Community Discussions

At some point embedded [LanceDB](https://saaskart.co/software/lancedb) may not be enough. What signals told you it was time to move to LanceDB Cloud?

How fast is index build and query in [LanceDB](https://saaskart.co/software/lancedb) for mid-to-large collections, and what tuning helped recall and speed?

Wiring [LanceDB](https://saaskart.co/software/lancedb) into a RAG framework should be simple given the integrations. Any gotchas with filtering or persistence?

The versioning built into the Lance format is unusual for a vector DB. How are teams using [LanceDB](https://saaskart.co/software/lancedb) versioning for reprod

[LanceDB](https://saaskart.co/software/lancedb) is lightweight, but how far does it scale for production RAG before you consider LanceDB Cloud or another store?

Storing images alongside vectors in [LanceDB](https://saaskart.co/software/lancedb) is handy for training pipelines. How are people using this in practice?

Both [LanceDB](https://saaskart.co/software/lancedb) and Chroma can run embedded. Which gave better performance and a nicer developer experience for your app?

Beyond vector search, the Lance format under [LanceDB](https://saaskart.co/software/lancedb) targets large ML datasets. Has anyone adopted it for training data

Storing vectors directly on S3 with [LanceDB](https://saaskart.co/software/lancedb) can cut costs. How is query latency when reading from object storage versus

Running [LanceDB](https://saaskart.co/software/lancedb) in-process with no server is appealing for simplicity. For which applications has the embedded model wor