🐛 [Lang] Cohere: Streaming Workflow is not working

[roma-glushko]

I have tried to test Cohere streaming workflow and faced error like:

After looking into Cohere documentation and some other LLM clients out there, it seems like Cohere is not using SSE protocol to stream chat chunks, they send serialised JSON chunks per stream line:

https://github.com/BerriAI/litellm/blob/main/litellm/llms/cohere.py#L178-L191

We need to fix this issue to make the stream workflow working.

[roma-glushko]

@mkrueger12 let's talk about this issue in discord