[Feature]: Need support for /v2/rerank

[jeromeroussin]

The Feature

Cohere introduced its v2 API in September 2024. The langchain cohere rerank class enforces the use of v2 only: https://github.com/langchain-ai/langchain-cohere/blob/main/libs/cohere/langchain_cohere/rerank.py#L47

Litellm does not have a /v2/rerank route however. This RFE is about adding /v2/rerank to the current list of rerank endpoints (/v1/rerank, /rerank)

Motivation, pitch

Continued use of litellm as a proxy to cohere and cohere like models.

Are you a ML Ops Team?

No

Twitter / LinkedIn details

https://www.linkedin.com/in/jeromeroussin/

[ishaan-jaff]

Would you expect the provider route to be cohere/v2 or cohere/

[jeromeroussin]

Ideally this would work

from cohere import ClientV2
client = ClientV2(api_key=<key>
                base_url= <litellm gateway url>)
response = client.rerank(...)

just like this works today:

from cohere import Client
client = Client(api_key=<key>
                base_url= <litellm gateway url>)
response = client.rerank(...)

That means support for /v2/rerank just like /v1/rerank works.

Are you standardizing that endpoint after something other than Cohere?