Microsoft has launched GraphRAG, a graph-based approach to Retrieval-Augmented Generation (RAG), on GitHub. This new tool aims to significantly improve question-answering over private or previously unseen datasets by using a hierarchical approach. GraphRAG combines pre-trained language models with external knowledge retrieval to enhance the accuracy and relevance of generated outputs. It outperforms naive RAG on comprehensiveness and diversity and performs better than source text summarization. The project has been open-sourced to encourage wider adoption and innovation within the AI community. GraphRAG is built on LanceDB and has been described as a structured and hierarchical approach to RAG, as opposed to naive semantic-search approaches using plain text. The tool integrates with Neo4j and LangChainAI, and has been developed with contributions from Microsoft Research.