CareLink - Post-Discharge Patient Monitoring System

An AI-powered patient monitoring and triage system for post-discharge care, featuring automated voice check-ins, intelligent risk assessment, and real-time alerts.

Features

• Automated Voice Check-ins: Daily phone calls to patients via Vapi
• AI-Powered Triage: Claude AI analyzes check-ins and assesses patient risk levels
• Smart Alerts: Multi-channel notifications (Slack, SMS, Email, Calendar) via Composio
• Video Escalation: LiveKit video rooms for urgent cases requiring immediate attention
• Patient Memory: Letta integration for contextual, long-term patient memory
• RAG System: Chroma vector database for discharge instructions and medical context
• Analytics Dashboard: Elastic Cloud + Kibana for insights and monitoring

Tech Stack

Frontend

• Next.js 15 - React framework with App Router
• TypeScript - Type-safe development
• Tailwind CSS - Utility-first styling

Backend

• Supabase - PostgreSQL database with real-time subscriptions
• Anthropic Claude - AI triage and analysis
• Vapi - Voice AI for patient calls
• Chroma - Vector database for RAG
• Letta - Patient memory management
• Composio - Workflow automation (Slack, SMS, Calendar, Gmail)
• LiveKit - Video call infrastructure
• Elastic Cloud - Logging and analytics

Getting Started

Prerequisites

• Node.js 18+ and npm
• Supabase account and project
• API keys for all services (see .env.example)

Installation

1. Clone and install dependencies:

cd CareLink
npm install

2. Set up environment variables:

cp .env.example .env

3. Fill in all required API keys in .env:

   • Supabase credentials
   • Anthropic API key
   • Vapi API key
   • Letta API key
   • Composio API key
   • LiveKit credentials
   • Elastic Cloud credentials

4. Set up Supabase database:

   Run the migration to create all tables:

   # In your Supabase dashboard SQL editor, run:
   supabase/migrations/00001_initial_schema.sql
   

   Or use Supabase CLI:

   supabase db push
   

5. Start Chroma (optional for local development):

docker run -p 8000:8000 chromadb/chroma

6. Run the development server:

npm run dev

Open http://localhost:3000 to view the dashboard.

Project Structure

CareLink/
├── src/
│   ├── app/                    # Next.js App Router pages
│   │   ├── api/               # API routes
│   │   │   ├── ingest/        # Vapi webhook handler
│   │   │   └── triage/        # AI triage endpoint
│   │   ├── dashboard/         # Main dashboard
│   │   ├── patients/          # Patient management
│   │   └── alerts/            # Alert management
│   ├── components/            # React components
│   ├── lib/                   # Utilities and integrations
│   │   ├── supabase.ts       # Supabase client
│   │   ├── chroma/           # RAG system
│   │   ├── letta/            # Patient memory
│   │   ├── composio/         # Workflow automation
│   │   └── livekit/          # Video calls
│   └── types/                # TypeScript types
├── supabase/
│   └── migrations/           # Database schema
└── .env.example             # Environment variables template

How It Works

1. Patient Enrollment

• Hospital staff adds patient to system with discharge details
• Discharge instructions are embedded into Chroma vector database
• Letta creates a memory profile for the patient

2. Daily Check-ins

• Vapi calls patient at preferred time
• Conversation is transcribed and analyzed
• Data sent to /api/ingest webhook

3. AI Triage

Claude analyzes check-in data with context from:

• Discharge instructions (RAG via Chroma)
• Patient history (Letta memory)
• Previous check-ins

Assigns risk score (0-100) and level (low/medium/high/critical) Identifies red flags and recommends actions

4. Alert Generation

• High/Critical Risk: Slack alert to care team
• Critical + Immediate Attention:
  • SMS to caregivers
  • Calendar event for care team
  • LiveKit video room creation
  • Email notification

5. Care Team Response

• Dashboard shows real-time alerts
• Can view patient details, check-in history
• Initiate video calls for urgent cases
• Track outcomes in Elastic/Kibana

API Endpoints

POST /api/ingest

Webhook receiver for Vapi call completions.

Request:

{
  "call": {
    "id": "call_123",
    "startedAt": "2025-01-25T14:30:00Z",
    "endedAt": "2025-01-25T14:35:00Z",
    "transcript": "...",
    "summary": "..."
  },
  "phoneNumber": "+15551234567"
}

POST /api/triage

AI triage assessment endpoint.

Request:

{
  "checkin_id": "uuid",
  "patient_id": "uuid"
}

Response:

{
  "success": true,
  "assessment_id": "uuid",
  "risk_level": "high",
  "requires_immediate_attention": true
}

Database Schema

See supabase/migrations/00001_initial_schema.sql for complete schema.

Main Tables:

• patients - Patient demographics and discharge info
• caregivers - Emergency contacts
• medications - Current prescriptions
• checkins - Voice call data and transcripts
• triage_assessments - AI risk assessments
• alerts - Notifications sent to care team
• discharge_instructions - Care instructions (with RAG)
• livekit_rooms - Video call sessions

Development

Running Tests

npm test

Building for Production

npm run build
npm start

Linting

npm run lint

Deployment

Vercel (Recommended)

vercel deploy

Note: Make sure to set all environment variables in your deployment platform.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

License

MIT License - see LICENSE file for details

Support

For issues or questions, please open a GitHub issue or contact the development team.