Datalevin

Datalevin's cost-based query optimizer makes complex joins up to 4x faster than SQLite and 2.4x faster than PostgreSQL, as demonstrated in the JOB benchmark, due to better plan generation.

It unifies key-value storage for EDN data, a document database with automatic path indexing, and a vector database with SIMD-accelerated search, reducing the need for multiple specialized databases.

Built-in MCP server, in-database embedding, OCR, and text generation via llama.cpp enable RAG applications without external services, as highlighted in the features section.

Can be embedded like SQLite, run in client/server mode with RBAC, or used as a Babashka pod for scripting, offering adaptability across different use cases.