Any thoughts here

You can define a GPTPineconeIndex with a blank list of documents but with your existing pinecone index

How does that work? what exactly is the use of documents then?

Also, is this specific to Pinecone or does it work with all different indexes?

I think it'll be less confusing for consumers if the documents array isn't required then, that way, it's less clear that we don't need it to query.

yeah, it's a good point. this seems to be a common use case that's popping up, let me take a look at resolving this!