LanceDB

Serverless embedded vector DB for multimodal AI with hybrid search.

LanceDB is an open-source, embedded multimodal vector database built on the columnar Lance file format, designed to run in-process without a separate server. It combines vector similarity search with full-text and SQL-style filtering, supports versioning and zero-copy access, and scales to billions of vectors across text, image, and video data. It is notable for being serverless/embeddable while still handling large-scale retrieval.