[visarga]

up to 100k rows you don't get faster by using vector store, just use numpy

[intalentive]

And often you have tags that filter it down even further.