Hey team, Pinecone integration is awesome! Thanks a ton for giving us this functionality. As of this writing, there are 2 primary use cases for pinecone relative to storage, one for vectors and the other for documents. Assistants (as you have integrated) can be set up around stored documents and queried. However indexes, when data is upserted, can be used slightly more generically to query indexes based on semantics, lexical, and other query types to find vector similarity.
I have found Glide’s integration limited in its ability to work with indexes, because you cannot store metadata in the request to upsert data to an index in Pinecone. This means that you cannot pre-filter results in a query (metadata filters also not supported in Glide), which is a powerful way to only get back similarity vectors that might be expected by a user.
I’d love to see metadata support in Glide, even if it requires us to form a JSON object to send along with the request during upsert, update, and query.
Thanks a ton!