3 ms·
the solution described in the blog post is currently in production at 100B vectors
by _peregrine_ 8mo ago
the solution described in the blog post is currently in production at 100B vectors
- rahimnathwani 8mo agoFor what/who?
- esafak 8mo agohttps://turbopuffer.com/customers/cursor https://turbopuffer.com/customers/cursor
- _peregrine_ 8mo agothis is actually not how cursor uses turbopuffer, as they index per codebase and thus need many mid-sizes indexes as opposed to one massive index as this post describes
- _peregrine_ 8mo agounfortunately i'm not able to share the customer or use case :( but the metrics that you see in the first charts in the post are from a production cluster