Michele LaurelliBlog

GraphRAG

GraphRAG: A RAG approach that integrates graphs of entities and relations to retrieve context and synthesize connections. The graph can guide local search, communities and global summaries. Building and maintaining it costs more than simple chunking and requires validation of entity and relation extraction.

In italiano: GraphRAGLegge I · Embedding, feature e spazi vettoriali

The graph can guide local search, communities and global summaries. Building and maintaining it costs more than simple chunking and requires validation of entity and relation extraction.

Examples

  • Map of people and organizations
  • Summaries of topical communities
  • Multi-hop relationship queries