Mark Pollack

Merged in d5b8123e60411f5d4a2cd5ee9eaf31f92a72c88c made a few minor changes. Note that `BeanOutputParser` is deprecated, so the tests should use `BeanOutputConverter` and other similar `OutputConverter` classes. The api keys you sent worked!

This seems to be as simple as ensuring that the String IDs you use for Documents are supported by your vector store of choice. A way to fail fast would...

The tests pass when running individually, but fail as a group. e.g. for `embed()` ``` 2024-10-23T18:45:59.256-04:00 DEBUG 627946 --- [ main] o.s.jdbc.core.JdbcTemplate : Executing SQL statement [DROP EXTENSION IF EXISTS...

Thanks for the contribution @eddumelendez !!

closing as the PR has been merged

I would be inclined to remove it. This comes from the very early says where I was reviewing langchain and llamaindex and langchain had this design. It is a mistake...

Yes, this code has already been updated in main. Thanks in anycase!!

I've cleaned up some odds and ends and it now merged in main via 067a33dbe2b3fa24a4016c86464b80ef6cfbae1b Thanks and sorry for the delay. I will close the other draft PR as well.

Ok, understood, moving to M2 for now to support whatever other similarity algorithms that pinecone has. BTW, it is valid to have negative numbers for cosine similarity

After thinking about it, is also likely that a v2 of the endpoint would have a different api so a different implementation class would be needed and that is where...