turbopuffer

Our first blog post is out "turbopuffer: fast search on object storage" https://lnkd.in/g3CGHhpK

A brief summary of what the turbopuffer team shipped in May '24: 🔍 BM25/Hybrid Search 📉 Up to 2x faster filtering for large namespaces (faster zero-copy bitmaps) 📉 Up to 2x faster queries on large namespaces with lots of attributes (zero-copy storage) 💰 Updated pricing to take attributes into account 🤖 Typescript client v0.5 with better connection pooling 📊 Dashboard and API support for prefix filtering of namespaces 📊 Named API Keys 🤖 Python Client async support with higher throughput See the rest here https://lnkd.in/eTVNvKs5

Learn more about our architecture https://lnkd.in/g278AdCB

Vector search might not seem that hard, but it's hard to make it fast with filters, high recall, and good cold start on object storage... Bojan Serafimov pulled off a monumental effort on the foundational query planner for this 🚀

And since we're in the mood for these launch announcements: you can now list Namespaces 📋✨🔥 https://lnkd.in/eX3eZAVg

but wait, there's more from Bojan Serafimov ... AND and OR filters for attributes🔥 https://lnkd.in/enPT5BUh

li'l API ergonomics from the newest member of our team, welcome Bojan Serafimov 🙌

We're excited to launch our *official* vector database integration with... turbopuffer! 🐡 Any customer can bring their turbopuffer vector database to use with Carbon. Carbon automatically synchronizes content from user data sources (Google Drive, Notion, etc) and embeddings in your turbopuffer vector database. It was a pleasure working with Simon and Justin (fellow Canadians 🇨🇦) to build out this integration! 🚀

After a nice refactor that'll open up for many types down the line, we've got string id support ✨

I'm faster. Cold latency down 2x 💫 https://lnkd.in/e4-7bSFk