    input_data = {
+        'metadata': metadata,
+        'path_flag': None,
+        'data': {'test': 'data'},
+        'timestamp': '2021-01-01',
+        'model_id': 1,
+        'model_name': metadata['metadata']['model_name'],
+        'prediction_confidence': 0,
+        'schema': 'test_schema',
+        'table_name': 'test_table',
+        'opc_output_config': {'test': 'config'},
+        'pi_web_api_output_config': {
+            'endpoint': 'https://test-pi-server.com',
+            'prediction_tags': {},
+            'confidence_tags': {},
+        },
+        'prediction_store_policy': 'erl:1',
+    }
+
+    prediction_data = MagicMock()
+    pi_web_api_data = MagicMock()
+    opc_metrics = MagicMock()
+
+    workflow_mock.execute_activity_method.side_effect = [
+        pi_web_api_data,  # write_pi_web_api_data
+        (prediction_data, opc_metrics),  # write_opc_data
+        MagicMock(),  # export_data_to_postgres
+        MagicMock(),  # write_metrics
+    ]
+
+    await format_and_export_prediction.run(input_data)
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_pi_web_api_data,
+                {
+                    'pi_web_api_output_config': input_data['pi_web_api_output_config'],
+                    'data': workflow_mock.execute_local_activity_method.return_value,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            ),
+            call(
+                Activities.write_opc_data,
+                {
+                    'opc_output_config': input_data['opc_output_config'],
+                    'data': pi_web_api_data,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            ),
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['table_name'],
+                    'data': prediction_data,
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                    'on_conflict': 'error',
+                    'unique_columns': ['model_id', 'timestamp'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_metrics,
+                {
+                    **metadata,
+                    'prediction': prediction_data,
+                    'opc_metrics': opc_metrics,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    assert workflow_mock.execute_activity_method.call_count == 4
+    assert workflow_mock.execute_local_activity_method.call_count == 1
+
+
+@mark.asyncio
+@patch(
+    'laborious.workflows.sub_workflows.format_and_export_prediction.workflow',
+    new_callable=AsyncMock,
+)
+async def test_run_default_path_flag_with_pi_web_api(workflow_mock, format_and_export_prediction):
+    input_data = {
+        'metadata': metadata,
+        'path_flag': 'default',
+        'data': {'test': 'data'},
+        'timestamp': '2021-01-01',
+        'model_id': 1,
+        'model_name': metadata['metadata']['model_name'],
+        'prediction_confidence': 0,
+        'schema': 'test_schema',
+        'table_name': 'test_table',
+        'pi_web_api_output_config': {
+            'endpoint': 'https://test-pi-server.com',
+            'prediction_tags': {},
+            'confidence_tags': {},
+        },
+        'comment': 'test_comment',
+    }
+
+    pi_web_api_data = MagicMock()
+
+    workflow_mock.execute_activity_method.side_effect = [
+        pi_web_api_data,  # write_pi_web_api_data
+        MagicMock(),  # export_data_to_postgres
+        MagicMock(),  # write_metrics
+    ]
+
+    await format_and_export_prediction.run(input_data)
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.write_pi_web_api_data,
+                {
+                    'pi_web_api_output_config': input_data['pi_web_api_output_config'],
+                    'data': workflow_mock.execute_local_activity_method.return_value,
+                    **metadata,
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    workflow_mock.execute_activity_method.assert_has_calls(
+        [
+            call(
+                Activities.export_data_to_postgres,
+                {
+                    **metadata,
+                    'schema': input_data['schema'],
+                    'table_name': input_data['table_name'],
+                    'data': pi_web_api_data,
+                    'timestamp_conversion': {
+                        'column': 'timestamp',
+                        'format': DATETIME_FORMAT_WITH_TZ,
+                    },
+                    'on_conflict': 'error',
+                    'unique_columns': ['model_id', 'timestamp'],
+                },
+                retry_policy=ANY,
+                start_to_close_timeout=ANY,
+            )
+        ]
+    )
+
+    assert workflow_mock.execute_activity_method.call_count == 3
+    assert workflow_mock.execute_local_activity_method.call_count == 1