tembo-io / pg_vectorize

The simplest way to build AI workloads on Postgres
https://tembo.io/pg_vectorize/
739 stars 33 forks source link

Add embedding provider for VoyageAI. #174

Closed palash25 closed 1 week ago

palash25 commented 3 weeks ago

/fixes #152 /claim #152

I was having some trouble with setting up the codebase locally, will look at it tomorrow. Till then I have tested just this code in isolation and the integration test works. Someone will need to a Voyage API key to the CI secrets for it pass though.

Will test it once again tomorrow with the full setup and open it up for review, although if anyone wants to skim through it till then, feel free to do so.

ChuckHend commented 2 weeks ago

@palash25 anything left here? I ran some tests and it seems good.

palash25 commented 1 week ago

Hi @ChuckHend sorry for the delayed response, I couldn't test this locally last week because of some trouble in setting up the codebase locally and then i was out sick.

anyway I tested it now with the following query and it returned embeddings. I am assuming this is the way to test new providers, let me know if I am missing something

select vectorize.transform_embeddings(
    input       => 'the quick brown fox jumped over the lazy dogs',
    model_name  => 'voyage/voyage-3-lite'
);

I will make the one change you requested regarding model names and fix the warnings the push in a few minutes then we can merge the PR.

palash25 commented 1 week ago

i am assuming the CI failures are unrelated