We will no longer be actively supporting KuzuDB (kuzudb.com)

The core Kuzu DB maintainers have announced they will no longer support the project. Kuzu is an MIT‑licensed, open‑source, embedded graph database designed to run in‑process (no external server) with on‑disk or in‑memory storage. It distinguished itself with columnar storage, vectorized processing, novel join algorithms, Cypher support for the property‑graph model, a graph‑native full‑text search index, an HNSW vector index, and an algorithms package. It also offered ready integrations with AI and data tooling such as LangChain, LlamaIndex, PyTorch Geometric, Pandas, Parquet and Iceberg, making it attractive for knowledge‑graph and vector‑embedding workflows. The announcement matters because teams that embedded Kuzu into AI/ML pipelines—especially for low-latency, in‑process knowledge graphs, vector search, or Cypher‑based analytics—now face maintenance, security and compatibility risks. The embedded nature complicates migration: you can’t just swap a remote vector DB without code changes. Practical next steps are to freeze and audit current deployments, export data (Parquet/CSV or graph dumps), evaluate community forks or take‑over, or migrate connectors to actively maintained alternatives (e.g., Neo4j/Memgraph for graph queries, Weaviate/Milvus for vector search). Because Kuzu is MIT‑licensed, the community can fork and continue development, but teams should plan for immediate operational contingencies and longer‑term migration or stewardship strategies.