GraphRAG is retrieval-augmented generation built on a knowledge graph: instead of (or alongside) similarity search over text chunks, the system retrieves entiti
Retrieval-augmented generation built on a knowledge graph: instead of, or alongside, similarity search over text chunks, the system retrieves entities and their relationships, enabling answers that require connecting facts, ownership chains, dependencies, or multi-hop reasoning.
For relationship-heavy enterprise domains like legal entities, supply chains, and org knowledge, where pure vector search loses the structure that matters. The cost is building and maintaining the graph, increasingly automated by LLM extraction.