LanceDB Blog
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications

OpenVINO quantization cut CLIP processing time for LanceDB by 75.4% on CPU
LanceDB shows how to speed up image search without a GPU: converting CLIP to OpenVINO saves 43% of the time, and 8-bit INT8 quantization saves 75.4%.

LanceDB: The IVF_PQ Index Compresses RAG Vectors 64x and Speeds Up Search
Embeddings eat memory in RAG systems. LanceDB explains the IVF_PQ index: a 4096-bit vector compresses to 64 bits, a 64x reduction, and search only scans the nearest cells.

Lance Adds Git-Style Dataset Branches: Experiments Stay Out of Production's Way
Lance, the data format, now supports branches, tags and shallow cloning, so dataset experiments run in a separate folder without breaking production. ByteDance and Netflix asked for the feature.

LanceDB Ships Realtime Engine: Fresh Data Searchable in 52 ms
LanceDB has added streaming writes with live indexes to Enterprise: a row is already durably stored in S3, and on average it appears in search after 52 milliseconds.

Second Dinner Turns Card Prototypes Around in Hours, and Says AI Tests Beat Manual Ones in 81% of Cases
The studio behind Hearthstone wired a LanceDB-backed RAG system into a Slack bot. Card mockups now take hours, and the studio says AI-generated tests beat hand-written ones in 81% of cases.

WeRide Uses LanceDB to Cut Search for Rare Dangerous Driving Footage from a Week to an Hour
Self-driving cars need rare, dangerous footage to train on. WeRide found it among billions of records with the LanceDB vector database: the search now takes 1 hour instead of 1 week.