- Document PluginStore wiring e2e scenarios (4.1.1-4.1.5) covering env
config, explicit branch, missing model, bad repo and bad credentials.
- Add unit tests for DataManagerRepository date parsing, timestamp
dropping, numeric coercion and timezone handling to keep coverage at
100%.
- Mark the cleanup guard in TrainModel as `no branch`, since the false
branch always exits through the re-raise.
- The sientia-module entrypoint now merges requirements.txt with the
Plugin Store runtime and installs both before boot (SIENTIAPDE-2223),
so the worker no longer calls install_runtime or runs the loaded
package version guard.
- Leave runtime libs unpinned in requirements files so the runtime
version wins; cap sqlalchemy<2.1 and evidently<0.7 for compatibility.
- Update README, CLAUDE.md and worker tests to match.
The probes only check app_up, which was set to 1 when the metrics server started,
before install_runtime. A pod that would fail the guard could turn Ready first, so a
rollout would remove the old pod. app_up now goes to 1 right after the guard.
Also: copy sys.modules keys before iterating, raise the sientia_model floor to the
runtime's >=0.14.0, pin the same runtime versions in requirements-local.txt, test
that the snapshot is taken before the install and the real importlib.metadata path,
and document the boot failure in README and CLAUDE.md.
Replaces the re-exec. requirements.txt now pins the union of the basic and xgboost
runtime libraries, so install_runtime has nothing to change in the packages the worker
already imported. If a runtime pin changes and requirements.txt is not updated, the
worker logs which packages changed and exits before starting the workers.
test_main_entrypoint assigned worker.main = AsyncMock() without restoring it, so every
later test of main() ran the mock; it now uses patch.object.
From the code-review-python pass on #34:
- README, CLAUDE.md and the module docstring: the worker re-executes
itself after install_runtime, and SIENTIA_RUNTIME_READY is internal;
setting it in .env/the chart would skip the install silently.
- "Installing runtime" is now logged only on the pass that installs; the
re-executed pass logs that the install was skipped.
- The ponytail comment notes that the install success metrics are lost
to the exec.
- New test: a failed install propagates without setting the flag or
calling os.execv (locks the install -> flag -> exec order).
The worker imports sklearn (metrics, evidently) before
PluginStore.install_runtime replaces it on disk with the runtime's pinned
version, the same order that breaks every sklearn prediction in laborious.
After installing the runtime, the worker now replaces its process once
(SIENTIA_RUNTIME_READY=1 skips the second install), so every module is
loaded from the runtime's packages.
generate_report wrote train_data.csv and test_data.csv with index=False, which
dropped the UTC time index. laborious drift reads test_data.csv from MLflow as
its reference and needs a timestamp column; without it, every model trained here
drifted against the first 30% of its own target window. Both CSVs now write the
index as timestamp. If a column already has that name, the CSV is written as
before and a warning is logged, so training never fails over drift.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Downgraded `sientia_do` version to 1.11.0 in `requirements.txt` to address compatibility issues.
- Updated `data_manager_repository.py` to save training and test data CSVs using float-cast versions of the data, ensuring consistency in the saved outputs.
- Enhanced test coverage in `test_data_manager_repository.py` to verify the presence of the target alias in the generated CSV files.
- Added a new fixture to manage runtime report artifacts in a writable temp directory during E2E tests, addressing permission issues in local CI/dev environments.
- Updated `conftest.py` to include a requirements.txt file in the model packaging path for training activities.
- Refactored existing fixtures to use `pytest.fixture` instead of `pytest_asyncio.fixture` for better compatibility.
- Enhanced the `Reports` class to include a target alias for report metrics, ensuring compatibility with Evidently's reporting requirements.
- Introduced new test scenarios to validate the handling of missing and whitespace-only `date_column` inputs in the training workflow.
These changes improve the robustness of the E2E testing framework and enhance the clarity of model reporting metrics.
- Made `date_column` a required field in `TrainModelParams`, ensuring it must be present in the input data.
- Updated related documentation in `input-sample.md`, `README.md`, and various test scenarios to reflect the change in requirement.
