[Feature]: supporting OpenRouter include_reasoning parameter

[jamesbraza]

The Feature

https://openrouter.ai/docs/parameters#include-reasoning

This opt-in flag include_reasoning from OpenRouter enables one to get reasoning traces out of a reasoning model hosted on OpenRouter, for example DeepSeek R1.

Motivation, pitch

Being able to get the "reasoning" when using OpenRouter.

Are you a ML Ops Team?

No

Twitter / LinkedIn details

No response

[jamesbraza]

I got it to work via this with openai==1.60.2:

client = OpenAI(
    base_url=os.getenv("OPENROUTER_API_URL", "https://openrouter.ai/api/v1"),
    api_key=os.environ["OPENROUTER_API_KEY"],
)
response = client.chat.completions.create(
    model="deepseek/deepseek-r1",
    messages=[{"role": "user", "content": "What is 1+1?"}],
    extra_body={"include_reasoning": True},
)
reply = response.choices[0].message

Or via curl:

curl -X POST https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-r1",
    "include_reasoning": true,
    "messages": [
      {"role": "user", "content": "What is 1+1?"}
    ]
  }'

I could not get it to work with litellm.completion

[krrishdholakia]

seems like a bug for passing provider-specific params, which litellm does support

@jamesbraza does it work via completion if you use extra_body?