An open-source Retrieval-Augmented Generation (RAG) system that allows you to chat with any website using AI.

Features

• Website Scraping: Automatically scrapes and processes website content
• Intelligent Chunking: Adapts chunking strategy based on website type (blog, documentation, e-commerce, etc.)
• Vector Search: Uses local sentence-transformers for embeddings (no API key needed!)
• Chat Interface: Clean, modern UI for chatting with website content
• No Authentication: Simple, ready-to-use setup

Tech Stack

• Next.js 15 - React framework
• Prisma - Database ORM (SQLite)
• Transformers.js - Local ML embeddings (all-MiniLM-L6-v2)
• OpenRouter - Chat completions (Llama 3.2 3B - Free tier)
• Puppeteer - Website scraping
• Tailwind CSS - Styling

Prerequisites

• Docker & Docker Compose (for Docker setup)
• Node.js 18+ or Bun (for local development)
• OpenRouter API key - Get one free at https://openrouter.ai/

Getting Started

Docker (Recommended)

# Clone the repo
git clone https://github.com/sepiropht/rag.git
cd rag
# Set up your API key
cp .env.example .env
# Edit .env and add your OpenRouter API key from https://openrouter.ai/
# Start with Docker
docker compose up -d
# Open http://localhost:3000 in your browser
# Add a website URL and start chatting!

Useful commands:

# View logs
docker compose logs -f
# Stop the application
docker compose down
# Rebuild after changes
docker compose up -d --build

Local Development

1. Clone the repository

git clone https://github.com/sepiropht/rag.git
cd rag

2. Install dependencies

bun install
# or
npm install

3. Set up environment variables

cp .env.example .env

Edit .env and add your OpenRouter API key:

OPENROUTER_API_KEY=your_openrouter_api_key_here
DATABASE_URL="file:./dev.db"

4. Set up the database

bunx prisma generate
bunx prisma migrate dev --name init

5. Run the development server

bun run dev
# or
npm run dev

Open http://localhost:3000 in your browser.

How It Works

1. Add a Website: Enter any website URL on the home page
2. Processing: The system scrapes the website, detects its type, and creates embeddings
3. Chat: Once processing is complete, click on the website card to start chatting
4. Ask Questions: Ask questions about the website content and get AI-powered answers

For a detailed explanation of how RAG works and the technical architecture, check out the blog post: Building an Open-Source RAG System

License

MIT License - feel free to use this project for any purpose.