Learning path · Production RAG · 57
GraphRAG
Combining knowledge graphs or community summaries with vector retrieval for global and relational questions over corpora.
Why it matters
- Helps "themes across the whole library" queries flat RAG misses.
- Surfaces entity relationships and community structure.
- Higher offline indexing cost than flat chunk indexes.
Key ideas
- Entity graphs
- Community summaries
- Global search
Top resources
- 01DocsMicrosoft
GraphRAG
Why this resource. Microsoft’s GraphRAG pipeline and query modes.
Covers in this concept
- community summaries
- global vs local search
- 02PaperMicrosoft Research
From Local to Global: A Graph RAG Approach to Query-Focused Summarization
Why this resource. The research write-up behind the library.
Covers in this concept
- knowledge graph
- query-focused summarization
GraphRAG extracts entities and relations, builds community summaries, and answers both local detail questions and global thematic ones. It shines on investigative workflows—connecting people, projects, and events across documents. Operate it as a batch indexing investment with clear query routing: use graph paths for synthesis questions, vector search for pinpoint fact lookup. Route queries explicitly—graph summaries for thematic questions, vector search for pinpoint facts—to avoid paying graph costs on every request.
Updated 2026-08-09 · Full learning path