PG Search and PGVector Feature Request

[Suraj Mirpuri]

Hi folks good morning,

I love this tool and would love to make a contribution to enhance some aspects. Mainly, I want to know if the scope of your gem is limited to full text search; or would you be open to allowing for integrating PGVector dense vectors potentially in your "features/" list, to allow for semantic search?

The use case is that one can do hybrid search between dense vector methods already, using an MIT licensed gem like Andrew Kane's Neighbor gem - https://github.com/ankane/neighbor; but one can also hybridize sparse vectors to get semantic and keyword search benefits if this were integrated into PGSearch.

It would also solve some problems firms using AI and embeddings face, such as using multi search, but being able to use it semantically also.

I feel comfortable working on this and opening a PR when I have something that fits your organization of your gem, and with tests that work, etc - but I wanted to ask what your opinions were on the above, and if such a thing is a good fit for this gem, before I embark upon that.

Thank you so much for your time and consideration!

Regards,

Suraj Mirpuri