PathRAG - Path-based Retrieval Augmented Generation

PathRAG is an implementation of the paper "PathRAG: Pruning Graph-based Retrieval Augmented Generation with Relational Paths". This system improves retrieval-augmented generation by extracting and utilizing key relational paths from an indexing graph.

System Overview

The application consists of four main components running in Docker containers:

1. Frontend (React + Vite)

   • Modern UI built with Material-UI
   • Interactive knowledge graph visualization
   • File upload and query interface
   • Real-time model selection and results display

2. Backend (FastAPI)

   • RESTful API endpoints for data processing
   • Knowledge graph construction
   • Path-based retrieval implementation
   • Integration with Weaviate and Ollama

3. Weaviate (Vector Database)

   • Stores and indexes knowledge graph nodes
   • Enables semantic similarity search
   • Manages relationships between entities

4. Ollama (LLM Service)

   • Local LLM deployment
   • Handles query processing
   • Generates natural language responses

Features

• Document Upload

  • Support for PDF, TXT, and DOCX files
  • Raw text input option
  • Multiple file upload capability
  • Progress tracking

• Knowledge Base Management

  • Create and manage multiple knowledge bases
  • Add descriptions and metadata
  • Organize documents by topic or domain

• Knowledge Graph Visualization

  • Interactive graph display
  • Node and relationship exploration
  • Path visualization for query results

• Query Interface

  • Natural language querying
  • Model selection dropdown
  • Query history tracking
  • Detailed response visualization

Installation

1. Prerequisites

   • Docker and Docker Compose
   • Git
   • 8GB+ RAM recommended
   • NVIDIA GPU (optional, for improved LLM performance)

2. Clone the Repository

   git clone https://github.com/your-username/PathRAG.git
   cd PathRAG

3. Environment Setup

   # Start all services
   docker-compose up -d

4. Access the Application

   • Frontend: http://localhost
   • Backend API: http://localhost:8000
   • Weaviate Console: http://localhost:8080
   • Ollama API: http://localhost:11434

Architecture

Frontend (Port 80)

• Vite + React application
• Material-UI components
• Nginx for static file serving and API proxying
• Environment configuration via VITE_API_URL

Backend (Port 8000)

• FastAPI application
• Uvicorn ASGI server
• File processing and KG construction
• Path-based retrieval implementation

Weaviate (Port 8080)

• Vector database for KG storage
• RESTful and GraphQL APIs
• Configurable vectorizer modules
• Persistent data storage

Ollama (Port 11434)

• Local LLM service
• Multiple model support
• REST API for inference
• Model management

API Endpoints

Upload API

• POST /upload/files - Upload documents
• POST /upload/knowledge_base - Create knowledge base
• GET /query/kbs - List knowledge bases

Query API

• GET /query/models - List available models
• POST /query/query - Execute queries
• GET /query/graph - Retrieve graph data

Configuration

Docker Compose

version: '3.8'
services:
  frontend:
    build: ./frontend
    ports:
      - "80:80"
    environment:
      - VITE_API_URL=

  backend:
    build: ./backend
    expose:
      - "8000"
    environment:
      - WEAVIATE_URL=http://weaviate:8080
      - OLLAMA_URL=http://ollama:11434

  weaviate:
    image: semitechnologies/weaviate:1.24.1
    environment:
      - QUERY_DEFAULTS_LIMIT=20
      - AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true

  ollama:
    image: ollama/ollama:latest
    volumes:
      - ollama-models:/root/.ollama

Development

Frontend Development

cd frontend
npm install
npm run dev

Backend Development

cd backend
pip install -r requirements.txt
uvicorn main:app --reload

Contributing

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

License

[Add your license information here]

References

• PathRAG Paper
• Original PathRAG Repository