[Feature]: Supports the volcengine new thinking mode for doubao-seed-1.6-250615

[gumutianqi]

The Feature

volcengine new model doubao-seed-1.6 add a new params "thinking": {"type": "xxx"}.
Links:

• https://www.volcengine.com/docs/82379/1593702#%E4%BD%BF%E7%94%A8%E6%96%87%E6%A1%A3

Motivation, pitch

I am working on AGI company as a Devoloper

LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?

Yes

Twitter / LinkedIn details

No response

[gumutianqi]

@krrishdholakia Thanks for picked this issue, I have a new test case for this feature.

I test the version v1.73.0.rc.1 with model doubao-seed-1-6, but this issue is not work with litellm proxy mode.

model_list:
  - model_name: doubao/doubao-seed-1-6
    litellm_params:
      model: volcengine/doubao-seed-1-6-250615
      api_key: os.environ/VOLCENGINE_API_KEY

> /v1/chat/completions reques

{
    "model": "doubao/doubao-seed-1-6",
    "messages": [
        {
            "role": "system",
            "content": "**Tell me your model detail information.**"
        }
    ],
    "user": "guest",
    "stream": true,
    "thinking": {
        "type": "disabled"
    }
}

I got this excepitons.

litellm.InternalServerError: InternalServerError: VolcengineException - AsyncCompletions.create() got an unexpected keyword argument 'thinking'. Received Model Group=doubao/doubao-seed-1-6\nAvailable Model Group Fallbacks=None