Snoozy

Personalized bedtime stories for kids.

Snoozy generates a fresh, age-appropriate bedtime story tailored to a child's name, chosen world, and mood ("vibe"), then narrates it with text-to-speech. With voice cloning, the story can be read in the voice of someone the child loves.

A child hearing their own mom or dad reading a personalized story when the parent isn't present isn't a TTS feature.

Repository layout

snoozy/
  snoozy-android/   # Expo (React Native) mobile app — primary codebase
  backend/          # Node.js / Express API server
  ios/              # Legacy Swift app (inactive)
  docs/             # Architecture, system design, product notes

How it works

1. The parent signs in and sets up a child profile (name, age, pronouns, voice).
2. They pick a world and a vibe for tonight's story.
3. The app calls the backend to generate the story text (LLM).
4. The app calls the backend to generate narration audio (TTS), which is cached and streamed back.
5. The story is saved locally and plays back with audio controls; finished stories live in the library.

HomeScreen → WorldPicker → VibePicker → Generating (story → audio → save) → StoryPlayer → StoryEnd

Frontend — snoozy-android/

An Expo / React Native app (TypeScript).

Stack highlights

• Navigation: no React Navigation — a single Zustand field currentScreen drives conditional Animated.View rendering in App.tsx.
• State: one global Zustand store (src/stores/storyStore.ts) owns screens, child details, saved stories, audio playback, and the generation pipeline.
• Auth: Clerk (@clerk/clerk-expo) — SignedIn / SignedOut wrappers; the JWT is attached to every backend call.
• Storage: expo-file-system for audio + story JSON; AsyncStorage for the child profile and avatar.
• Design tokens: src/config/tokens.ts — use Spacing.*, Radii.*, Fonts.*, Colors.*.

Commands

cd snoozy-android
npx expo start          # start the Metro dev server
npx expo run:android    # build & run on Android device/emulator
npx expo run:ios        # build & run on the iOS simulator
npx tsc --noEmit        # TypeScript type-check

There is no lint script or test suite configured for the app.

See snoozy-android/CLAUDE.md for full architecture and styling constraints (e.g. no shadows, Nunito font only, consistent Spacing.* tokens).

Backend — backend/

An Express API that proxies the LLM and TTS providers and protects the keys.

Endpoints (all under /api, JWT-protected via Clerk requireAuth()):

• POST /api/generate-story — Azure OpenAI chat → returns { title, storyText }.
• POST /api/generate-audio — ElevenLabs or Fish Audio TTS → returns an MP3 buffer.
• GET /health — liveness check.

Notable details

• TTS provider is selected per-request by the client via the provider field (elevenlabs | fishaudio).
• Story text is pre-processed in src/utils/ttsPreprocessor.js (strips markdown, injects SSML break tags per vibe).
• ElevenLabs output is loudness-normalized via src/utils/audioNormalizer.js.
• In-memory audio cache keyed by SHA-256 of (processedText + voiceId), capped at 200 entries / 2h TTL — swap for Redis above ~500 stories/day.
• Per-user rate limits: 15 stories/day and 30 audio requests/day.
• ElevenLabs voice resolution order: explicit voiceId → region default (REGION_VOICE_DEFAULTS) → global fallback (Rachel).

Commands

cd backend
npm install
npm run dev    # node --watch (auto-restart on save)
npm start      # production start

Configuration

Copy backend/.env.example to backend/.env and fill in the keys:

Variable	Purpose
AZURE_OPENAI_API_KEY / AZURE_OPENAI_ENDPOINT / AZURE_OPENAI_CHAT_DEPLOYMENT	Story generation (required)
AZURE_OPENAI_TTS_DEPLOYMENT	Azure TTS deployment
FISH_API_KEY / FISH_AUDIO_VOICE_ID	Fish Audio TTS + voice cloning
ELEVENLABS_API_KEY / ELEVENLABS_VOICE_ID	ElevenLabs TTS
CLERK_SECRET_KEY	Clerk auth (backend secret key)
STRIPE_SECRET_KEY	Stripe payments
RAZORPAY_KEY_ID / RAZORPAY_KEY_SECRET	Razorpay payments (India)

Config is loaded from .env via src/config.js. The service deploys on Render (render.yaml, rootDir: backend).

Documentation

More detail lives in docs/:

• ARCHITECTURE.md
• SYSTEM_DESIGN.md
• PRODUCT_ANALYSIS.md
• TECHNICAL_DEBT.md

License

Proprietary — all rights reserved.