RAG AI Agent Backend

A minimal, production-ready FastAPI backend demonstrating Retrieval-Augmented Generation (RAG) with vector similarity search. Built for educational purposes and easy frontend integration.

📹 Full YouTube Guide: Youtube link

🚀 X Post: X link

💡 Try the RAG AI Agent: App link

☕️ Buy me a coffee: Cafe Latte

🤖️ Discord: Invite link

🎯 Features

• FastAPI backend with automatic API documentation
• Supabase integration with pgvector for vector similarity search
• Multi-AI Provider support (OpenAI & Anthropic)
• Vector embeddings with semantic search
• Citation-based answers with source tracking
• Frontend-ready architecture for NextJS integration
• Docker containerization for easy deployment

🏗️ Architecture

yt-rag/
├── app/
│   ├── core/           # Infrastructure (config, database)
│   ├── models/         # Pydantic data models
│   ├── services/       # Business logic (RAG, embeddings)
│   └── main.py         # FastAPI application
├── sql/
│   └── init_supabase.sql  # Database initialization script
└── requirements.txt

🚀 Quick Start Guide

Complete setup from clone to asking questions in ~10 minutes

Prerequisites

• Python 3.11+
• Supabase account
• OpenAI API key
• Anthropic API key (optional, for Claude)

Step 1: Clone and Install Dependencies

# Clone the repository
git clone https://github.com/ShenSeanChen/yt-rag.git
cd yt-rag

# Create virtual environment
python3.11 -m venv venv_yt_rag
source venv_yt_rag/bin/activate  # On Windows: venv_yt_rag\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Step 2: Get API Keys (5 minutes)

Supabase Setup:

1. Go to supabase.com and create a new project
2. Wait for project to be ready (~2 minutes)
3. Go to Settings → API and copy:
   • Project URL (e.g., https://abc123.supabase.co)
   • Anon public key (starts with eyJ...)
   • Service role secret key (starts with eyJ...)

OpenAI Setup:

1. Go to platform.openai.com
2. Create account/sign in → API Keys → Create new key
3. Copy the key (starts with sk-...)

If you meet issues accessing overseas services due to the lack of a Visa or Mastercard, please check out the alternative version of this repository. That version uses Chinese APIs to ensure the project runs smoothly (Supabase is still required, but the free tier fully covers this project).

Step 3: Configure Environment

# Copy environment template
cp .env.example .env

# Edit with your real API keys
nano .env  # or use your preferred editor

Update .env with your values:

# Supabase Configuration
SUPABASE_URL=https://your-project-ref.supabase.co
SUPABASE_ANON_KEY=your_anon_key_here
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key_here

# OpenAI Configuration (using latest models)
OPENAI_API_KEY=sk-your_openai_key_here
OPENAI_EMBED_MODEL=text-embedding-3-large
OPENAI_CHAT_MODEL=gpt-4o

# AI Provider
AI_PROVIDER=openai

# Optional: Anthropic
ANTHROPIC_API_KEY=your_anthropic_key_here
ANTHROPIC_CHAT_MODEL=claude-3-5-sonnet-20241022

Step 4: Initialize Database (2 minutes)

1. Open Supabase Dashboard → SQL Editor
2. Click "New query"
3. Copy entire contents of sql/init_supabase.sql
4. Paste and click "Run"

✅ This creates everything needed:

• pgvector extension
• rag_chunks table with VECTOR(3072) for latest embeddings
• Performance indexes
• Vector search functions
• RLS policies for future auth

Step 5: Test Setup (Optional)

# Test your complete setup
python test_setup.py

This verifies:

• ✅ Dependencies installed
• ✅ API keys configured
• ✅ Database connected
• ✅ Schema initialized
• ✅ RAG pipeline working

Step 6: Start the Server

uvicorn main:app --reload --port 8000

The API will be available at:

• API: http://localhost:8000
• Interactive Docs: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

