LanceDB is an open-source, embedded database for AI data that stores vectors alongside text, images, video and other media, and searches them by vector, keywords or SQL.
In plain words
Many vector databases run as a separate server. The open-source version of LanceDB runs inside your own program, like a library, and can store its data on your own disk or in object storage. It is built on Lance, an open columnar file format made for AI data such as images and video. You can store embeddings next to the original files and metadata, then search by vector similarity, full-text search or SQL. It also versions your data automatically and is aimed at model training as well as retrieval.
A quick example
Imagine a folder of product photos with captions. You store each photo, its caption and its embedding in one LanceDB table on your laptop. You can then search for similar photos by vector, find captions with a keyword, or filter with SQL, without running a database server.
Where you'll meet it
- Local or embedded vector search inside apps.
- Multimodal search across text, images and video.
- Managing training data for AI models.
Related
Sources
- Docslancedb/lancedb, LanceDB (GitHub) · read 28 Oct 2026
- DocsLanceDB Documentation, LanceDB · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 28 Oct 2026. Spotted an error? Report it.