Cognee

An open-source **persistent agent-memory framework** that builds a hybrid graph-plus-vector memory layer for LLM agents. Cognee ingests unstructured data (documents, conversations, tool outputs), automatically extracts entities and relationships into a knowledge graph (NetworkX, Neo4j, KuzuDB), and dual-indexes the same content into a vector store (LanceDB, Qdrant, Weaviate, Milvus) — giving agents both relational reasoning (graph traversal) and semantic retrieval (similarity search) from a single ingest pipeline.

Pure-vector memory loses the ability to answer questions that require *connecting* entities ("which of these documents involve the same legal entity?"). Pure-graph memory loses semantic similarity ("find documents that *feel* like this one even without exact entity overlap"). Most enterprise memory needs are both. Cognee productizes the dual-index pattern so developers don't hand-roll the entity-extraction + graph-construction + vector-embedding pipeline themselves.

Document-heavy agentic workflows (legal research, due diligence, medical record synthesis) where both relational queries and semantic retrieval matter; multi-source data ingestion pipelines where entity resolution across sources is the differentiator; building Animesis-CMA-style constitutional memory layers on top (Cognee provides the substrate; CMA provides the governance).