fix: complete vLLM tool_choice"auto" compatibility fix

[jilin6627-spec]

fix: complete vLLM tool_choice="auto" compatibility fix

Problem

vLLM OpenAI-compatible server returns 400 error when receiving tool_choice="auto", unless the server was started with --enable-auto-tool-choice --tool-call-parser <...>.

Original PR only fixed TokenRecordingModelWrapper, but there are three more code paths that still pass through tool_choice="auto" to the model client:

1. QwenPawReactAgent._reasoning() in src/qwenpaw/agents/react_agent.py
2. ToolGuardMixin._reasoning() in src/qwenpaw/agents/tool_guard_mixin.py
3. RoutingChatModel.__call__() in src/qwenpaw/agents/routing_chat_model.py

This causes the 400 error even with the original fix applied, because tool_choice="auto" still reaches the vLLM server through these other paths.

Solution

Apply the same fix to all three paths: when tool_choice == "auto", set it to None before passing to the underlying model client. This matches what was already done in TokenRecordingModelWrapper.

Testing

• Tested by building arm64 docker image and running QwenPaw connected to a standard vLLM server without --enable-auto-tool-choice
• Before this fix: 400 error on first query
• After this fix: query works normally, no 400 error

[github-actions]

Welcome to QwenPaw! 🐾

Hi @jilin6627-spec, this is your 2nd Pull Request.

📋 About PR Template

To help maintainers review your PR faster, please make sure to include:

• ✅ Description - What this PR does and why
• ✅ Type of Change - Bug fix / Feature / Breaking change / Documentation / Refactoring
• ✅ Component(s) Affected - Core / Console / Channels / Skills / CLI / Documentation / Tests / CI/CD / Scripts
• ✅ Checklist:
  • Run and pass pre-commit run --all-files
  • Run and pass relevant tests (pytest or as applicable)
  • Update documentation if needed
• ✅ Testing - How to test these changes
• ✅ Local Verification Evidence:

  pre-commit run --all-files
  # paste summary result
  
  pytest
  # paste summary result

Complete PR information helps speed up the review process. You can edit the PR description to add these details.

🙌 Join Developer Community

Thanks so much for your contribution! We'd love to invite you to join the official QwenPaw developer group! You can find the Discord and DingTalk group links under the "Developer Community" section on our docs page:
https://qwenpaw.agentscope.io/docs/community

We truly appreciate your enthusiasm—and look forward to your future contributions! 😊

We'll review your PR soon.

---

Tip

⭐ If you find QwenPaw useful, please give us a Star!

Staying ahead

Star QwenPaw on GitHub and be instantly notified of new releases.

Your star helps more developers discover this project! 🐾

[xieyxclack]

Fixed in #3295 ?

[xieyxclack]

Closed as invalid

[jilin6627-spec]

#3295 只改了一个文件 (src/qwenpaw/token_usage/model_wrapper.py)，但还有另外三个文件也需要修复： #3295 只修复了一个文件。#3295 只修复了 model_wrapper.py，但还有三个调用路径没有修复： 1. token_usage/model_wrapper.py - ✅ #3295 已修复 2. agents/react_agent.py (QwenPawAgent._reasoning) - ❌ 未修复 3. agents/tool_guard_mixin.py (_reasoning) - ❌ 未修复 4. agents/routing_chat_model.py (endpoint.model) - ❌ 未修复 我们的 PR #3438 补全了剩下三个文件的修复，这才是完整的。误以为全部修好了，但实际上只修了一处。

…

On Thu, Apr 16, 2026 at 7:10 PM Yuexiang XIE ***@***.***> wrote: *xieyxclack* left a comment (agentscope-ai/QwenPaw#3438) <#3438 (comment)> Closed as invalid — Reply to this email directly, view it on GitHub <#3438 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CABG3ARYOEEEBWF6NLYOOJ34WC5STAVCNFSM6AAAAACX2BSJPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DENJZGU2DGNBZGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>