GraphRAG

GraphRAG is a Python-based framework designed to build, manage, and query Graph Retrieval-Augmented Generation (GraphRAG) systems. It combines advanced graph processing tools like Neo4j and NetworkX with natural language capabilities powered by LLMs (Large Language Models) to create, visualize, and query knowledge graphs.

Features

Knowledge Graph Construction: Extract entities and relationships from text and construct a knowledge graph.
Graph Management:
- Store and query graphs in Neo4j.
- Perform analysis and operations on NetworkX graphs.
Graph Summarization: Analyze and summarize communities within the graph.
Natural Language Interaction: Generate answers to queries using the knowledge graph and language models.
Interactive Visualization: Explore and interact with knowledge graphs using Gradio.
Modular Design: A highly organized and reusable codebase for scalable development.

File Structure

├── graphRAG # Main project directory

├── init.py #Marks the directory as a Python package 

├── main.py # Entry point to run the GraphRAG pipeline 

├── community_summaries.pkl # Serialized summaries of graph communities 

├── entities_extraction.py # Extract entities and relationships from text 

├── generating_answers.py # Generate answers to queries using the graph 

├── get_communities.py # Identify and analyze graph communities 

├── gradio_viz.py # Gradio-based interactive graph visualization 

├── graph_builder.py # Build the knowledge graph 

├── graph_nx.py # Manage and analyze NetworkX-based graphs 

├── KG_classes.py # Defines classes for knowledge graph objects 

├── subgraphs.py # Extract and manage subgraphs 

├── utils.py # Shared utility functions 

├── graph.gpickle # Serialized NetworkX graph file 

├── README.md # Project documentation (this file)