JanSahayak AI 🇮🇳

JanSahayak AI — India's autonomous welfare navigator. Multilingual AI that helps every citizen discover, verify, and apply for government welfare schemes through voice and chat.

---

🎯 Project Goal

To democratize access to government welfare schemes for the 65% of India's population residing in rural areas. JanSahayak AI bridges the digital divide by removing language barriers and complex navigation challenges — acting as a personal welfare consultant that citizens can talk to in their native language via WhatsApp or web.

₹73,000 Cr in government benefits go unclaimed every year. JanSahayak AI fixes that.

---

✨ Features

• 🗣️ Multilingual Voice — Hindi, Tamil, Telugu, and 9 more Indian languages via Bhashini
• ✅ Smart Eligibility Check — Autonomous agent verifies age, income, occupation, and documents
• 💬 WhatsApp Integration — Works on Meta WhatsApp Cloud API; no app download needed
• 📄 Document Guidance — Simplified checklist for every scheme application
• 📜 Jargon-Free Summaries — Complex legal text converted to plain language
• 🤖 Agentic Loop — OBSERVE → REASON → ACT → VERIFY cycle via LangGraph orchestration
• 🔄 Multi-LLM Fallback — Automatic failover across LLM providers for reliability

---

🛠️ Tech Stack

Layer	Technology
Frontend	React (Vite) PWA — deployed on Vercel
Backend	Python 3.11, FastAPI
AI Orchestration	LangGraph — autonomous agentic loop
LLM	Multi-LLM fallback chain (no single vendor lock-in)
Voice (STT/TTS)	Bhashini API — 12 Indian languages
WhatsApp	Meta WhatsApp Cloud API (free tier, 1000 conv/month)
Knowledge Base	ChromaDB (vector search)
Database	SQLite (dev)
Deployment	Vercel (frontend) · Hugging Face Spaces (backend)

---

🤖 The Agentic Loop

User message
     │
     ▼
 OBSERVE — parse input, extract profile fields (age, income, location)
     │
     ▼
 REASON  — query knowledge base, evaluate eligibility rules
     │
     ▼
   ACT    — fetch scheme details, generate multilingual response
     │
     ▼
 VERIFY  — cross-check facts, confirm document requirements
     │
     ▼
 Reply to user (WhatsApp / Web)

---

📂 Folder Structure

JanSahayak-AI/
├── backend/                  # FastAPI application
│   ├── app/
│   │   ├── main.py           # Entry point + CORS
│   │   ├── config.py         # Settings (pydantic-settings)
│   │   ├── agents/           # LangGraph agentic loop
│   │   ├── routes/           # API endpoints
│   │   │   ├── chat.py
│   │   │   ├── whatsapp_meta.py  # Meta WhatsApp webhook
│   │   │   └── health.py
│   │   ├── services/         # Voice, messaging, eligibility
│   │   └── models/           # SQLAlchemy DB models
│   ├── Dockerfile
│   └── requirements.txt
├── frontend/                 # React (Vite) PWA
│   ├── src/
│   │   ├── App.jsx           # React Router — / = landing, /app = chat
│   │   ├── pages/
│   │   │   ├── Landing.jsx   # Marketing landing page
│   │   │   └── index.jsx     # Chat interface
│   │   └── components/
│   └── vercel.json
├── docs/
│   └── Presentation/         # Hackarena 2.0 pitch deck (HTML)
└── README.md

---

🚀 How to Run

Backend

cd backend
python -m venv .venv
# Windows
.venv\Scripts\activate
# macOS/Linux
source .venv/bin/activate

pip install -r requirements.txt

# Copy and fill in env vars
cp .env.example .env

uvicorn app.main:app --reload
# API live at http://localhost:8000

Frontend

cd frontend
npm install
npm run dev
# App live at http://localhost:5173
# / = landing page   /app = chat interface

WhatsApp Bot (local demo)

# 1. Start backend (above)
# 2. Expose with ngrok
ngrok http 8000
# 3. Set webhook in Meta Developer Console:
#    https://<your-ngrok-url>/meta/webhook
#    Verify Token: jansahayak2024

---

🌐 Live Deployments

Service	URL
Landing page	jansahayak.vercel.app
Backend API	Hugging Face Spaces

---

👥 Team

Name	Role
Ritesh Bonthalakoti	Lead Developer & AI Architect
Shreya Goyal	UX Researcher & Domain Expert

---

📄 License

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