🏥 Swasthya-Setu — स्वास्थ्य-सेतु

AI for Bharat Hackathon 2026 Submission

Swasthya-Setu ("Health Bridge") demonstrates how Generative AI on AWS can support frontline healthcare workers and patients in rural India — with fast AI triage, a guideline-grounded medical chatbot, and a patient management system — all deployed live on AWS.

---

🌐 Live Demo

Service	URL
Frontend (CloudFront)	https://d74rck1rcmkfd.cloudfront.net
Backend API (API Gateway)	https://5zdiwo4lfl.execute-api.ap-south-1.amazonaws.com/api/v1
API Docs (Swagger)	https://5zdiwo4lfl.execute-api.ap-south-1.amazonaws.com/docs

---

🎯 Problem Statement

Rural India faces a severe healthcare accessibility crisis:

• 600M+ people live in rural areas with limited access to doctors
• Frontline health workers (ASHAs, ANMs) lack structured decision-support tools
• Patients travel hours to clinics for conditions that can be triaged remotely
• No affordable, multilingual, AI-powered health guidance exists at scale

Swasthya-Setu is a working proof-of-concept that shows how AI can bridge this gap — today.

---

✨ Features (All Live & Functional)

1. 🧑‍⚕️ Patient Management

• Register patients with name, age, and gender (Male/Female)
• Persistent patient profiles stored in AWS RDS PostgreSQL
• Full consultation history per patient (triage + chat logs)
• Search and browse all registered patients

2. 🚨 AI Symptom Triage (RED / YELLOW / GREEN)

• Enter symptoms in Hindi or English — multilingual input supported
• Powered by GPT-4o-mini for fast, structured classification
• Returns one of three clinically meaningful levels:
  • 🔴 RED — Emergency: Immediate hospital referral
  • 🟡 YELLOW — Priority: Visit clinic or pharmacy
  • 🟢 GREEN — Non-urgent: Home care with guidance
• Includes confidence score, actionable advice, and clinical reasoning
• Every triage is logged to the database for analytics

3. 💬 Medical AI Chatbot (RAG-Based)

• Conversational Q&A powered by GPT-4o-mini + Retrieval-Augmented Generation
• Answers grounded in WHO guidelines, NHM advisories, and Ayushman Bharat documentation
• Shows cited sources for every response — transparent and trustworthy
• Session memory for multi-turn conversations
• Embeddings stored in pgvector on AWS RDS via Amazon Bedrock Titan

4. 📊 Analytics Dashboard

• Real-time consultation counts
• Triage distribution breakdown (RED / YELLOW / GREEN)
• Chat interaction metrics
• Live data from production database

---

🏗️ Architecture

User (Browser)
      │
      ▼
Amazon CloudFront  ──────────────────  Amazon S3
(CDN, HTTPS, SSL)                    (React Frontend)
      │
      ▼
Amazon API Gateway (ap-south-1)
      │
      ▼
AWS Lambda — FastAPI via Mangum adapter
      │
      ├── OpenAI GPT-4o-mini  (Triage + Chatbot LLM)
      ├── Amazon Bedrock Titan (Embeddings for RAG)
      ├── Amazon RDS PostgreSQL + pgvector (Patient data + Vector store)
      └── LangChain (RAG orchestration)

AWS Services Used

Service	Purpose
AWS Lambda	Serverless backend (FastAPI app via Mangum)
Amazon API Gateway	HTTP routing and CORS
Amazon RDS (PostgreSQL)	Patient data + pgvector embeddings
Amazon S3	Frontend static hosting
Amazon CloudFront	CDN, HTTPS, global delivery
Amazon ECR	Docker image registry for Lambda
Amazon Bedrock (Titan)	Vector embeddings for RAG knowledge base

---

🗂️ Tech Stack

Layer	Technology
Frontend	React 18, Vite, Tailwind CSS, Framer Motion
Backend	Python 3.13, FastAPI, Mangum
AI / LLM	OpenAI GPT-4o-mini
RAG	LangChain, pgvector, Amazon Bedrock Titan Embeddings
Database	PostgreSQL (AWS RDS), SQLAlchemy
Deployment	Docker, AWS Lambda (arm64), ECR, API Gateway

