feat(api,ui): forecast explainability & driver attribution slice (#224)

[w7-mgfcode]

Summary

Adds the explainability vertical slice — rule-based driver attribution for baseline forecasts — closing #224.

• New app/features/explainability/ slice: models, schemas, service, routes, explainers, reason_codes + full test suite (unit + integration).
• Three endpoints under /explain:
  • POST /explain/forecast — explain an ad-hoc baseline h=1 forecast.
  • GET /explain/runs/{run_id} — explain a registry model_run.
  • GET /explain/jobs/{job_id} — explain a completed predict job.
• Baseline models only (naive / seasonal_naive / moving_average); lightgbm / regression → clean RFC 7807 400. Time-safe (reads only <= as_of_date).
• New forecast_explanation table (migration f84258c4cb44) persists each explanation as an audit trail.
• Frontend: ExplanationPanel component + use-explanations hook, wired into the run-detail and forecast-visualization pages.

Bundled fix — fix(ui) (#230)

The slice's dogfood surfaced a pre-existing, repo-wide bug: api.ts only treated application/json as JSON, so RFC 7807 application/problem+json error bodies were never parsed and the raw JSON string leaked into every ErrorDisplay. Fixed here (one line + regression test) because the ExplanationPanel error UX depends on it. Root cause tracked in #230.

Validation

• ruff / mypy --strict / pyright --strict — clean.
• Backend: 1288 unit + 8 explainability integration tests pass.
• Frontend: 104 vitest tests pass; tsc --noEmit + lint clean.
• Real-browser dogfood (PRP Level 5): verified the Forecast Explanation card on /explorer/runs/{id} (full panel), /visualize/forecast (full panel for an explainable predict job), and the graceful neutral message for an unexplainable job.

Notes

• uv.lock has an unrelated local version-string drift (0.2.8 → 0.2.14) left out of this PR.

Closes #224.

[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: 01d4eb32-8dba-4610-8e40-e61aa6ef2f38

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/explainability-driver-attribution

---

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.}