🚀 Scheme Navigator AI

Scheme Navigator AI is an intelligent government scheme advisory platform powered by multi-agent AI. It analyzes user profiles, intelligently matches them to eligible schemes, identifies missed benefits, and provides actionable guidance for scheme applications.

Live App: scheme-navigator-ai.vercel.app

🛠️ Tech Stack

Languages
Python 3.10+ JavaScript

Frameworks & UI
React 19 FastAPI Vite TailwindCSS

Databases & Search
MongoDB ChromaDB Vector Embeddings

Services & AI
Groq LLM Firebase Auth Axios

✨ Key Features

🤖 1. Multi-Agent AI System

Query Agent: Understands user intent and extracts demographic information (age, income, location, sector)
RAG Agent: Retrieves relevant government schemes from vector database with semantic search
Eligibility Agent: Validates scheme eligibility based on government criteria and user profile
Compliance Agent: Ensures accurate, hallucination-free responses with government policy validation
Action Agent: Provides step-by-step application guidance for scheme enrollment
Alert Agent: Tracks user profile and alerts on new scheme eligibility

🎯 2. Scheme Discovery & Matching

Intelligent Matching: Semantic matching of user profiles to government schemes
23+ Schemes Indexed: Complete database of Indian government welfare schemes
Multi-Criteria Filtering: Income levels, age groups, occupation, region-specific eligibility
Missed Benefits Identification: Highlights schemes user qualifies for but hasn't applied
Real-time Eligibility Check: Instant validation against scheme criteria
Ranking Algorithm: Prioritizes schemes by relevance and applicability

👤 3. Smart User Profiling

Profile Creation: Capture income, age, occupation, state, and family details
Dynamic Classification: Automatically categorize users (farmer, student, elderly, entrepreneur)
Profile Updates: Track eligibility changes as user profile evolves
Secure Storage: Firebase authentication + MongoDB persistence
WhatsApp Integration: Receive scheme alerts on WhatsApp

📊 4. Application Tracking Dashboard

Eligibility Summary: View all schemes user qualifies for
Application Status: Track submitted applications
Benefit Status: Monitor active benefits and payments
Eligibility Timeline: See when new schemes become available
Alerts Panel: Real-time notifications for new opportunities

🗣️ 5. Voice-Enabled Interaction

Voice Input: Speak queries in natural language
Voice Output: Listen to scheme details and application steps
Multi-Language Ready: Foundation for Hindi, Tamil, Kannada, Bengali support
Accessibility: Helps illiterate and elderly users navigate schemes

📱 6. Responsive Mobile-First Design

Mobile Optimized: Works seamlessly on phones, tablets, desktops
Offline Capable: Core functionality works without internet
Touch-Friendly: Large buttons and swipe navigation
Dark Mode Ready: Eye-friendly interface for long sessions
Progressive Enhancement: Graceful degradation for older devices

🔐 7. Enterprise-Grade Security

Firebase Auth: Secure user authentication with MFA support
Data Privacy: Encrypted user profiles and application data
Role-Based Access: Admin dashboards and user restrictions
Audit Logging: Track all user interactions for compliance
HTTPS Only: Secure communication across all layers

⚡ 8. Performance Optimizations

Vector Search: Fast semantic matching with ChromaDB
Keyword Fallback: Ensures results even with vector search misses
Response Caching: Frequently accessed schemes cached in-memory
Lazy Loading: Progressive loading of scheme details
Optimized Embeddings: 25 pre-computed embeddings for instant search

