Enables force reasoning based on chat template for Qwen3-Thinking

[JustinTong0323]

Motivation

Adds a mechanism to enforce reasoning based on patterns detected in the chat template.

The changes introduce a force_reasoning flag to the TemplateManager and ReasoningParser. This flag is set if reasoning patterns are detected in the chat template, such as <think> tags.

This ensures that reasoning is enforced even when the user does not explicitly enable it.

Modifications

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[gemini-code-assist]

Summary of Changes

Hello @JustinTong0323, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request introduces a significant enhancement to how reasoning is handled within the system, particularly for models like Qwen3-Thinking. It establishes a new capability to automatically detect if a loaded chat template implies a need for reasoning (e.g., through specific tags like <think>) and, if so, enforces this reasoning behavior. This ensures that the model's output consistently includes reasoning steps when the template design dictates it, improving adherence to model-specific output formats without requiring explicit user configuration.

Highlights

• Automatic Reasoning Enforcement: Implemented a mechanism to automatically detect and enforce reasoning (e.g., via <think> tags) based on the chat template itself, ensuring reasoning is applied even if not explicitly enabled by the user. This is particularly relevant for models like Qwen3-Thinking.
• Template Manager Enhancements: Added a force_reasoning property and a _detect_reasoning_pattern method to the TemplateManager. This new method scans the loaded chat template for specific patterns (like <think>, </think>, or reasoning_content) to determine if reasoning should be implicitly enforced.
• Reasoning Parser Integration: Modified the ReasoningParser and its specific model detectors (including DeepSeekR1Detector, Qwen3Detector, and KimiDetector) to accept and utilize the new force_reasoning flag. This flag, propagated from the TemplateManager, dictates whether reasoning parsing should be active regardless of explicit user settings.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[zhyncs]

After #8363 do we still need this

[CatherineSue]

Thanks! Might be a better implementation than #8363. Can we integrate the UTs here?

[JustinTong0323]

After #8363 do we still need this在 #8363 之后，我们是否还需要这个

I talked with Qwen team and they think remaining one reasoning parser is better for users, and avoid confusion.

[zhyncs]

hold on please

[JustinTong0323]

At present, the behavior is consistent with that of the main branch. Users can choose to use either "qwen3" or "qwen3-thinking". Regarding the Qwen3-Thinking model, these two options exhibit the same behavior.

[CatherineSue]

Overall LGTM. Left two nit comments.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!