📚 API Usage

Health Check

curl http://localhost:8000/healthz

Seed Knowledge Base

# Seed with default documents
curl -X POST http://localhost:8000/seed

# Or seed with custom documents
curl -X POST http://localhost:8000/seed \
  -H "Content-Type: application/json" \
  -d '{
    "docs": [
      {
        "chunk_id": "policy_returns_v1#window",
        "source": "https://help.example.com/returns",
        "text": "You can return unworn items within 30 days of purchase..."
      }
    ]
  }'

Ask Questions (RAG)

curl -X POST http://localhost:8000/answer \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Can I return shoes after 30 days?",
    "top_k": 6
  }'

Example Response:

{
  "text": "Based on our return policy, you can return unworn shoes within 30 days of purchase [policy_returns_v1#window]. Items must be in original condition...",
  "citations": ["policy_returns_v1#window", "policy_returns_v1#conditions"],
  "debug": {
    "top_doc_ids": ["policy_returns_v1#window", "policy_returns_v1#conditions"],
    "latency_ms": 1250
  }
}

🔧 Configuration Options

AI Providers

OpenAI (Recommended)

AI_PROVIDER=openai
OPENAI_API_KEY=your_key
OPENAI_EMBED_MODEL=text-embedding-3-small  # 1536 dimensions
OPENAI_CHAT_MODEL=gpt-4o-mini

Anthropic Claude

AI_PROVIDER=anthropic
ANTHROPIC_API_KEY=your_key
ANTHROPIC_CHAT_MODEL=claude-3-haiku-20240307

# Note: Still need OpenAI key for embeddings
OPENAI_API_KEY=your_openai_key

RAG Parameters

Adjust in app/core/config.py:

• chunk_size: Token limit per chunk (default: 400)
• chunk_overlap: Overlap between chunks (default: 60 tokens)
• default_top_k: Number of chunks to retrieve (default: 6)
• temperature: LLM creativity (default: 0.1)

🐳 Docker Deployment

# Build image
docker build -t yt-rag .

# Run container
docker run -p 8080:8080 --env-file .env yt-rag

🔮 NextJS Frontend Integration

This backend is designed for seamless frontend integration:

Frontend Setup (NextJS)

// lib/supabase.js
import { createClient } from '@supabase/supabase-js'

const supabase = createClient(
  process.env.NEXT_PUBLIC_SUPABASE_URL,
  process.env.NEXT_PUBLIC_SUPABASE_ANON_KEY
)

// API calls to your backend
const response = await fetch('http://localhost:8000/answer', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ query: 'user question' })
})

Future Auth Integration When adding Google authentication:

1. Enable Google Auth in Supabase
2. The RLS policies are already configured
3. Frontend and backend will share the same Supabase client
4. No backend changes needed!

📁 Project Structure

yt-rag/
├── app/
│   ├── core/
│   │   ├── config.py      # Environment & settings
│   │   └── database.py    # Supabase client & operations
│   ├── models/
│   │   ├── requests.py    # API request schemas
│   │   ├── responses.py   # API response schemas
│   │   └── entities.py    # Database entities
│   ├── services/
│   │   ├── embedding.py   # AI provider abstraction
│   │   ├── rag.py        # RAG pipeline logic
│   │   └── chunker.py    # Text processing utilities
│   └── main.py           # FastAPI app & routes
├── sql/
│   └── init_supabase.sql # Database setup script
├── .env.example          # Environment template
├── requirements.txt      # Python dependencies
├── Dockerfile           # Container configuration
└── README.md           # This file

🛠️ Development

Running Tests

# Install dev dependencies
pip install pytest pytest-asyncio httpx

# Run tests (coming soon)
pytest

Code Quality

# Format code
black app/
isort app/

# Lint code
flake8 app/

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

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

🙋‍♂️ Support

• 📚 Documentation: Check the /docs endpoint when running
• 🐛 Issues: GitHub Issues
• 💬 Discussions: GitHub Discussions

---

Built with ❤️ for the developer community

This project demonstrates modern RAG architecture patterns and is perfect for learning, prototyping, or building production applications.

💖 Support / Sponsors

If my open-source work saves you time, consider becoming a sponsor. It keeps these repos maintained, documented, and free.