---

🚀 Run Locally (Docker)

# Clone the repo
git clone https://github.com/Sumit-ai-dev/swasthya-setu-prototype.git
cd swasthya-setu-prototype

# Start everything with Docker Compose
docker compose up --build

Service	URL
Frontend	http://localhost:5173
Backend	http://localhost:8000
API Docs	http://localhost:8000/docs

Required environment variables — copy backend/.env.example to backend/.env and fill in:

OPENAI_API_KEY=your_openai_key
DATABASE_URL=postgresql://postgres:password@localhost:5432/swasthya_setu
EMBEDDING_PROVIDER=local   # or 'bedrock' for AWS Titan
TRIAGE_MODEL_ID=gpt-4o-mini
CHATBOT_MODEL_ID=gpt-4o-mini

---

☁️ AWS Production Deployment

Backend (Lambda via ECR)

cd backend
bash scripts/deploy.sh

This script:

1. Builds the Docker image for linux/arm64
2. Pushes to Amazon ECR in ap-south-1
3. Creates or updates the Lambda function
4. Applies environment configuration

Frontend (S3 + CloudFront)

cd frontend
npm run build
aws s3 sync dist/ s3://swasthya-setu-frontend-bucket --delete
aws cloudfront create-invalidation --distribution-id <ID> --paths "/*"

Database Initialization

cd backend
export DATABASE_URL=postgresql+psycopg2://...

# Create tables
python3 -m app.scripts.init_db

# Seed WHO/NHM medical knowledge base
python3 -m app.scripts.seed_medical_data

---

🔬 Clinical Standards

The triage engine is calibrated against WHO Emergency Triage Assessment and Treatment (ETAT) guidelines:

• 🔴 RED — Airway, Breathing, Circulation, or Disability compromise; cancer/malignancy indicators; severe fever in infants
• 🟡 YELLOW — Stable but serious conditions requiring clinical investigation; persistent symptoms; chronic conditions
• 🟢 GREEN — Mild self-limiting symptoms; home care appropriate

⚠️ Disclaimer: Swasthya-Setu is an AI demonstration prototype. It is not a substitute for professional medical diagnosis or treatment. Always consult a qualified healthcare provider.

---

🧪 Testing

cd backend
pytest app/tests/test_clinical_triage.py -v

Tests verify triage classification correctness across critical symptom profiles including fever, chest pain, respiratory distress, and oncology red flags.

---

📁 Project Structure

swasthya-setu-prototype/
├── backend/
│   ├── app/
│   │   ├── api/api_v1/endpoints/   # patients, triage, chatbot, analytics
│   │   ├── core/config.py          # Pydantic settings
│   │   ├── db/                     # SQLAlchemy models + session
│   │   ├── schemas/                # Pydantic request/response schemas
│   │   ├── services/               # RAG service, LLM calls
│   │   └── scripts/                # DB init, seed scripts
│   ├── Dockerfile                  # Lambda-compatible container
│   ├── requirements.txt
│   └── scripts/deploy.sh           # AWS Lambda deployment
├── frontend/
│   ├── src/
│   │   ├── components/             # PatientDashboard, Triage, Chatbot, Analytics
│   │   ├── api.js                  # Axios API client
│   │   └── App.jsx
│   └── scripts/deploy.sh           # S3 + CloudFront deployment
└── docker-compose.yml

---

🛣️ Future Roadmap

• Offline-first Android app — DistilBERT on-device triage for no-internet zones
• Indian Sign Language (ISL) support — accessibility for deaf patients
• eSanjeevani / ABDM integration — connect with national health ID
• Voice input in regional languages — Telugu, Tamil, Bengali, Marathi
• Village-level health worker dashboards — ASHA/ANM portal
• Outbreak detection — anomaly detection from triage data patterns

---

👨‍💻 Team

Built by Sumit Das for the AI for Bharat Hackathon 2026 🇮🇳

---

स्वास्थ्य-सेतु — Bridging India's healthcare gap, one consultation at a time.