fix: honor prefers-reduced-motion for page transitions (MON-194)

[Walid-peach]

What

Closes the remaining gap in prefers-reduced-motion support flagged by the RGAA self-audit (MON-194, RGAA 13.8/13.9 / WCAG 2.3.3), and removes the now-fixed item from the /accessibilite declaration.

Why

The audit found prefers-reduced-motion unreferenced by literal grep, but the landing page's cinematic scroll hero (AssemblyScrollExperience.tsx) already gates on framer-motion's useReducedMotion() and falls back to StaticExperience — that's why the string search came up empty even though the behavior existed. Auditing the actual remaining continuous/large-motion effects turned up two real gaps: PageTransition.tsx, an unguarded fade + 6px translate that runs on every route change site-wide, and no CSS-level prefers-reduced-motion fallback in globals.css for anything not already gated in script (e.g. the nudge scroll-hint keyframe).

Changes

• frontend/src/components/PageTransition.tsx: read useReducedMotion() and skip the fade/translate animation when set, showing content immediately instead.
• frontend/src/app/globals.css: add a global @media (prefers-reduced-motion: reduce) override that collapses animation/transition durations and disables smooth scrolling, as a CSS-level fallback in addition to the existing JS gates (AssemblyScrollExperience, LiveAssemblyPulse, QuizDeck, and now PageTransition).
• frontend/src/app/accessibilite/page.tsx: remove the "Animations non désactivables" non-conformity item now that it's fixed, following the same pattern as the MON-192/193/195 fixes.

Testing

• npm run lint — clean.
• npm test — 25 suites / 197 tests pass.
• npm run build — compiles and type-checks cleanly; the /votes prerender step fails locally with a 500 from the API, but that's a pre-existing local-environment gap (no local API/DB running against which to statically generate that page) — confirmed the change only touches PageTransition.tsx, globals.css, and the static /accessibilite page, none of which touch /votes data fetching.
• Not manually tested in a real browser with OS-level "reduce motion" toggled (no interactive browser session in this environment) — reasoned through the code path instead: useReducedMotion() from framer-motion reflects matchMedia('(prefers-reduced-motion: reduce)') synchronously on first client render, so AssemblyScrollExperience renders StaticExperience and PageTransition skips its transition, for a user with the OS preference set.

Risks / Notes

• No schema, deploy config, or CI workflow changes.
• The global CSS override in globals.css is a broad * selector; it only fires under prefers-reduced-motion: reduce and only clamps animation/transition duration, so it shouldn't visually break anything — it just makes the very short hover micro-transitions (e.g. HemicycleChart seat dots) effectively instant too, which is expected and desired.

Breaking Changes

None.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
mon-elu	Ready	Preview	Aug 1, 2026 10:37pm

[github-actions]

🧪 dbt data-health check — ✅ passed

22:31:22  52 of 62 PASS source_not_null_raw_votes_vote_id ................................ [PASS in 0.91s]
22:31:22  56 of 62 START test source_unique_raw_deputies_deputy_id ....................... [RUN]
22:31:23  53 of 62 PASS source_not_null_raw_votes_voted_at ............................... [PASS in 0.92s]
22:31:23  57 of 62 START test source_unique_raw_votes_vote_id ............................ [RUN]
22:31:23  56 of 62 PASS source_unique_raw_deputies_deputy_id ............................. [PASS in 0.89s]
22:31:23  58 of 62 START test unique_mart_deputy_scorecard_deputy_id ..................... [RUN]
22:31:23  54 of 62 PASS source_relationships_raw_vote_positions_deputy_id__deputy_id__source_raw_deputies_  [PASS in 1.02s]
22:31:23  59 of 62 START test unique_mart_party_alignment_deputy_id ...................... [RUN]
22:31:23  55 of 62 PASS source_relationships_raw_vote_positions_vote_id__vote_id__source_raw_votes_  [PASS in 1.12s]
22:31:23  60 of 62 START test unique_mart_vote_summary_vote_id ........................... [RUN]
22:31:24  57 of 62 PASS source_unique_raw_votes_vote_id .................................. [PASS in 0.88s]
22:31:24  61 of 62 START test unique_stg_deputies_deputy_id .............................. [RUN]
22:31:24  58 of 62 PASS unique_mart_deputy_scorecard_deputy_id ........................... [PASS in 0.96s]
22:31:24  62 of 62 START test unique_stg_votes_vote_id ................................... [RUN]
22:31:24  59 of 62 PASS unique_mart_party_alignment_deputy_id ............................ [PASS in 0.98s]
22:31:24  60 of 62 PASS unique_mart_vote_summary_vote_id ................................. [PASS in 0.98s]
22:31:25  61 of 62 PASS unique_stg_deputies_deputy_id .................................... [PASS in 1.00s]
22:31:25  62 of 62 PASS unique_stg_votes_vote_id ......................................... [PASS in 0.89s]
22:31:26  
22:31:26  Finished running 62 data tests in 0 hours 0 minutes and 20.56 seconds (20.56s).
22:31:26  
22:31:26  Completed successfully
22:31:26  
22:31:26  Done. PASS=62 WARN=0 ERROR=0 SKIP=0 TOTAL=62

Validates prod data health. A failure means current prod data is stale/broken, not necessarily that this PR is wrong.

[Walid-peach]

Attention Score

100/100 - SAFE TO SKIM

• Base 100
• No Must Fix findings
• No Should Fix findings
• Verdict is not "Needs changes" or "Scope should be split"
• No high-risk paths touched (no migrations, workflows, railway.json, api/main.py CORS/rate-limit, requirements*.txt)
• Diff is small: 21 additions / 6 deletions across 3 files (well under 400 lines / 15 files)
• No new api/, rag/, or scripts/ code, so the "no test changes" penalty doesn't apply
• Testing / Validation Gaps section is non-empty (see below) but doesn't cross a threshold worth penalizing given how mechanical the change is and how well-established the pattern already is elsewhere in the codebase
• Nothing the agent couldn't verify from the diff/CI: build, lint, and full Jest suite already pass in CI
  Reason you can skim: three small, mechanical, well-precedented changes with all CI green and no risky surface touched.

Summary

This PR closes the RGAA 13.8/13.9 / WCAG 2.3.3 gap for prefers-reduced-motion. It correctly identifies that the audit's headline claim was stale — AssemblyScrollExperience.tsx already gates its cinematic scroll hero behind useReducedMotion() — and instead fixes the two real remaining gaps: PageTransition.tsx (an unguarded fade+translate on every route change) and a missing CSS-level @media (prefers-reduced-motion: reduce) fallback in globals.css. It also removes the now-resolved item from /accessibilite, following the exact pattern of the three prior RGAA-fix PRs (MON-192/193/195).

Must Fix

None.

Should Fix

None.

Nice to Have

• PageTransition.tsx has no test coverage for the reduceMotion branch. QuizClient.test.tsx already shows the pattern for mocking useReducedMotion from framer-motion in this codebase, so a similar test here (asserting opacity/transform are set immediately and the requestAnimationFrame path is skipped) would be cheap to add and would guard the behavior long-term. Not blocking given how small and mechanical the change is.
• The global CSS override in globals.css uses a broad *, *::before, *::after selector. This is the standard, well-known idiom for this problem (and intentionally uses 0.001ms rather than 0s so transitionend listeners still fire), so it's not a concern — just worth a reviewer's awareness that it also collapses the very short hover micro-transitions in HemicycleChart.tsx (0.1–0.15s) under reduced motion, which is expected/desired but slightly broader than "large motion" framing in the issue title.

Testing / Validation Gaps

• No new automated test was added for the PageTransition reduced-motion branch (see Nice to Have above).
• No manual browser verification with OS-level "reduced motion" toggled was performed (per the PR's own Testing section) — verified by code reading only. Low risk given the identical pattern already ships and works for AssemblyScrollExperience, LiveAssemblyPulse, and QuizDeck.

Documentation / Reviewer Notes

• /accessibilite update correctly follows the established pattern (MON-192/193/195) for removing a fixed non-conformity item — no further doc changes needed.
• Reviewers should double check in a real browser with "reduce motion" enabled that page navigation no longer shows the fade/translate, since that's the one path not verified end-to-end here.

Verdict

Ready to merge