Seekdb

SeekDB is an open source, AI native search database that keeps relational, vector, full text, JSON and GIS data in a single engine. Its purpose is to let developers build retrieval and RAG applications without stitching together a separate vector store, search index and SQL database. The main capability is hybrid search. A single SQL statement can combine vector based semantic search with keyword search, and results can be reranked with weighted scores, Reciprocal Rank Fusion or an LLM based reranker. Relational filters are pushed down to storage, and multi table joins are supported. The database supports dense and sparse vectors with several distance metrics, including Manhattan, Euclidean, inner product and cosine. It can also run embedding, reranking, LLM inference and prompt management inside the database, supporting a document in, data out RAG workflow. Underneath, it is built on the OceanBase engine, which the vendor says provides real time writes and queries, full ACID compliance and compatibility with the MySQL ecosystem. Developers can use it through a Python client installed with pip, in embedded mode or in client and server mode. The home page also announces a release with lower resource overhead and a variant aimed at agents, MCP and RAG. It suits developers and teams building AI applications who want one database for structured and unstructured data. The software is open source with code on GitHub, and the site does not publish pricing.