### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

I faced this issue over codiumai/pr-agent:latest where the model is my local model, for the interference backend I'm using vLLM

litellm includes stream_options: {"include_usage": True} in the request body even when the call is non-streaming. vLLM strictly rejects this with 400 — stream_options is only valid when stream=True.

This was silently hidden when using Ollama as backend (ignores unknown fields). Switching to vLLM exposed the bug.

### Steps to Reproduce

import litellm

response = litellm.completion( model="openai/my-model", messages=[{"role": "user", "content": "say hi"}], api_base="http://my-vllm-host:8000/v1", api_key="my-key", # stream not set / stream=False )

VLLM -> VLLMValidationError: Stream options can only be defined when `stream=True`.

Or over curl curl https://my-vllm-host:8000/v1/chat/completions \ -H "Authorization: Bearer my-key" \ -H "Content-Type: application/json" \ -d '{ "model": "my-model", "messages": [{"role": "user", "content": "say hi"}], "stream_options": {"include_usa…