LanceDB is an open-source (Apache 2.0) serverless vector database using the Lance columnar format—designed for zero-copy integration with ML frameworks. Free and self-hosted; LanceDB Cloud in development. The Lance format enables direct access from Python, JavaScript, and ML tools without data copying or serialization overhead. Embedded mode runs in-process (like SQLite for vectors) with no server deployment required. Cloud mode (coming) will provide managed serverless experience. Features include multi-modal support (images, text, video), automatic data versioning, and full-text search alongside vector search. IVF-PQ indexing provides memory-efficient approximate search. For text-to-SQL applications: embed query examples and schema in Lance format accessible directly from Python data pipelines. Best suited for data science workflows, ML pipelines, and applications wanting embedded vector capability without operational overhead. Page should cover: Lance format advantages, embedded vs server deployment, multi-modal capabilities, versioning features, comparison with Chroma and SQLite-based solutions, and ML pipeline integration patterns.