Graph RAG

GraphRAG combines knowledge graphs with LLMs to enable structured reasoning and global document summarization across complex, private datasets.

Microsoft Research developed GraphRAG to solve the 'needle in a haystack' problem inherent in standard vector search. By constructing a structured knowledge graph from raw text (extracting entities like 'Project Silica' or 'Artemis') and clustering them into hierarchical communities, the system provides LLMs with a high-level map of the data. This architecture allows the model to answer complex, multi-document queries (e.g., 'What are the top three risks across all 500 project reports?') that traditional RAG often misses. It is the definitive choice for operators managing massive, interconnected document stores where global context matters as much as local retrieval.

https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/