Aurora — AI Voice Health Companion

A voice-first mobile health app built with React Native (Expo SDK 56). Track hydration, sleep, habits, and nutrition through natural conversation with an AI companion powered by Gemini, Supabase, and Deepgram.

Features

Voice AI companion — speak naturally to log water, sleep, meals, and habits
Auto-stop silence detection — recording stops automatically after 3 seconds of silence (real RMS level detection on web, timer on native)
Two-step AI pipeline — Gemini classifies intent → Supabase executes tools → Gemini generates a personalised response
Groq fallback — if Gemini is unavailable, Groq (Llama 3.3 70B) takes over automatically, then OpenAI, then deterministic replies
Animated health score — ring with live beating heart + travelling ECG pulse, color-coded to score
Supabase backend — Row Level Security on all 9 tables, service role key never leaves the server
Google OAuth + email/password auth — via Supabase Auth
Cross-platform — Android, iOS, and web (Expo)
EAS Build ready — build APK without Android Studio

Tech Stack

Layer	Technology
Mobile	React Native 0.85 + Expo SDK 56
Language	TypeScript (strict)
Navigation	React Navigation v7 (conditional auth flow)
Backend API	Node.js + Express 5
Database	Supabase (PostgreSQL + Row Level Security)
Auth	Supabase Auth (email + Google OAuth)
AI — Intent & Response	Google Gemini 2.0 Flash Lite
AI — Fallback	Groq (Llama 3.3 70B) → OpenAI → Deterministic
Voice STT	Deepgram Nova-2 / OpenAI Whisper
Voice TTS	expo-speech
Audio recording	expo-audio (native) + MediaRecorder + AnalyserNode (web)
Animations	React Native Animated API + react-native-svg
Deployment	Render (API server) + EAS Build (APK)

Project Structure

Aurora/
├── src/
│   ├── screens/          # Home, Companion, Sleep, Nutrition, Habits, Profile…
│   ├── components/       # Shared UI — PremiumBottomTabBar, visuals, AnimatedSvg
│   ├── hooks/            # useVoiceRecorder — cross-platform VAD + silence detection
│   ├── navigation/       # AppNavigator — conditional screen registration auth flow
│   ├── services/         # agentApi.ts — text + voice API calls to backend
│   ├── store/            # HealthContext — global state + fire-and-forget Supabase sync
│   ├── lib/              # Supabase client, push notifications
│   ├── theme/            # Design tokens (colors, spacing, typography)
│   └── types/            # TypeScript types (health, navigation)
├── server/
│   └── index.js          # Express API — two-step AI pipeline, Supabase tool functions
├── supabase/
│   └── schema.sql        # Full PostgreSQL schema + RLS policies + auto-profile trigger
├── assets/               # App icons, splash screen
├── .env.example          # Environment variable template (safe to commit)
├── app.json              # Expo app config
├── eas.json              # EAS Build profiles (preview APK, production)
└── render.yaml           # Render.com one-click deploy config

Getting Started

Prerequisites

Node.js 18+
A free Supabase project
A free Gemini API key
A free Deepgram API key
A free Groq API key

1. Clone and install

git clone https://github.com/your-username/aurora.git
cd aurora
npm install

2. Configure environment variables

cp .env.example .env

Open .env and fill in your keys:

# Points the Expo app at the API server
EXPO_PUBLIC_API_URL=http://localhost:4000

# Supabase (get from Supabase → Project Settings → API)
EXPO_PUBLIC_SUPABASE_URL=https://your-project.supabase.co
EXPO_PUBLIC_SUPABASE_ANON_KEY=your-anon-key
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_ANON_KEY=your-anon-key
SUPABASE_SERVICE_ROLE_KEY=your-service-role-key   # server only, never expose

# AI providers
GEMINI_API_KEY=your-gemini-key
GROQ_API_KEY=your-groq-key
DEEPGRAM_API_KEY=your-deepgram-key
OPENAI_API_KEY=                                    # optional — enables Whisper STT

3. Set up the database

Go to Supabase dashboard → SQL Editor
Paste the full contents of supabase/schema.sql
Click Run

This creates 9 tables, all indexes, RLS policies, and an auto-profile trigger that fires on every new signup.

If you get "type already exists" errors, run this first to reset enums:

drop type if exists gender_type cascade;
drop type if exists activity_level_type cascade;
drop type if exists habit_period_type cascade;
drop type if exists meal_type cascade;
drop type if exists chat_role_type cascade;

4. Run locally

npm run dev          # starts API server + Expo together

Or separately:

npm run api          # Express API on http://localhost:4000
npm start            # Expo dev server

On a physical Android device: update .env with your machine's local IP:

EXPO_PUBLIC_API_URL=http://192.168.x.x:4000

Then scan the QR code with Expo Go.

Tunnel mode (if not on same network):

npx expo start --tunnel

Deploying the API Server

The Express server must be publicly accessible for the mobile app to reach it.

