The paper introduces a Graph RAG approach to query-focused summarization (QFS) over private text corpora, addressing the limitations of retrieval-augmented generation (RAG) in handling global questions directed at entire text corpora. The proposed method combines the strengths of RAG and QFS by using a large language model (LLM) to build a graph-based text index in two stages: first, deriving an entity knowledge graph from the source documents, and then pregenerating community summaries for closely-related entities. For a given question, each community summary is used to generate a partial response, which are then summarized into a final response. The approach is evaluated on two real-world datasets, showing significant improvements over a naive RAG baseline in terms of comprehensiveness and diversity of generated answers. An open-source, Python-based implementation of both global and local Graph RAG approaches is also announced.The paper introduces a Graph RAG approach to query-focused summarization (QFS) over private text corpora, addressing the limitations of retrieval-augmented generation (RAG) in handling global questions directed at entire text corpora. The proposed method combines the strengths of RAG and QFS by using a large language model (LLM) to build a graph-based text index in two stages: first, deriving an entity knowledge graph from the source documents, and then pregenerating community summaries for closely-related entities. For a given question, each community summary is used to generate a partial response, which are then summarized into a final response. The approach is evaluated on two real-world datasets, showing significant improvements over a naive RAG baseline in terms of comprehensiveness and diversity of generated answers. An open-source, Python-based implementation of both global and local Graph RAG approaches is also announced.