SevaSetu AI

A civic welfare discovery and guidance platform that helps citizens find, understand, and apply for government schemes using AI-powered assistance.

Architecture

┌──────────────────────────────────────────────────────────┐
│                    Frontend (React + Vite)                │
│                    apps/web/                              │
└──────────────────────────┬───────────────────────────────┘
                           │ HTTP / REST API
┌──────────────────────────▼───────────────────────────────┐
│                 API Gateway (FastAPI)                     │
│                 services/api_gateway/                     │
│  ┌────────┬────────┬────────┬────────┬────────┬────────┐ │
│  │  Auth  │Profile │Schemes │  Chat  │  Apps  │ Admin  │ │
│  │ routes │ routes │ routes │ routes │ routes │ routes │ │
│  └────────┴────────┴────────┴────────┴────────┴────────┘ │
└──────────────────────────┬───────────────────────────────┘
                           │
        ┌──────────────────┼──────────────────┐
        ▼                  ▼                  ▼
┌───────────────┐ ┌──────────────┐ ┌──────────────────┐
│ PostgreSQL /  │ │   ChromaDB   │ │  MinIO (Object   │
│   SQLite      │ │  (Vector DB) │ │    Storage)      │
└───────────────┘ └──────────────┘ └──────────────────┘
                           │
┌──────────────────────────▼───────────────────────────────┐
│               Orchestrator Service                        │
│               services/orchestrator/                      │
│  ┌──────────────────────┐  ┌───────────────────────────┐ │
│  │   Rule Engine        │  │  RAG Pipeline             │ │
│  │   (Eligibility)      │  │  (Scheme Retrieval + LLM) │ │
│  └──────────────────────┘  └───────────────────────────┘ │
└───────────────────────────────────────────────────────────┘

Features

• User Authentication — Register and login with JWT-based auth
• Profile Management — Capture citizen demographics (age, income, category, location, etc.)
• Scheme Discovery — Semantic search over government schemes using vector embeddings
• Eligibility Checking — Rule-based eligibility evaluation against citizen profiles
• AI Chat Assistant — RAG-powered conversational guidance using Groq LLM
• Application Tracking — Draft and track scheme applications
• Multi-language Support — English, Hindi, and Bengali

Tech Stack

Component	Technology
Frontend	React, TypeScript, Vite
API Gateway	FastAPI, Python
Database	SQLite (dev) / PostgreSQL (prod)
Vector DB	ChromaDB + sentence-transformers
LLM	Groq (llama-3.3-70b-specdec)
Object Store	MinIO
Auth	JWT (python-jose) + bcrypt
Embeddings	all-MiniLM-L6-v2 (HuggingFace)

Getting Started

Prerequisites

• Python 3.11+
• Node.js 18+
• Docker (optional, for PostgreSQL + MinIO)

Backend Setup

# Clone the repository
git clone <repo-url>
cd IBM-Project

# Create a virtual environment
python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Or install the project with the optional CPU-only RAG stack
uv sync --extra rag-cpu

# Configure environment
cp .env.example .env
# Edit .env with your settings (Groq API key, etc.)

# Run the API server
uvicorn services.api_gateway.main:app --host 127.0.0.1 --port 8000 --reload

The rag-cpu extra restores ChromaDB, LangChain, and sentence-transformers with the CPU-only PyTorch wheel. Vercel installs only the default project dependencies so its serverless function stays below the bundle-size limit.

Frontend Setup

cd apps/web
npm install
npm run dev

The frontend will be available at http://127.0.0.1:5173.

Database & Infrastructure (Optional)

docker compose up -d

This starts PostgreSQL and MinIO containers. The app uses SQLite by default for local development.

API Endpoints

Authentication

Method	Endpoint	Description
POST	/api/v1/auth/register	Create an account
POST	/api/v1/auth/login	Sign in

Profile

Method	Endpoint	Description
GET	/api/v1/profile/me	Get user profile
PATCH	/api/v1/profile/me	Update profile

Schemes

Method	Endpoint	Description
POST	/api/v1/schemes/search	Semantic search

Chat

Method	Endpoint	Description
POST	/api/v1/chat/	AI assistant message

Applications

Method	Endpoint	Description
POST	/api/v1/applications/	Create application draft
GET	/api/v1/applications/me	List user's applications

Environment Variables

Key configuration in .env:

Variable	Default	Description
APP_NAME	SevaSetu AI	Application name
JWT_SECRET_KEY	(change me)	JWT signing secret
JWT_ACCESS_TOKEN_EXPIRE_MINUTES	30	Token expiry
DATABASE_URL	SQLite path	Database connection string
GROQ_API_KEY	(set yours)	Groq LLM API key
CORS_ORIGINS	JSON array of URLs	Allowed CORS origins

Project Structure

IBM-Project/
├── apps/web/                 # React frontend
├── data/
│   ├── fixtures/             # Dataset CSV files
│   ├── relational_db/        # SQLAlchemy models + DB
│   ├── vector_db/            # ChromaDB config
│   └── object_storage/       # MinIO config
├── integrations/external/    # External service integrations
├── packages/shared/          # Shared code (auth, config, schemas)
├── scripts/                  # Bootstrap and utility scripts
├── services/
│   ├── api_gateway/          # FastAPI routes
│   └── orchestrator/         # Rule engine + RAG pipeline
└── tests/                    # Test suite

License

Proprietary — Internal project