Graph RAG overview

Graph RAG as a cross-cutting topic — combining knowledge graphs with retrieval-augmented generation for multi-hop reasoning over structured and unstructured data. Covers Microsoft's GraphRAG pattern (community detection, summarization hierarchies, global search), entity-relation extraction from documents into graph form, graph traversal as retrieval, and the advantage over flat vector RAG for questions requiring synthesis across multiple documents or reasoning about relationships between entities. Graph RAG excels at "what is the overall theme?" queries that flat retrieval cannot answer.