feat(spec-detail): 👍 top-left, 👎 bottom-left on the plot, one vote per image

[MarkusNeusinger]

Summary

• 👍 / 👎 directly on the plot. The spec detail view now renders a thumbs-up button top-left and a thumbs-down button bottom-left over the image, on both the static preview and the interactive surface, in place of the .report() flag (reporting stays on the hub page link and in the footer). One tap inks the thumb (brand green / matte red), flashes >>> .liked / >>> .disliked in the centre like .copied, and posts a reaction-only entry to /feedback naming the implementation.
• Votes are countable per image. New Alembic revision a7c3e9d1f5b8 adds feedback.library_id + feedback.language and a (spec_id, library_id) index; FeedbackRepository.reaction_counts(spec_id, library_id, language?) returns {thumbs_up, thumbs_down} counting the newest row per session (ordered by created_at, then id, so ties resolve the same way every run). Nothing is displayed yet — this is the seam for a later "show counts" step.
• One vote per image per session. The choice is kept in localStorage (anyplot_plot_votes) so the thumb stays inked across reloads and carousel switches, the other thumb is natively disabled (faded, tooltip .rated() on a wrapping span), and the server silently drops a repeat 👍/👎 from the same session for the same implementation (FeedbackRepository.has_plot_vote). A vote is final; nobody can flip up and down at will.
• Reaction-only rate limit. Free text stays at 5/min per IP, now counted over message-bearing rows only; a 👍/👎 tap is capped at 30/min, so rating several plots in a row neither trips the free-text limit nor blocks a message written afterwards.
• Shared feedback helpers (FEEDBACK_SESSION_KEY, newFeedbackSessionId, specIdFromPath) move out of FeedbackWidget into utils/feedback.ts and routes/paths.ts; the widget keeps its behaviour (page-level reactions carry no library_id, so they never count as plot votes).

Supersedes #8126 (both thumbs top-left, no per-library field, three months behind main with the files since moved). Placement per owner decision: 👍 top-left, 👎 bottom-left.

Test plan

• /verify-core: uv run pytest tests/unit tests/integration — 2004 passed, 1 skipped (pre-existing MonoLisa italic skip); ruff check + ruff format --check + mypy api core clean. New tests: repository has_plot_vote, reaction_counts (newest-per-session, language filter, session-less rows), count_recent_by_ip(messages_only=); router unit tests for the vote guard, the looser reaction limit and the page-level bypass; integration tests for persisted library_id/language and one-vote-per-session.
• /verify-migrations: throwaway Postgres 16 (pgserver) — alembic heads single head a7c3e9d1f5b8, upgrade head, alembic check "No new upgrade operations detected.", downgrade -1 + upgrade head roundtrip clean.
• /verify-frontend: yarn type-check, yarn lint (0 warnings), Prettier clean, yarn vitest run — 71 files / 640 tests passed. Drove /scatter-basic/python/matplotlib in headless Chromium at 1280px and 390px, light and dark: both thumbs render, tap 👍 → inked + toast + exactly one POST (spec_id, library_id, language, session_id), 👎 locked, vote survives reload. The POST was stubbed locally; no row was written to the shared production DB.
• uv run python -m tools.changelog check --base origin/main passes.
• Copilot review round 1: three findings (native disabled, stable ranking tiebreaker, tz-naive test cutoff) applied in the second commit; threads resolved.

Checklist

• changelog.d/thumbs-on-plot.md added (Added / Changed / Removed), no new bullet in CHANGELOG.md
• Related documentation updated: docs/reference/plausible.md (feedback_submitted mode plot_overlay + library prop, report_issue source), docs/contributing.md (where "report issue" now lives)

Deploy note: the migration only adds nullable columns and an index; sync-postgres.yml runs alembic upgrade head after merge. The frontend sends library_id/language regardless; if the API deploys first, the extra fields are still accepted (they are optional in FeedbackRequest).

🤖 Generated with Claude Code

https://claude.ai/code/session_01Eckdg6kWrjasZMwNn8A1v8

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

🟡 Changes recommended

There are correctness/accessibility issues to address (nondeterministic vote counting ordering and disabled-button semantics, plus a test contract mismatch around tz-naive timestamps).

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR updates the spec detail experience to support one-tap 👍/👎 voting directly on the plot (static and interactive views), persists “one vote per image per session” in localStorage + backend, and extends the feedback pipeline (API, DB, tests, and docs) to store/count implementation-scoped reactions and apply separate rate limits for message vs reaction-only submissions.

Changes:

• Add plot-overlay 👍/👎 UI in SpecDetailView, persist votes per implementation, and submit reaction-only feedback with library_id + language.
• Extend backend feedback schema/router + repository with plot-vote guard, reaction-only rate limit, and per-implementation reaction counting.
• Add migration + model fields for feedback.library_id/feedback.language, plus tests and documentation/changelog updates.

File summaries

File	Description
tests/unit/api/test_feedback_router.py	Adds unit coverage for plot vote guard, persisted impl context, and reaction-only rate limiting behavior.
tests/integration/test_repositories.py	Adds integration coverage for message-only rate limiting, plot-vote detection, and per-implementation reaction counts.
tests/integration/api/test_api_endpoints.py	Extends integration tests to assert persisted library_id/language and one-vote-per-session semantics.
docs/reference/plausible.md	Documents updated analytics semantics for feedback_submitted and report_issue.
docs/contributing.md	Updates contributor guidance for where “report issue” now lives in the UI.
core/database/repositories.py	Implements messages_only rate-limit counting, plot-vote guard lookup, and per-implementation reaction counting.
core/database/models.py	Adds nullable library_id/language fields and the (spec_id, library_id) index definition.
changelog.d/thumbs-on-plot.md	Adds changelog fragment describing the new plot-overlay voting and backend seams.
app/src/utils/feedback.ts	Extracts shared feedback/session helpers and vote keying for per-implementation persistence.
app/src/utils/feedback.test.ts	Adds unit tests for the new feedback utilities.
app/src/sections/spec-detail/SpecDetailView.tsx	Renders thumbs overlay buttons (👍 top-left, 👎 bottom-left), persists vote state, and posts reactions via hook.
app/src/sections/spec-detail/SpecDetailView.test.tsx	Adds tests for overlay button rendering, one-vote lock, optimistic UI, rollback on failure, and persistence across remounts.
app/src/routes/paths.ts	Adds specIdFromPath() for shared route parsing.
app/src/routes/paths.test.ts	Adds tests for specIdFromPath().
app/src/pages/SpecPage.tsx	Wires specId into SpecDetailView and removes the in-plot report action wiring.
app/src/hooks/useQuickReaction.ts	New hook for submitting plot-overlay reaction-only feedback with impl context and analytics.
app/src/components/FeedbackWidget.tsx	Switches to shared specIdFromPath and shared session-id helpers.
api/schemas.py	Extends FeedbackRequest with optional library_id/language.
api/routers/feedback.py	Implements message-vs-reaction rate limits and one-vote-per-image-per-session drop logic; persists impl context.
alembic/versions/a7c3e9d1f5b8_feedback_library_and_language.py	Adds nullable DB columns and (spec_id, library_id) index via Alembic migration.

Review details

• Files reviewed: 20/20 changed files
• Comments generated: 3
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.