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).
Migrated to a shared sientia-module image and umbrella Helm chart, removing the standalone Dockerfile, .dockerignore, and values.yaml. Introduced CLAUDE.md for AI-assisted development. Updated quality-gate.yml with push triggers, concurrency, and refined permissions. Bumped project version to 2.0.0 to reflect these architectural changes.
