Heya! There is a new kiddo around the block which is aiming to be faster than the current and only vector similarity search/storage extension.
Pg Vector is kindof showing some liabilites, which pgvecto.rs aims to fix…
I request for this to be added, hopefully expecting a positive response
Looking at the code of pgvecto.rs, it does not seem to be using Postgres’ buffer cache, page format, nor WAL system, and is thus fundamentally incompatible with Neon’s current architecture.
Could you share what liabilities?
I’d say the Vector Dimension limit is one of its most recognised liability.
More like transaction issues is the ones that come to mind