Render (recommended — free tier)

Push this repo to GitHub
render.com → New → Web Service → connect repo
Render auto-detects render.yaml — click Apply
Add these environment variables in the Render dashboard:

Key	Value
`SUPABASE_URL`	your Supabase project URL
`SUPABASE_ANON_KEY`	your anon key
`SUPABASE_SERVICE_ROLE_KEY`	your service role key
`GEMINI_API_KEY`	your Gemini key
`GROQ_API_KEY`	your Groq key
`DEEPGRAM_API_KEY`	your Deepgram key

Deploy — you get a URL like https://aurora-api.onrender.com
Update EXPO_PUBLIC_API_URL in eas.json to this URL before building the APK

Free tier note: Render spins down after 15 min of inactivity. Upgrade to $7/mo Starter to keep it always-on.

Building the APK

EAS builds in the cloud — no Android Studio or Java needed locally.

# Install EAS CLI (one time)
npm install -g eas-cli

# Login to your Expo account (create free at expo.dev)
eas login

# Link this project to your Expo account
eas init

# Build the APK
eas build --platform android --profile preview

Build takes ~5–10 minutes. EAS gives you a direct .apk download link — share it with anyone to sideload on Android.

Before building: set EXPO_PUBLIC_API_URL in eas.json to your Render URL (not localhost).

AI Pipeline Architecture

User speaks / types
        │
        ▼
[Step 1] Gemini: classify intent
         Input:  user message only
         Output: { intent, params, confidence }
         (Gemini never sees database credentials)
        │
        ▼
[Step 2] Supabase tool execution (server-side)
         logWater / logSleep / createHabit / logMeal
         getHealthSummary / getWeeklyReport / saveMemory
        │
        ▼
[Step 3] Gemini: generate response
         Input:  sanitised tool result + user profile + memories
         Output: warm coach-like reply (1–3 sentences)
        │
        ▼
        If Gemini fails → Groq → OpenAI → Deterministic fallback

Security contract:

SUPABASE_SERVICE_ROLE_KEY is server-only — never sent to client or AI providers
Gemini receives only sanitised action results, never credentials
Deepgram receives only audio bytes
All tables have RLS — users can only read/write their own rows

Voice Pipeline

Mic open
   │
   ├─ Web:    AnalyserNode samples RMS every 120ms
   │          Speech detected → reset 3s silence timer
   │          3s silence → auto-stop → transcribe → AI pipeline
   │
   └─ Native: 3s timer from recording start → auto-stop
              (or manual tap to stop earlier)
              → transcribe → AI pipeline

STT priority: OpenAI Whisper (if key set) → Deepgram Nova-2 → error

VAD confidence gating:

< 0.3 → show error / confirmation card
0.3–0.7 → show transcript confirmation ("Did I hear this correctly?")
≥ 0.7 → auto-proceed, add to chat, sync to Supabase

Environment Variables Reference

Variable	Used by	Description
`EXPO_PUBLIC_API_URL`	Client	URL of the Express API server
`EXPO_PUBLIC_SUPABASE_URL`	Client	Supabase project URL
`EXPO_PUBLIC_SUPABASE_ANON_KEY`	Client	Supabase anon key (safe to expose)
`SUPABASE_URL`	Server	Supabase project URL
`SUPABASE_ANON_KEY`	Server	Supabase anon key
`SUPABASE_SERVICE_ROLE_KEY`	Server only	Bypasses RLS — never expose
`GEMINI_API_KEY`	Server	Primary AI provider
`GROQ_API_KEY`	Server	Fallback AI provider
`DEEPGRAM_API_KEY`	Server	Voice transcription
`OPENAI_API_KEY`	Server	Optional — Whisper STT + GPT fallback
`OPENAI_MODEL`	Server	Default: `gpt-4.1-mini`

License

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 14 Commits
.claude		.claude
assets		assets
server		server
src		src
supabase		supabase
.env.example		.env.example
.gitignore		.gitignore
AGENTS.md		AGENTS.md
App.tsx		App.tsx
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
app.json		app.json
eas.json		eas.json
index.ts		index.ts
package-lock.json		package-lock.json
package.json		package.json
render.yaml		render.yaml
tsconfig.json		tsconfig.json

Navigation Menu

Repository files navigation

Aurora — AI Voice Health Companion

Features

Tech Stack

Project Structure

Getting Started

Prerequisites

1. Clone and install

2. Configure environment variables

3. Set up the database

4. Run locally

Deploying the API Server

Render (recommended — free tier)

Building the APK

AI Pipeline Architecture

Voice Pipeline

Environment Variables Reference

License

About

Releases

Packages

Contributors

Languages

Folders and files

Latest commit

History

Repository files navigation

Aurora — AI Voice Health Companion

Features

Tech Stack

Project Structure

Getting Started

Prerequisites

1. Clone and install

2. Configure environment variables

3. Set up the database

4. Run locally

Deploying the API Server

Render (recommended — free tier)

Building the APK

AI Pipeline Architecture

Voice Pipeline

Environment Variables Reference

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages