Cognee

Builds a knowledge graph plus vector store for hybrid GraphRAG.

Cognee is an open-source Python framework for building memory layers for AI agents and LLM applications. It ingests data (documents, conversations, code) and constructs queryable knowledge graphs combined with vector embeddings, replacing ad-hoc RAG pipelines with a structured, persistent memory. It supports multiple graph and vector store backends and exposes a pipeline-based API for loading, cognifying, and searching data.