This project is an open-source Retrieve-Augment-Generate (RAG) system based on knowledge graphs by Microsoft. It uses a large language model to generate knowledge graph data, transforming unstructured text into knowledge graph data with labels, thereby enhancing the output results of Large Language Models (LLMs). Compared to the baseline RAG, which relies on vector similarity, GraphRAG, which is based on knowledge graphs, performs better in answering more abstract (relational) and summarizing questions.