Graph-RAG - Open Source Hybrid Search with Knowledge Graphs

A completely free and open source Retrieval-Augmented Generation (RAG) system that combines Vector Search (PostgreSQL with pgvector) and Knowledge Graphs (Neo4j) for semantic understanding and context-aware question answering.

🎯 Key Features

• 100% Open Source: MIT License, free forever
• Hybrid Search: Vector embeddings + knowledge graph traversal
• Free LLM Models: Use Ollama locally or any API provider
• Auto-Ingestion: Watch directories and automatically process documents
• Knowledge Extraction: Extract entities and relationships automatically
• REST API: FastAPI-based endpoints for integration
• CLI Interface: Interactive command-line chat
• Docker Ready: Pre-configured containerized setup

📚 Documentation

• INSTALLATION.md - Setup and installation guide

🚀 Quick Start (5 minutes)

Prerequisites

• Docker & Docker Compose
• Python 3.11+
• Optional: Ollama for free local LLM

1. Clone and Setup

git clone https://github.com/EnggTalha/Graph_Rag.git
cd Graph_Rag

2. Start Docker Services

docker compose up -d

3. Install Python Dependencies

conda create -n graph_rag python=3.11 -y
conda activate graph_rag
pip install -r requirements.txt

4. Configure Environment

cp .env.example .env
# Edit .env to set your LLM provider
# Default: uses local Ollama (free, no API key needed)

5. Initialize Database

python init_database.py

6. Start Using

Interactive CLI:

python cli.py

REST API Server:

python -m agent.api
# Visit http://localhost:8000/docs for API documentation

🛠️ Technology Stack

Component	Technology
Backend	FastAPI (Python)
Vector DB	PostgreSQL + pgvector
Graph DB	Neo4j
LLM Models	Ollama (local), OpenAI, Anthropic, HuggingFace
Embeddings	HuggingFace (free) or OpenAI
Deployment	Docker & Docker Compose

📁 Project Structure

Graph_Rag/
├── agent/              # Core RAG agent and API
├── ingestion/          # Document processing pipeline
├── frontend/           # Web interface
├── data/               # Your documents go here
├── sql/                # Database schemas
├── docker-compose.yml  # Service configuration
├── cli.py              # Interactive CLI
└── requirements.txt    # Python dependencies

📖 Common Commands

# Start services
docker compose up -d

# Ingest documents
python -m ingestion.ingest --documents ./data --watch --verbose

# Start API server
python -m agent.api

# Start CLI
python cli.py

# View logs
docker compose logs -f

# Stop services
docker compose down

🤖 Using Free Local Models (Ollama)

No API keys needed! Run everything on your machine:

# Install Ollama from https://ollama.ai

# Download a free model
ollama pull mistral

# In .env, set:
LLM_PROVIDER=ollama
LLM_MODEL=mistral

# Start using
python cli.py

💰 Using Commercial APIs (Optional)

If you prefer paid providers with more powerful models:

# In .env, set:
LLM_PROVIDER=openai
LLM_API_KEY=sk-proj-your-key-here

Supported providers:

• OpenAI (GPT-4, GPT-3.5)
• Anthropic (Claude)
• HuggingFace (free cloud models)
• Any OpenAI-compatible API

📋 Supported File Formats

• Text: .txt, .md, .markdown, .rst
• Documents: .pdf, .doc, .docx
• Data: .json, .csv, .xlsx
• Code: .py, .js, .ts, etc.

🔗 API Documentation

Once API server is running:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

🐛 Troubleshooting

Docker services won't start:

# Check if ports are in use
lsof -i :5432    # PostgreSQL
lsof -i :7687    # Neo4j
lsof -i :6379    # Redis

# View logs
docker compose logs

Import errors:

# Reinstall dependencies
pip install -r requirements.txt --force-reinstall

Database connection issues:

# Check database is initialized
python init_database.py

# View database logs
docker compose logs postgres

📞 Support

For issues and questions:

• Check INSTALLATION.md for detailed setup
• Review docker compose logs: docker compose logs
• Check .env configuration

📄 License

MIT License - See LICENSE file for details

🎉 Getting Started

See INSTALLATION.md for complete setup instructions for all platforms (Linux, macOS, Windows, WSL2).

---

Built with ❤️ using FastAPI, PostgreSQL, Neo4j, and Ollama.