- Adjusted the handling of `date_format` to default to `yyyy-MM-dd HH:mm:ss` if omitted, enhancing usability.
- Refined test scenarios to include new examples and ensure compliance with the updated parameter structure.
These changes improve the robustness of the model training workflow and clarify the expectations for input data.
- Added a method to persist computed regression metrics (MSE, MAE, R²) as MLflow parameters during model training, enhancing model evaluation and tracking.
- Updated the Training class to log the equation path if available, improving artifact management.
- Changed project name in values.yaml from "sientia-dataops-model-manager" to "sientia-model-manager".
- Added new environment variables for GitHub repository and branch configuration.
- Refactored cleanup paths to use a centralized REPORTS_TEMP_DIR constant for consistency.
- Updated runtime configurations and adjusted volume mounts for better resource management.
- Enabled SSH access for the model manager and disabled Grafana dashboard creation.
- Updated tests to reflect changes in directory paths and environment variable usage.
- Replaced synchronous MinIO repository calls with asynchronous counterparts in the Training class for improved performance.
- Enhanced logging throughout the training process to provide better insights into model metadata loading, parameter validation, and training execution.
- Updated the train_test_split function to enforce DataFrame input type, ensuring consistency in data handling.
- Removed the deprecated model_repository.py file to streamline the codebase.
- Adjusted cleanup schedule logic to improve error handling and logging during schedule reconciliation.
- Updated tests to reflect changes in the training workflow and repository interactions.
- Modified `.env.example` to set local defaults for PostgreSQL, MLflow, and MinIO configurations.
- Added MongoDB configuration parameters to the environment setup.
- Updated `README.md` to reflect changes in workflow input parameters and task queue naming conventions.
- Removed the `ModelServing` class to streamline the codebase, as it was deemed unnecessary.
- Adjusted `connectors_config.py` to align with new environment variable names and improve clarity.
- Updated tests to reflect changes in configuration handling and removed tests related to the deleted `ModelServing` class.
- Updated `.env.example` to include new environment variables for MinIO and PyPI configuration.
- Refactored `create_cleanup_schedule` to utilize runtime-specific task queues and improve schedule reconciliation logic.
- Enhanced `Activities` class to require a default bucket in MinIO configuration.
- Adjusted `requirements.txt` to specify version for `evidently`.
- Updated tests to reflect changes in schedule creation and configuration handling.
- Added new ignore rule for Ruff to allow temporary paths in tests.
- Introduced MyPy overrides for specific modules to ignore errors.
- Refactored `Cleanup` and `ExperimentTracking` classes to remove async keywords from methods, improving consistency in method signatures.
- Updated `Training` class methods to handle synchronous operations, enhancing performance and clarity.
- Adjusted `requirements.txt` to remove unnecessary Git dependency, streamlining project setup.
- Updated `Activities` class to improve garbage collection handling.
- Enhanced error messaging in `ExperimentTracking` for better clarity on update failures.
- Refactored `Training` class to streamline exception handling and improve type hints.
- Introduced new methods in `TrainModelParams` for better handling of experiment run IDs and model metadata.
- Added functionality to extract model equations in `DataManagerRepository` for linear regression models.
- Introduced a new activity to load model metadata from the model store.
- Refactored training logic to utilize new model metadata and improved parameter handling.
- Updated the `TrainModelParams` class to include additional fields for model configuration.
- Replaced deprecated utility functions with a custom train-test split implementation.
- Removed unused utility functions and cleaned up the data manager repository.
- Adjusted experiment tracking to include model-specific metadata in notifications.
- Added PluginStore integration for model management.
- Replaced StorageRepository with MinIORepository in Activities, Cleanup, and Training classes.
- Updated training logic to handle validation files and improved data management.
- Enhanced configuration for MinIO and PluginStore in connectors.
- Removed deprecated model repository and storage repository files.
- Updated environment variable handling for new configurations.
This parameter allows customizing the threshold (1-1000) used when rem_static_win is enabled, defaulting to 1 if null.
Updates include parameter definition, business rule validation, repository logic for passing the threshold, documentation in README.md and PIPELINE_PARAMS_CHANGELOG.md, and new unit and integration tests.
