feat(api,ui): showcase-completion E2 — safe replay & workspace lifecycle (#408)

[w7-mgfcode]

Summary

Implements epic #408 (umbrella #406) per PRPs/PRP-showcase-completion-E2-safe-replay-lifecycle.md:

• Safe replay — every panel Replay now opens a confirmation dialog with a recorded-vs-sent config preview (buildReplayRequest is the single source for both the preview and the wire request). A reset=true workspace escalates to destructive copy + a destructive-styled "Replay & wipe database" button. No code path starts a replay without the dialog. This supersedes the E4 (feat(api,ui): showcase-workspace E4 — workspace restore/replay (list/load workspaces + replay through the run path) #393) "deliberately no confirm dialog" decision, as committed by feat(api,ui): showcase workspace completion — the forecastlab control story #406.
• Lineage — replayed rows carry a replay badge; the loaded view renders the replayed_from_workspace_id ancestor chain (depth-capped at 5, deleted ancestors render "(original deleted)" — never an error), ancestors clickable.
• Lifecycle panel — search (q), show-archived toggle, allow-listed sort, clickable tag chips, pin toggle, per-row dropdown (pin / archive / edit details / delete), rename/notes/tags editor over the E1 PATCH endpoint.
• Server-side list filters — GET /demo/workspaces gains q (name ILIKE), repeated tags (JSONB containment), include_archived (default false — archived rows now hidden by default), allow-listed sort_by/sort_order (unknown → silent default, dimensions precedent); pinned rows always order first; total respects the active filters (scenarios precedent — shared _apply_filters on count + rows).
• Multi-select delete — N sequential single DELETEs behind one confirmation; no new bulk endpoint (product-vision guardrail).
• Compare page — /showcase/compare?a=&b= (frontend-only diff over two served details, mirroring run-compare): config diff, result diff with sign-only WAPE delta, created-objects presence matrix, lineage note, partial-run badges; bad ids degrade to the picker.
• Link health — new GET /demo/workspaces/{id}/health probes the row's soft references in-process via httpx.ASGITransport (the pipeline._Client mechanism — no cross-slice imports): 2xx→alive, 404→dead, anything else→unknown (a probe never 500s the route). Loaded-workspace artifact cards get dead-link warnings + an alive/dead summary chip + a partial-run warning.

CONTRACT(E1) reconciliation

All seven CONTRACT(E1) points reconciled against the merged #407 code with zero divergence — columns, WorkspaceUpdateRequest, PATCH route, GET response fields, DemoRunRequest.replayed_from_workspace_id (+ frontend send), job_ids slot shape all matched the assumed names. One verify-or-add fallback fired: the frontend WorkspaceListItem/WorkspaceDetail TS types did not yet carry the E1 fields — added additively here.

Deferral (Decision 1 of the PRP)

A replay-policy picker (exact / safe-keep / modified) is out of scope — replay stays verbatim. A "modified replay" already exists as Load → edit controls → Run; the confirm dialog carries a "Use Load instead" hint.

Validation

• All five backend gates green (ruff check + format --check, mypy --strict, pyright --strict, pytest -m "not integration" — 2005 passed); demo integration suite green against real Postgres (30 passed, incl. new filter + health tests); strict-mode AST walker green.
• Frontend pnpm lint clean + pnpm test --run green (403 passed; 66 files).
• Level 4 smoke + browser dogfood on a seeded stack: keep-run → health probe (3 alive) → alias delete flips the probe dead + renders the card warning marker and ✕ 1 dead chip → confirmed replay records provenance and the replay badge → lineage strip renders the chain → pin/tags/notes round-trip → compare page renders config/result diff + lineage note → multi-select delete removed 2 metadata rows, created objects verified intact.

[sourcery-ai]

Sorry @w7-mgfcode, your pull request is larger than the review limit of 150000 diff characters

[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: 3f2e7281-fba2-4a13-b639-18b31cb83296

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-completion-e2-safe-replay-lifecycle

---

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.}