Kids Camera 神奇相機（睿睿）
Status: active
Started: 2026-07-12
Path: /opt/data/kids-camera/
Port: 5053
Proxy: Ming's Note → /kids/
Core Question
How to turn a 3-year-old's love of photo-taking and voice recording into a tiny game that supports real 3–5 developmental learning — without building an education platform?
What It Is
A web app for 睿睿 (3yo) — take a photo, say what it is, hear it back. Three big buttons: 影相 → 講嘢 → 再播. Mobile-first, family use.
Key Features
- Camera capture (getUserMedia) + photo album fallback
- ≤8s voice recording with early stop
- Instant playback + redo
- Free mode + Flashcard mode (vision AI identifies objects)
- AI vision: Xiaomi MiMo-V2-Omni identifies objects, suggests sentences
- Favourite entries
- Local gallery with chronological view
MVP Scope
- ✅ Photo + audio capture
- ✅ Instant playback
- ✅ Vision AI identification
- ✅ Free + Flashcard modes
- ✅ Favourites
- ❌ Video, task cards, stickers, login, cloud album
Why It Matters
- Uses skills the child already enjoys (high engagement)
- Maps to preschool development domains (language, cognition, emotion, motor)
- Core loop: 創作 → 即時回放自己 — matches 睿睿's natural behaviour
- Parent co-play > AI grading
Tech Stack
- Backend: Express + SQLite (better-sqlite3, WAL mode)
- Frontend: HTML/CSS/JS SPA
- Vision API: Xiaomi MiMo-V2-Omni (XIAOMI_API_KEY from /opt/data/.env)
- Upload: multer (10MB per file)
- Storage limit: 1GB
- Dependencies: express, better-sqlite3, multer, dotenv
Storage
- DB: kids-camera.db (WAL mode, 6 entries as of 2026-07-14)
- Uploads: uploads/ (photos as JPEG, audio as WebM)
- ~593KB used / 1GB limit
- Orphan files: 3 (2 corrupted 6-byte photos, 1 audio without DB record)
- Two media routes: Express static + custom
/api/media/{*filepath}
Open Questions
- [ ] Complete orphan file cleanup?
- [ ] Add task cards /寻宝 mode?
- [ ] Family sharing (Henna's phone)?
- [ ] Backup strategy for uploads + DB?
Research
Developmental research for 3–5 year olds saved in:
- _sandbox/sessions/2026-07-12-kids-photo-voice-minigame/output/research-3to5-photo-voice.md
- _sandbox/sessions/2026-07-12-kids-photo-voice-minigame/output/option1-magic-camera-detail.md
Key insight: 3–5 歲最重要係語言表達、認知分類、社交情緒、動作協調。睿睿已自己揀咗兩個超好入口：影相（認知+動作）同錄音（語言+情緒）。
Related Files
- Wiki: [[preschool-development-photo-voice]] (developmental research)
- Sandbox:
_sandbox/sessions/2026-07-12-kids-photo-voice-minigame/
- Plan:
.hermes/plans/2026-07-12_kids-camera-mvp.md
Notes
- DB was recovered from upload filenames on 2026-07-14 (was emptied somehow)
created_at column is INTEGER (Unix timestamp), not TEXT- Vision prompt is in 粵語, returns JSON: {name, category, sentence}
- Media served via
/api/media/uploads/ (static) and /api/media/{*filepath} (custom route)