Gemini Alert is a real-time emergency assistance platform. Users can broadcast distress alerts, see nearby helpers on a live map, and receive guidance from a Groq-hosted AI assistant. The project is split into a Vue 3 web client and a Flask backend that proxies Groq API calls and persists data in Firebase services.

• Python 3.14 (or >= 3.10 compatible with groq)
• Node.js 18+ (tested with Node 20/24)
• Firebase project with Authentication and Firestore enabled
• Google Maps JavaScript API key (with proper HTTP referrer restrictions)
• Groq API key

Copy the example files first, then fill in real values:

Frontend values:

VUE_APP_API_BASE_URL=http://localhost:5001
VUE_APP_FIREBASE_API_KEY=your-firebase-web-api-key
VUE_APP_FIREBASE_AUTH_DOMAIN=your-project.firebaseapp.com
VUE_APP_FIREBASE_PROJECT_ID=your-project-id
VUE_APP_FIREBASE_STORAGE_BUCKET=your-project.appspot.com
VUE_APP_FIREBASE_MESSAGING_SENDER_ID=1234567890
VUE_APP_FIREBASE_APP_ID=1:1234567890:web:abcdef123456
VUE_APP_FIREBASE_MEASUREMENT_ID=
VUE_APP_GOOGLE_MAPS_API_KEY=your-google-maps-api-key
VUE_APP_GOOGLE_MAPS_MAP_ID=

FIREBASE_SERVICE_ACCOUNT_KEY_PATH should point to a JSON file stored outside the repo or in another untracked location. In Cloud Run, prefer FIREBASE_SERVICE_ACCOUNT_KEY via Secret Manager rather than shipping a credential file with the code. ALLOWED_ORIGINS accepts a comma-separated list of origins and trims trailing slashes; if you leave it empty, the backend will not allow cross-origin browser requests. Do not commit Firebase Admin credentials, .env files, or any Firebase messaging service worker with inline config.

# Backend
cd backend
python -m venv ../venv
source ../venv/bin/activate   # Windows: ..\venv\Scripts\activate
pip install --upgrade pip
pip install -r requirements.txt

# Frontend
cd ../frontend
npm install

# Terminal 1 - Backend
cd backend
source ../venv/bin/activate
python app.py
# Flask serves on http://localhost:5001

# Terminal 2 - Frontend
cd frontend
npm run dev
# Vue dev server runs on http://localhost:8080

The backend reads backend/.env and the frontend reads frontend/.env.local. If you copy the example files above and fill them in, the app should bootstrap without any committed secrets.

The frontend proxies AI chat calls to the backend (VUE_APP_API_BASE_URL). Ensure both services are running for full functionality.

• Recommended command: gunicorn app:app --bind 0.0.0.0:5001 --workers 2
• Deploy the backend/ directory to Google Cloud Run using the included backend/Dockerfile
• Set environment variables in your hosting provider (GROQ_API_KEY, FIREBASE_SERVICE_ACCOUNT_KEY*, ALLOWED_ORIGINS, PORT, etc.)
• For Cloud Run, inject FIREBASE_SERVICE_ACCOUNT_KEY from Secret Manager instead of relying on a JSON file in the repository or container image
• Point ALLOWED_ORIGINS at your deployed frontend origins, for example https://your-app.vercel.app,https://your-domain.com

cd frontend
npm run build:prod
# Output emitted to frontend/dist/

• Deploy dist/ to static hosting (Vercel, Firebase Hosting, Netlify, S3, etc.)
• The checked-in vercel.json is now frontend-only. Set the Vercel project to this repo root and provide the VUE_APP_* environment variables in Vercel.
• Configure the same environment variables in your hosting dashboard (prefixed with VUE_APP_)
• Set VUE_APP_API_BASE_URL to your deployed Google Cloud backend URL
• Ensure the Google Maps API key allows the deployed domain(s)

• POST /ask – Groq completion via REST
• POST /ask-stream – Groq streaming responses (Server-Sent Events)
• GET /chats – Fetch authenticated user chat history
• POST /api/nearest-users – Find nearby users using Firestore location snapshots
• POST /api/send-sos – Broadcast SOS alert to nearby helpers
• POST /api/alerts/nearby – Retrieve active alerts near the user
• POST /api/alerts/<alertId>/respond – Record assistance responses
• POST /api/devices/register – Register a push token for a user device
• DELETE /api/devices/<token> – Remove a push token
• POST /api/conversations – Create or fetch a direct conversation
• GET /api/conversations – List the authenticated user's conversations
• GET /api/conversations/<conversationId> – Fetch conversation metadata
• GET /api/conversations/<conversationId>/messages – Fetch conversation messages
• POST /api/conversations/<conversationId>/messages – Send a direct message
• POST /api/emotion/analyze – Return deterministic emotion analysis for a message

Bug reports, feature requests, and pull requests are welcome. Please open an issue first to discuss major changes.

This project is licensed under the MIT License. See the LICENSE file for details.