awslabs/graphrag-toolkit

Python toolkit for building graph-enhanced GenAI applications

What it solves

This toolkit provides tools for building graph-enhanced Generative AI applications, specifically addressing the limitations of standard RAG by using structured knowledge graphs to improve the accuracy of question-answering systems.

How it works

The toolkit contains two primary components:

  1. Lexical Graph: A framework that automates the construction of a hierarchical lexical graph from unstructured data. It allows for the creation of question-answering strategies that query this graph to retrieve relevant information.
  2. BYOKG-RAG: A Knowledge Graph Question Answering (KGQA) approach that integrates LLMs with existing structured knowledge graphs, allowing users to "bring your own knowledge graph" (BYOKG) to perform complex queries.

Who it’s for

Developers building GenAI applications who want to move beyond simple vector search and implement multi-hop retrieval or complex question answering over structured data.

Highlights

  • Automated Graph Construction: Automates the creating of hierarchical lexical graphs from unstructured text.
  • Hybrid Queries: Supports combining vector search and graph search to enhance retrieval accuracy.
  • BYOKG Support: Allows the use of existing structured knowledge graphs for LLM-based question answering.
  • Multi-Hop Retrieval: Specifically designed to improve retrieval for complex queries that require connecting multiple pieces of information.

Related

  • Project
    docker/genai-stack

    A Docker-based starter stack for building GenAI applications, featuring integrated LLMs and Neo4j for RAG and knowledge graph capabilities.

  • Project
    HKUDS/LightRAG

    A lightweight, graph-based RAG framework that combines knowledge graphs and vector embeddings to provide deep contextual understanding and efficient incremental updates.

  • Project
    neo4j/neo4j-graphrag-python

    The official Neo4j Python package for building GraphRAG applications, enabling the automated construction of knowledge graphs and hybrid retrieval for LLMs.

  • Project
    InternScience/GraphGen

    A framework for generating knowledge-driven synthetic data using knowledge graphs to fill knowledge gaps in LLMs for improved supervised fine-tuning.

  • Project
    DEEP-PolyU/LinearRAG

    LinearRAG is a relation-free graph construction method for GraphRAG that eliminates LLM token costs during graph building to enable efficient, scalable multi-hop reasoning on large corpora.