LanceDB
Embedded vector database built on a columnar format for multimodal data.
Aperçu
LanceDB stores vectors alongside the data they describe in a columnar format designed for versioned, on-disk datasets, and runs embedded rather than as a server. It solves the problem that embeddings usually point at something - an image, an audio clip, a row - and keeping the pointer in a different system means both have to be kept in step. It exists because for local and multimodal workloads, co-locating them is simpler than joining them.
Graphe des technologies
Comment LanceDB se rattache au reste de la pile.
Utilise et dépend de
- Embedding concept Techniques LanceDB uses Embedding High ✓ source
Appliqué à
- Multimodal AI application Multimodal LanceDB used for Multimodal AI High ✓ source