langchain-ai/langchain

Pinecone: Support starter tier

StankoKuveljic opened this issue ยท 6 comments

Feature request

Adapt the pinecone vectorstore to support upcoming starter tier. The changes are related to removing namespaces and delete by metadata feature.

Motivation

Indexes in upcoming Pinecone V4 won't support:

  • namespaces
  • configure_index()
  • delete by metadata
  • describe_index() with metadata filtering
  • metadata_config parameter to create_index()
  • delete() with the deleteAll parameter

Your contribution

I'll do it.

@StankoKuveljic @jamescalam
Can you please point me to the v4 spec documentation on pinecone.
With support for namespace removed, does the free tier restrict users to a single vector space? Using metadata filtering may be one way to bypass this limitation.

@StankoKuveljic - I searched, but can't find any information about "upcoming V4" of Pinecone. Can you point us to documentation about v4 and removing namespaces?

Indexes in upcoming Pinecone V4 won't support:
namespaces

Is this information available to the public. I've not seen any communication on pinecone removing namespaces?

I'm sorry for the confusion.
This is related to upcoming starter tier https://docs.pinecone.io/docs/projects#gcp-starter-environment-public-preview it won't affect standard and enterprise tiers. The commit is reverted and I'll reopen an issue if there are breaking changes in examples and tests that won't support starter tier.

@StankoKuveljic Sounds good! If we are using namespaces, is it safe to unpin the langchain version?

@david-engelmann
The commit is reverted 2babe30 so now everything should be as before.

However, keep in mind that if you are using free tier gcp-starter environment you won't be able to use namespaces alongside with other limitations. Also, you will have issues running integration tests for Pinecone.
The integration test is probably going to change soon to allow contributors running it with free gcp-starter.