Real-time AI-powered health emergency detection and patient triage system with voice alerts for healthcare facilities.

• Real-time Emergency Detection - AI analyzes video feeds for falls, choking, seizures, and other medical emergencies
• Patient Risk Triage - AI-powered risk scoring and prioritization using Wood Wide AI
• Multi-Room Monitoring - Dashboard view of multiple patient rooms with vital signs
• Voice Alert System - Natural voice announcements via ElevenLabs when emergencies are detected
• Two-Way Voice Communication - Staff can interact with the AI assistant using voice commands (LiveKit)
• AI Observability Dashboard - Monitor detection accuracy and model performance (Arize)
• Multi-Language Support - Alerts in English, Spanish, and Mandarin

• Frontend: Next.js 16 (App Router), React 19, TypeScript, Tailwind CSS 4
• AI Detection: Google Gemini 3 Flash (gemini-3-flash-preview)
• Risk Prediction: Wood Wide AI (patient triage, ML model training)
• Object Detection: YOLO service (optional, for person detection)
• Voice Synthesis: ElevenLabs API
• Voice Agent: LiveKit
• Observability: Arize Phoenix
• Notifications: Sonner (toast notifications)

• Node.js 18+
• npm or yarn
• API keys for:
  • Google Gemini
  • ElevenLabs (optional - for voice alerts)

1. Clone the repository:

git clone https://github.com/yourusername/mediwatch.git
cd mediwatch

2. Install dependencies:

npm install

3. Create a .env.local file:

cp .env.example .env.local

4. Add your API keys to .env.local:

# Required
GOOGLE_GEMINI_API_KEY=your_gemini_api_key
ELEVENLABS_API_KEY=your_elevenlabs_api_key

# Optional - LiveKit voice agent
LIVEKIT_API_KEY=your_livekit_api_key
LIVEKIT_API_SECRET=your_livekit_api_secret
NEXT_PUBLIC_LIVEKIT_URL=wss://your-app.livekit.cloud

# Optional - Wood Wide AI patient triage
WOODWIDE_API_URL=https://api.woodwide.ai
WOODWIDE_API_KEY=your_woodwide_api_key

# Optional - YOLO person detection
YOLO_SERVICE_URL=http://localhost:8000

# Optional - Supabase persistence
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key

5. Run the development server:

npm run dev

6. Open http://localhost:3000 in your browser.

1. Start Camera - Click "Start Camera" to enable the video feed
2. Start Analysis - Click "Start Analysis" to begin AI monitoring
3. Voice Agent - Click "Connect" to enable voice interactions
4. View Alerts - Emergency alerts appear in the Alert History panel
5. Acknowledge - Click to acknowledge alerts or use voice command

1. View Priority Queue - See patients ranked by risk score
2. Select Patient - Click a patient card to view detailed risk factors
3. Review Factors - See contributing factors and recommended actions
4. Monitor Vitals - Real-time heart rate and oxygen saturation display

1. Multi-Room View - Monitor multiple patient rooms simultaneously
2. Camera Cards - Each room shows live feed with vital overlays
3. Risk Indicators - Color-coded risk badges (Low/Moderate/Elevated/High)
4. Voice Alerts - Automatic announcements when emergencies detected

• "What happened?" - Get details about the most recent alert
• "Status" - Check current monitoring status
• "Acknowledge" - Acknowledge the latest alert
• "Help" / "Emergency" - Dispatch emergency response

To test emergency detection without a real scenario:

1. Start the camera and analysis
2. Simulate a fall by quickly moving out of frame or lying down
3. The AI will detect the potential emergency and trigger alerts

/src
  /app
    /page.tsx                        # Main dashboard
    /triage/page.tsx                 # Patient triage page
    /realtime/page.tsx               # Real-time monitoring page
    /api
      /analyze/route.ts              # Gemini Vision + YOLO detection
      /tts/route.ts                  # ElevenLabs TTS endpoint
      /stats/route.ts                # Observability stats endpoint
      /triage/route.ts               # Wood Wide AI triage recommendations
      /livekit/route.ts              # LiveKit token generation
      /woodwide/[[...path]]/route.ts # Proxy to Wood Wide AI backend
  /components
    /VideoFeed.tsx                   # Webcam capture and display
    /OvershootVideoFeed.tsx          # Alternative video feed component
    /AlertHistory.tsx                # Emergency alert timeline
    /StatsPanel.tsx                  # Monitoring statistics
    /VoiceAgent.tsx                  # LiveKit voice interaction
    /ObservabilityPanel.tsx          # Arize metrics display
    /TriagePanel.tsx                 # Patient priority queue
    /PatientRiskDetail.tsx           # Detailed risk view
    /CameraCard.tsx                  # Room camera with vitals
    /VideoCard.tsx                   # Video display card
    /Sidebar.tsx                     # Navigation sidebar
    /Navbar.tsx                      # Top navigation bar
    /ui
      /Button.tsx                    # Button component
      /Badge.tsx                     # Badge component
      /RiskBadge.tsx                 # Risk level indicator
  /lib
    /gemini.ts                       # Gemini Vision API client
    /elevenlabs.ts                   # ElevenLabs TTS client
    /observability.ts                # Arize tracing utilities
    /woodwide.ts                     # Wood Wide AI client
  /types
    /index.ts                        # TypeScript type definitions

Healthcare - AI-powered patient safety monitoring for hospitals and care facilities.

The easiest way to deploy your Next.js app is to use the Vercel Platform.

Check out the Next.js deployment documentation for more details.

MIT

Built for NexHacks 2026