Sanchalak - Unified Government Scheme Eligibility System

Overview

Sanchalak is a robust, modular platform for conversational data collection, eligibility checking, and scheme management for government benefit programs (e.g., PM-KISAN). It integrates LLM-driven chat, scheme logic, and data storage, and is designed for extensibility, reliability, and real-world deployment.

---

Features

• Conversational AI: Multilingual, voice-enabled chat for data collection and eligibility.
• Scheme Management: Canonical scheme definitions, Prolog-based eligibility, and dynamic field extraction.
• Modular Architecture: Decoupled services for UI, API, data, and orchestration.
• LLM Integration: Pluggable LLM backend (LM Studio, Qwen2-57B-Instruct, etc.).
• Production-Ready: Dockerized, scalable, and easy to deploy.

---

Architecture

See the full architecture diagram: docs/sanchalak_architecture.svg

• UI Frontend: Next.js/React, multilingual, voice-enabled chat.
• UI Backend: Node.js GraphQL server for audio, TTS, and chat integration.
• Schemabot: Python GraphQL API for conversational data extraction and eligibility.
• Scheme Server: Python FastAPI, Prolog, and MongoDB for scheme logic and canonical models.
• EFR Server: Python FastAPI for farmer registry and CRUD operations.
• LM Studio: LLM inference (Qwen2-57B-Instruct or compatible).
• MongoDB: Data storage for scheme server and EFR.

---

Quick Start

1. Local Development

# Install Python and Node.js dependencies
pip install -e src
cd src/app/new_ui && npm install && cd ../../..

# Start all services
bash start_sanchalak.sh

• Access the UI: http://localhost:3000
• All service endpoints and logs are printed in the terminal.

2. Docker Compose (Production/Server)

docker-compose up --build

• All services (UI, APIs, MongoDB, LM Studio) are orchestrated automatically.
• Configure environment variables in docker/.env.

---

Service Endpoints

• UI Frontend: http://localhost:3000
• UI Backend: http://localhost:3001
• EFR Server: http://localhost:8001
• Scheme Server: http://localhost:8002
• Schemabot GraphQL: http://localhost:8003/graphql
• LM Studio: http://localhost:1234
• MongoDB: mongodb://localhost:27017

---

Directory Structure

/README.md
/docker-compose.yml
/docker/                # All Dockerfiles and Docker env
/src/
  app/new_ui/           # Next.js frontend and Node.js backend
  schemabot/            # Python GraphQL API (chat, extraction)
  scheme_server/        # Python FastAPI + Prolog + MongoDB
  efr_server/           # Python FastAPI (farmer registry)
  pipeline/             # Data extraction, normalization, etc.
  translation/          # (Optional) Translation APIs
/docs/                  # Architecture, project docs, cards
/data/                  # Data download scripts
/notebooks/             # EDA, results, figures

---

Environment Variables

• All sensitive keys and service URLs are managed via .env.local (for dev) and docker/.env (for Docker).
• See the sample in docker/.env for all required variables.

---

Documentation

• See docs/ for:
  • Architecture diagram
  • Project card and summary
  • Detailed service and API documentation
  • Scheme definitions and canonical models

---

Contributing

• Fork, branch, and submit PRs.
• Please keep code modular and document new features.

---

License

MIT (c) AnnamAI Team