add ChatTemplateKwargs to ChatCompletionRequest

[justa-cai]

Describe the change
add chat_template_kwargs param support

Provide OpenAI documentation link
vllm qwen3 thinking modes

curl http://localhost:8000/v1/chat/completions -H "Content-Type: application/json" -d '{
  "model": "Qwen/Qwen3-8B",
  "messages": [
    {"role": "user", "content": "Give me a short introduction to large language models."}
  ],
  "temperature": 0.7,
  "top_p": 0.8,
  "top_k": 20,
  "max_tokens": 8192,
  "presence_penalty": 1.5,
  "chat_template_kwargs": {"enable_thinking": false}
}' 

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.93%. Comparing base (774fc9d) to head (9ed7474).
Report is 104 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #980       +/-   ##
===========================================
- Coverage   98.46%   85.93%   -12.53%     
===========================================
  Files          24       43       +19     
  Lines        1364     2268      +904     
===========================================
+ Hits         1343     1949      +606     
- Misses         15      300      +285     
- Partials        6       19       +13     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull Request Overview

This PR adds support for a new parameter "chat_template_kwargs" to the ChatCompletionRequest structure to allow for non-standard parameters in chat completions.

• Added the ChatTemplateKwargs field to ChatCompletionRequest to support extra configuration parameters (e.g., think mode for Qwen3).

[Copilot]

[nitpick] Consider updating this comment to reference 'ChatTemplateKwargs' to match the field name and improve grammar by capitalizing 'such' to 'Such' to enhance clarity.

Suggested change

	// ExtraBody provides a way to add non-standard parameters to the request body. such as think mode for qwen3
	// ChatTemplateKwargs provides a way to add non-standard parameters to the request body. Such as think mode for qwen3.

[sashabaranov]

Suggested change

	ChatTemplateKwargs map[string]interface{} `json:"chat_template_kwargs,omitempty"`
	ChatTemplateKwargs map[string]any `json:"chat_template_kwargs,omitempty"`

[torrischen]

@sashabaranov
Hi, I wonder if the pr will be merged soon?
Our project is using this repo and currently we are using qwen3, we urgently need the feature to support switching between thinking and non-thinking

[sashabaranov]

@justa-cai @torrischen there are still a few pending comments and I'd also add that we should explicitly name the variable so it would be obvious that the parameter is not related to OpenAI first-party API. I'd be happy to merge after all this is fixed 🫶

[justa-cai]

#ChatTemplateKwargs
Additional kwargs to pass to the template renderer. Will be accessible by the chat template.
VLLM

[justa-cai]

@sashabaranov help meger

[sashabaranov]

Thank you for update!