Hi. Would filtering on metadata when doing a query makes the response time faster (as there are less data) or does it actually make it slower?
The answer, as is usual, is “it depends.” In most cases the response time will be a bit faster for exactly why you listed, the query itself is being run on a smaller subset of the whole corpus. But there is a performance cost to performing the metadata query itself, and if the resulting subset isn’t small enough to make a difference in vector query time you might see a slight increase in latency.
We recommend testing with your data to be sure.