feat(api,ui): showcase-completion E4 — run-config phase controls (#410)

[w7-mgfcode]

Summary

Implements E4 — run-config phase controls (epic #410, umbrella #406): start-frame-time run configuration for the showcase pipeline. An operator on /showcase can pick which forecasting models the pipeline trains/backtests, tune the backtest split + winner-ranking metric (WAPE/MAE/RMSE), preview exactly what will run, and find the config recorded on the saved workspace and replayed verbatim. No mid-run re-entry — the linear single-asyncio.Lock pipeline is preserved; all config is start-frame-only. Every change is additive: a legacy frame behaves byte-identically.

Changes

Backend

• app/shared/model_taxonomy.py — public KNOWN_MODEL_TYPES allow-list (drift-locked) so the demo slice validates models without importing a sibling slice.
• app/features/demo/schemas.py — DemoBacktestConfig; DemoRunRequest.{train_model_types,backtest} (additive Optional); WorkspaceListItem.run_config.
• app/features/demo/models.py + migration b7c1d9e3f204 — nullable run_config JSONB replay-input column (not an E1 story slot; see DOMAIN_MODEL D1).
• app/features/demo/workspace.py — create_workspace records run_config (NULL on defaults).
• app/features/demo/pipeline.py — ResolvedRunConfig threads selection/split/metric; step_train fail-fast on a disabled opt-in flag (D6); step_backtest unified custom-selection loop + metric-driven winner (D4); step_v2_train honors the horizon (D8); pipeline_complete echoes the config.
• app/features/model_selection/ — CandidateModelInfo.enabled overlaid in the service from the forecast_enable_* flags (D3; capabilities.py stays pure).

Frontend

• run-config-utils.ts (defaults, dirty-rule helpers, train-plan, split-fit warning, parseRunConfig); DemoBacktestSettingsForm; RunConfigPanel (enabled-filtered picker + form + preview); showcase.tsx wiring (dirty-only inclusion, Load/Replay repopulation); WorkspacePanel config badge; buildReplayRequest forwards run_config verbatim; types/api.ts types.

Docs — additive notes in API_CONTRACTS, DOMAIN_MODEL, RUNBOOKS (incidents 30–31).

Validation

• L1 — ruff + ruff format --check, mypy strict, pyright 0 errors, pnpm lint ✅
• L2 — 2066 backend unit + 446 frontend vitest ✅
• L3 — migration up→down→up clean · 29 demo integration tests · live HTTP probes: catalog enabled overlay, 422 paths (unknown/duplicate/gap≥horizon), full custom run green (winner by rmse, run_config persisted on list+detail), legacy run byte-identical ✅
• L4 — browser dogfood: panel renders, disabled opt-ins hidden, picker/form/preview work ✅

Notes

• GET /model-selection/models isn't in docs/_base/API_CONTRACTS.md (the slice predates the heuristic doc), so the enabled field behavior is documented in the demo rows rather than a one-off endpoint row.

Closes #410.

[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: bc12e73f-cb55-48a8-80f1-c322435ea396

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/showcase-run-config-phase-controls

---

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.}