turbopuffer

turbopuffer is a serverless vector database that stores data on object storage (S3) with a memory/SSD caching layer in front. This architecture keeps costs low while handling billions of vectors. It supports vector search, BM25 full-text search, and hybrid queries combining both.

Namespaces scale independently to 500M+ documents each. You can pin frequently queried namespaces to cache for predictable sub-10ms latency. No infrastructure to manage: you write documents, turbopuffer handles indexing, caching, and storage tiering automatically.