🏗️ System Architecture

┌─────────────────────────────────────────────────────────────┐
│                   PRESENTATION LAYER                        │
│          (React 19 + Vite on Development)                   │
│  ┌──────────────────────────────────────────────────────┐   │
│  │  • User Login & Authentication                       │   │
│  │  • Profile Creation & Management                     │   │
│  │  • Chat Interface with Voice I/O                     │   │
│  │  • Scheme Discovery Dashboard                        │   │
│  │  • Eligibility & Application Tracking                │   │
│  │  • Alerts & Notifications                            │   │
│  │  • WhatsApp Integration                              │   │
│  └──────────────────────────────────────────────────────┘   │
│                    ↓ (Axios REST)                           │
├─────────────────────────────────────────────────────────────┤
│                   API GATEWAY LAYER                         │
│           (FastAPI + Python on Port 8000)                   │
│  ┌──────────────────────────────────────────────────────┐   │
│  │  Routes:                                             │   │
│  │  • POST /chat - Query processing with agents         │   │
│  │  • POST /eligibility - Scheme eligibility check      │   │
│  │  • POST /profile - User profile management           │   │
│  │  • GET /schemes - Scheme browsing & filtering        │   │
│  │  • POST /apply - Application submission              │   │
│  │  • GET /alerts - Alert management                    │   │
│  │                                                      │   │
│  │  Middleware:                                         │   │
│  │  • Firebase authentication                           │   │
│  │  • CORS protection                                   │   │
│  │  • Request validation & logging                      │   │
│  └──────────────────────────────────────────────────────┘   │
│                ↓ (Python)       ↓ (HTTP)                    │
├─────────────────────────────────────────────────────────────┤
│           AI AGENTS LAYER         │   DATA LAYER            │
│       (Multi-Agent Orchestration) │ (Persistence & Search)  │
│ ┌─────────────────────────────┐   │ ┌─────────────────────┐ │
│ │ 1. Query Agent              │   │ │ MongoDB:            │ │
│ │    • Intent extraction      │   │ │  • Users            │ │
│ │    • Entity recognition     │   │ │  • Profiles         │ │
│ │                             │   │ │  • Applications     │ │
│ │ 2. RAG Agent                │   │ │  • Scheme data      │ │
│ │    • Vector search          │   │ │                     │ │
│ │    • Semantic matching      │   │ │ ChromaDB:           │ │
│ │    • Fallback keyword       │   │ │  • Embeddings       │ │
│ │                             │   │ │  • Vector index     │ │
│ │ 3. Eligibility Agent        │   │ │                     │ │
│ │    • Rule-based logic       │   │ │ Firebase:           │ │
│ │    • Criteria validation    │   │ │  • Auth tokens      │ │
│ │    • Score calculation      │   │ │  • User sessions    │ │
│ │                             │   │ └─────────────────────┘ │
│ │ 4. Compliance Agent         │   │                         │
│ │    • Response validation    │   │                         │
│ │    • Hallucination blocking │   │                         │
│ │    • Source citation        │   │                         │
│ │                             │   │                         │
│ │ 5. Action Agent             │   │                         │
│ │    • Step generation        │   │                         │
│ │    • Document requirements  │   │                         │
│ │    • Timeline estimation    │   │                         │
│ │                             │   │                         │
│ │ 6. Alert Agent              │   │                         │
│ │    • Profile monitoring     │   │                         │
│ │    • Eligibility tracking   │   │                         │
│ │    • Notification creation  │   │                         │
│ │                             │   │                         │
│ │  ↓ (Groq LLM API)           │   │                         │
│ │  Groq (llama-3.3-70b)       │   │                         │
│ └─────────────────────────────┘   │                         │
└─────────────────────────────────────────────────────────────┘

🚀 Complete Setup Guide

Step 1: Clone Repository

git clone https://github.com/your-repo/scheme-navigator-ai.git
cd scheme-navigator-ai

Step 2: Environment Configuration

Create .env file in project root:

# LLM Configuration
GROQ_API_KEY=your_groq_api_key_here

# Firebase Configuration
FIREBASE_API_KEY=your_firebase_api_key
FIREBASE_AUTH_DOMAIN=your-project.firebaseapp.com
FIREBASE_PROJECT_ID=your-project-id
FIREBASE_STORAGE_BUCKET=your-project.appspot.com
FIREBASE_MESSAGING_SENDER_ID=your_sender_id
FIREBASE_APP_ID=your_app_id

# MongoDB
MONGODB_URL=mongodb+srv://user:password@cluster.mongodb.net/scheme-navigator

# Server Configuration
API_HOST=0.0.0.0
API_PORT=8000

Step 3: Install Dependencies

Backend:

cd backend
pip install -r requirements.txt

Frontend:

cd frontend
npm install

Step 4: Start Services (3 Terminals)

