PgVectorscale annoucement

[cometkim]

Timescale team just announced a new extension "pgvectorscale" which supports a new index built with pgrx, using pgvector's data type

• Explainer: https://www.timescale.com/blog/pgvector-is-now-as-fast-as-pinecone-at-75-less-cost/
• Repo: https://github.com/timescale/pgvectorscale

They claim big improvements in most benchmarks. I hope we can try it out on pradedb too.

[philippemnoel]

You'd like for this extension to be released within the ParadeDB Dockerfile? Ideally this should be contributed upstream to pgvector.

[cometkim]

If I understand correctly, users just need to CREATE INDEX over USING diskann instead of USING hnsw if the extension exists.

It interoperates without replacing anything, doesn't it make sense to offer it as an add-on?

[philippemnoel]

Yes, you're right. We're adding it here: #1262