Knowledge Graph RAG System

A comprehensive implementation of Retrieval-Augmented Generation (RAG) systems using knowledge graphs and Neo4j. This repository demonstrates how to build, query, and leverage knowledge graphs for enhanced AI applications.

Overview

This project showcases different approaches to building and utilizing knowledge graphs for RAG systems:

1. Simple Knowledge Graph - Basic Neo4j implementation for creating and querying knowledge graphs
2. Knowledge Graph RAG - Advanced RAG system using knowledge graphs for structured information retrieval
3. Healthcare Knowledge Graph - Domain-specific implementation for healthcare data
4. Text Preparation for RAG - Tools for preparing and processing text data for RAG systems

Features

• Neo4j integration for knowledge graph storage and querying
• LangChain integration for RAG pipeline implementation
• Entity extraction and relationship mapping
• Full-text search capabilities
• Hybrid retrieval combining structured and unstructured data
• Conversation history handling for contextual responses
• Domain-specific implementations (healthcare, Roman Empire)

Project Structure

knowledge-graph-rag/
├── simple_kg/                  # Basic knowledge graph implementation
│   └── kg_simple.py            # Simple Neo4j operations
├── kgraph_rag/                 # Advanced RAG with knowledge graphs
│   └── roman_emp_graph_rag.py  # Roman Empire knowledge graph RAG
├── prep_text_for_rag/          # Text preparation tools
│   └── app.py                  # Text processing application
├── healthcare/                 # Healthcare domain implementation
│   ├── health_care_kg.py       # Healthcare knowledge graph
│   ├── health_care_langchain.py # Healthcare RAG with LangChain
│   └── healthcare.csv          # Sample healthcare data
├── requirements.txt            # Project dependencies
└── .env                        # Environment variables (not tracked in git)

Prerequisites

• Python 3.8+
• Neo4j Aura instance or local Neo4j database
• OpenAI API key

Installation

1. Clone the repository:

   git clone https://github.com/Badribn0612/knowledge-graph-rag.git
   cd knowledge-graph-rag

2. Install dependencies:

   pip install -r requirements.txt

3. Create a .env file with your credentials:

   NEO4J_URI=your_neo4j_uri
   NEO4J_USERNAME=your_neo4j_username
   NEO4J_PASSWORD=your_neo4j_password
   AURA_INSTANCENAME=your_aura_instance_name
   OPENAI_API_KEY=your_openai_api_key
   

Usage

Simple Knowledge Graph

from simple_kg.kg_simple import connect_and_query

# Connect to Neo4j and run a simple query
connect_and_query()

Knowledge Graph RAG

from kgraph_rag.roman_emp_graph_rag import chain

# Ask a question about the Roman Empire
response = chain.invoke({
    "question": "Who was the first Roman emperor?",
})

print(response)

Healthcare Knowledge Graph

from healthcare.health_care_langchain import health_care_chain

# Ask a question about healthcare
response = health_care_chain.invoke({
    "question": "What are the symptoms of diabetes?",
})

print(response)

Key Components

Knowledge Graph Creation

The system supports multiple approaches to knowledge graph creation:

• Manual entity and relationship creation
• Automated extraction from text using LLMs
• Import from structured data (CSV)

RAG Implementation

The RAG system combines:

• Structured data retrieval from the knowledge graph
• Unstructured data retrieval using vector search
• Context-aware question answering with conversation history

Entity Extraction

The system uses LLMs to extract entities from text, which are then stored in the knowledge graph for future retrieval.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgments

• Neo4j for the graph database
• LangChain for the RAG framework
• OpenAI for the language models