Terminal 1: Backend API

cd backend
python -m uvicorn main:app --reload --host 0.0.0.0 --port 8000
# Runs on http://localhost:8000
# Swagger UI: http://localhost:8000/docs

Terminal 2: Frontend Application

cd frontend
npm run dev
# Runs on http://localhost:5173

Terminal 3: Run Tests (Optional)

cd backend
python test_api.py

Open Browser: http://localhost:5173

📖 Usage Examples

Example 1: Profile-Based Scheme Discovery

User: "I'm a 32-year-old farmer from Maharashtra with ₹2.5L annual income"
System: 
  Identified as: Farmer, Below middle income
  Found 12 eligible schemes
  Top 3: PM-KISAN, PMAY, Pradhan Mantri Yojana

Example 2: Direct Eligibility Check

User: "Am I eligible for PM-KISAN?"
System:
  YES - You qualify!
  Monthly benefit: ₹500-1000
  Application deadline: Open
  Next steps: Upload land documents

Example 3: Missed Benefits Alert

User: "Show me schemes I might have missed"
System:
  Missed 3 schemes!
  • Pradhan Mantri Gramin Awas Yojana (Housing)
  • PM-WANI (Internet access)
  • State-specific agriculture subsidies

🧪 Testing

# Run all tests
cd backend
python test_api.py

# Expected Output:
# Test 1: Query intent extraction - PASSED
# Test 2: RAG scheme retrieval - PASSED
# Test 3: Eligibility validation - PASSED
# Test 4: Compliance check - PASSED
# Test 5: Action guidance - PASSED
# Test 6: Alert generation - PASSED
# Test 7: End-to-end chat - PASSED
# 
# Result: 7/7 PASSED (100%)

📁 Project Structure

scheme-navigator-ai/
├── backend/
│   ├── main.py                    # FastAPI server
│   ├── orchestrator.py            # Agent coordination
│   ├── agents/
│   │   ├── query_agent.py         # Intent extraction
│   │   ├── rag_agent.py           # Scheme retrieval
│   │   ├── eligibility_agent.py   # Eligibility check
│   │   ├── compliance_agent.py    # Response validation
│   │   ├── action_agent.py        # Application steps
│   │   └── alert_agent.py         # Profile management
│   ├── config/
│   │   ├── settings.py            # Configuration
│   │   └── prompts.py             # Agent prompts
│   ├── data/
│   │   ├── schemes_data.json      # Scheme database
│   │   └── chroma_db/             # Vector embeddings
│   ├── services/
│   │   ├── llm_service.py         # Groq LLM interface
│   │   ├── firebase_service.py    # Authentication
│   │   └── database.py            # Data persistence
│   └── requirements.txt
│
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   │   ├── ChatInterface.jsx     # Main chat UI
│   │   │   ├── ProfileForm.jsx       # User profiles
│   │   │   ├── SchemeCard.jsx        # Scheme display
│   │   │   ├── VoiceInputButton.jsx  # Voice input
│   │   │   ├── MissedBenefitsPanel.jsx
│   │   │   ├── AlertsPanel.jsx       # Notifications
│   │   │   └── MessageBubble.jsx
│   │   ├── hooks/
│   │   │   ├── useChat.js
│   │   │   ├── useProfile.js
│   │   │   └── useSpeechRecognition.js
│   │   ├── services/
│   │   │   ├── api.js               # Backend API client
│   │   │   └── voiceService.js      # Voice processing
│   │   ├── pages/
│   │   │   ├── Home.jsx
│   │   │   ├── ChatPage.jsx
│   │   │   ├── ProfilePage.jsx
│   │   │   └── AlertsPage.jsx
│   │   ├── App.jsx
│   │   └── main.jsx
│   ├── package.json
│   ├── vite.config.js
│   └── tailwind.config.js
│
├── .env                          # Environment variables
├── README.md                     # This file
├── QUICK_START.md               # 30-second setup
└── requirements.txt

🔄 Data Flow Pipelines

Chat Query: User → Query Agent (intent extraction) → RAG Agent (semantic search) → Eligibility Agent (criteria check) → Compliance Agent (validation) → Response

User Profile: Profile Form → Firebase/MongoDB → Alert Agent (scheme matching) → Notifications (in-app/email/WhatsApp)

