Celebration layer

The celebration layer (EXP-008, Phase 55) adds the emotional feedback on top of the existing mechanical tracking. It is frontend-only and works in both storage modes.

Pieces

Module	Role
lib/praise/phrase-picker.ts	No-repeat praise-phrase cycling, reads data/praise/{lang}.json (bundled from backend/config/praise/*.yaml via make sync-praise).
lib/feedback/feedbackPref.ts	Intensity (subtle/normal/enthusiastic) + sound prefs in localStorage; reduced-motion override; per-level gating + frequency helpers.
hooks/useFeedbackIntensity.ts	Live effective intensity (re-reads on pref change + reduced-motion change).
lib/feedback/milestones.ts	Pure threshold detection (streak/mastery/level).
lib/feedback/celebrationQueue.ts	Module-level FIFO with id de-dup; one listener (the host).
lib/praise/celebration-bus.ts	emitCelebration (plays the mapped sound + notifies subscribers) and the celebrate* milestone helpers.
lib/feedback/celebration-stats.ts	Snapshots gamification + drives milestone/badge celebration at completion.
lib/audio/sound-effects.ts	Runtime-synthesized sounds (no audio files).
components/exercises/AnswerCelebration.tsx	Per-answer haptic + praise + bus emit.
components/feedback/Confetti.tsx	CSS-only confetti burst.
components/feedback/MilestoneHost.tsx + MilestoneOverlay.tsx	Drains the queue, shows one overlay at a time.

Flow

1. An exercise renders <AnswerCelebration isCorrect>; it fires a haptic, picks a phrase (intensity + frequency gated), and emits answer_correct/answer_wrong on the bus.
2. emitCelebration plays the mapped sound (self-gated on the sound preference) and notifies subscribers.
3. On lesson completion the page snapshots gamification, runs the award, then celebrateProgressSince detects milestones + badges and routes them through celebrateMilestone, which enqueues an overlay and emits a sound. MilestoneHost shows the overlays sequentially.

Rules

• All animations are CSS (transform + opacity); no animation libraries. prefers-reduced-motion suppresses every animation and forces the effective intensity to subtle.
• Sounds are synthesized at runtime from additive recipes (renderSamples is the pure, testable core); the AudioContext is created lazily on the first play inside a user gesture.
• The celebration layer is supplementary: every cue is also visible. A failure in it must never break the lesson flow (the stat reads are defensive).