Overview
GraphRAG augments Retrieval-Augmented Generation by building a knowledge graph over the source corpus, clustering entities into communities, and generating hierarchical summaries that feed the LLM at query time.
Key Capabilities
- Entity & relationship extraction into a graph
- Leiden-based community detection with auto-summaries
- Global, local and DRIFT search modes for nuanced retrieval
- Prompt-tuning guide and modular CLI / Python APIs
- Designed for complex reasoning over private datasets