Application Help: User Query → Action Agent (step-by-step guide) → Document templates + FAQ → Formatted response

🌟 API Endpoints

Method	Endpoint	Purpose
`POST`	`/chat`	Process user query through agent pipeline
`POST`	`/eligibility`	Check scheme eligibility for user
`POST`	`/profile`	Create/update user profile
`GET`	`/schemes`	Browse all indexed schemes
`POST`	`/apply`	Submit scheme application
`GET`	`/alerts`	Fetch user alerts & notifications
`POST`	`/voice`	Process voice input

🔧 API Endpoints

Health Check

GET http://localhost:8000/

Chat Query

POST http://localhost:8000/chat
Content-Type: application/json

{
  "query": "I am a farmer from Maharashtra",
  "profile": {
    "age": 30,
    "state": "Maharashtra",
    "occupation": "farmer",
    "income": 250000
  }
}

Get All Schemes

GET http://localhost:8000/schemes

Save Profile

POST http://localhost:8000/profile
Content-Type: application/json

{
  "profile": {
    "age": 30,
    "income": 250000,
    "state": "Maharashtra",
    "occupation": "farmer"
  }
}

Missed Benefits

GET http://localhost:8000/missed?age=30&state=Maharashtra&occupation=farmer

📊 Data Ingestion

Load scheme data into ChromaDB for semantic search:

python -m backend.data.ingestion add       # Add schemes to ChromaDB
python -m backend.data.ingestion list      # List indexed schemes
python -m backend.data.ingestion reset     # Clear database

🧪 Testing

Run All Tests

python test_api.py

Manual Testing

Open http://localhost:5173 in browser
Fill profile (age, state, occupation)
Send message: "I am a farmer from Maharashtra"
Verify schemes appear with details
Test voice input
Check missed benefits section

🤝 Architecture Overview (Multi-Agent RAG Pipeline)

        Web Browser (React)
             ↓ (API)
         Frontend UI
             ↓ (HTTP POST /chat, /profile)
       FastAPI Backend
             ↓
      Pipeline Orchestrator
             ↓
    ┌─────────┴─────────┐
    ↓                   ↓
  Agents            Database (ChromaDB + JSON)
  1. Query          23 Government Schemes
    2. RAG (ChromaDB)  User Profiles (MongoDB)
  3. Eligibility    
  4. Compliance
  5. LLM Service (Groq Llama-3.3)
  6. Action
  7. Alert (Missed Benefits)

📈 Performance

API Response Time: <2 seconds
Frontend Load: <1 second
Memory Usage: <500MB

📄 License

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

🙏 Acknowledgments

Built for Economic Times GenAI Hackathon
Government scheme data from official sources
LLM powered by Groq
Vector embeddings with ChromaDB

Name		Name	Last commit message	Last commit date
Latest commit History 74 Commits
backend		backend
frontend		frontend
.gitignore		.gitignore
QUICK_START.md		QUICK_START.md
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

🚀 Scheme Navigator AI

🛠️ Tech Stack

✨ Key Features

🤖 1. Multi-Agent AI System

🎯 2. Scheme Discovery & Matching

👤 3. Smart User Profiling

📊 4. Application Tracking Dashboard

🗣️ 5. Voice-Enabled Interaction

📱 6. Responsive Mobile-First Design

🔐 7. Enterprise-Grade Security

⚡ 8. Performance Optimizations

🏗️ System Architecture

🚀 Complete Setup Guide

Step 1: Clone Repository

Step 2: Environment Configuration

Step 3: Install Dependencies

Step 4: Start Services (3 Terminals)

📖 Usage Examples

Example 1: Profile-Based Scheme Discovery

Example 2: Direct Eligibility Check

Example 3: Missed Benefits Alert

🧪 Testing

📁 Project Structure

🔄 Data Flow Pipelines

🌟 API Endpoints

🔧 API Endpoints

Health Check

Chat Query

Get All Schemes

Save Profile

Missed Benefits

📊 Data Ingestion

🧪 Testing

Run All Tests

Manual Testing

🤝 Architecture Overview (Multi-Agent RAG Pipeline)

📈 Performance

📄 License

🙏 Acknowledgments

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages