love it. vector dbs are so bloated and unnecessarily expensive/complicated for very little return.

oh, promising.

can you upload benchmarks on the readme file with FAISS, pg_vector, and the like?