datalevin

Why we included this project

Datalevin is an embedded database that speaks Datalog instead of SQL, which makes it a natural fit for developers who prefer Datalog's implicit joins and recursive rules to relational queries. It runs on LMDB with ACID guarantees, and you can either embed it in your own process or run it as a client-server database, with official bindings for Clojure, Java, Python, and Node. What makes it interesting is that relational, graph, document, full-text, and vector search all live under a single schema and query model, so one datastore can cover workloads that would otherwise require several specialized engines. The vector support is aimed at AI applications like semantic search and retrieval-augmented generation, and it ships with local embedding, text generation, and OCR providers plus MCP tooling. For teams that want one self-contained store instead of a stack of single-purpose components, this is worth a close look.