LanceDB
An open-source vector database that runs locally or in your cloud, with multimodal storage, vector and full-text search, and automatic versioning.
LanceDB is a vector database for developers building AI retrieval applications and teams working with multimodal datasets. Its embedded library can run locally or in your own cloud.
- Multimodal tables: Store text, images, videos and point clouds alongside embeddings and metadata, so searchable vectors and their original content share one table.
- Search and filtering: Use vector similarity and full-text search, with SQL queries and filters; the broader platform also supports hybrid search.
- Data versions: Automatic versioning lets you manage dataset changes without separate versioning infrastructure.
- Application integration: Python, TypeScript and Rust SDKs connect the database to your applications, and GPU support is available for building vector indexes.
The embedded library is open source under Apache 2.0. LanceDB also offers Cloud and Enterprise products with managed infrastructure; those are distinct from the local library.
The broader platform covers dataset curation, feature engineering and training storage. It supports deduplication and finding edge cases for labeling, Python feature pipelines with automatic embedding updates, and adding columns without rewriting existing data. Training storage provides random access and global shuffling to feed GPUs.
Logotyper
