feat: ForecastOps Control Center — Full Version (model health, incident export, action layer)

[w7-mgfcode]

Implements PRP-25 (PRPs/PRP-25-forecastops-control-center-full.md), taking the ForecastOps Control Center from its PRP-24 MVP to the Full Version. Closes #219.

What changed

Phase A — Model Health & Drift (feat(api))

• New read-only GET /ops/model-health endpoint. Classifies forecast-error (WAPE) performance drift per (store, product) grain from the full successful-run history — improving / stable / degrading / unknown, ±10% relative band.
• classify_drift() — a pure, totally-defined helper (never raises, tolerates None gaps); mirrors PRP-24's score_retraining_candidate pattern.
• New /ops Model Health section with drift badges + WAPE delta.

Phase B — Incident Report Export (feat(ui))

• incident-report.ts — pure builders: a CSV column set + a Markdown incident report assembled entirely client-side from already-loaded page data.
• "Export report" dropdown in the page header — CSV (attention items) and Markdown (full report).

Phase C — Action Layer (feat(ui))

• Retraining Queue gains row checkboxes + "Retrain selected (N)" → confirm dialog → sequential POST /jobs (never Promise.all — jobs run synchronously server-side), per-item toast.
• "Promote to alias" on Model Health / candidate rows → confirm dialog → POST /registry/aliases.

Vision-alignment notes (per .claude/rules/product-vision.md)

• The /ops page is now an action launcher. Phase C calls the existing, already-sanctioned POST /jobs and POST /registry/aliases endpoints — no new POST /ops/* mutating endpoint is added. The ops backend slice stays read-only (no models.py, no Alembic migration). This mild tension is accepted: the Forecast page already triggers train jobs the same way.
• WebSocket live job updates were assessed and deliberately declined. The feature brief lists WS job updates under "Full Version", but that conflicts with the non-streaming product vision. Existing polling (/ops/summary 15s, use-jobs.ts 5s) already delivers live job state. A WS would be a separate PRP if ever wanted.

Validation

• ruff + ruff format, mypy --strict, pyright --strict — clean.
• pytest — 1158 unit + 205 integration passing (fresh DB).
• Frontend tsc + lint (0 errors) + vitest — 84 tests passing.
• Browser-dogfooded on /ops: Model Health renders with drift badges, CSV + Markdown export download, bulk-retrain creates a (verified completed) train job, promote creates an alias.

No new dependency.

[sourcery-ai]

Sorry @w7-mgfcode, you have reached your weekly rate limit of 500000 diff characters.

Please try again later or upgrade to continue using Sourcery

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: a4d25bdf-fd60-4851-baa7-b706a48865f3

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/ops-control-center-full

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}