[ENHANCEMENT] new version have no Tool protocol setting

[lftx1986]

Problem (one or two sentences)

The previous version of Roo code had a tool protocol setting in the provider's advanced configuration, but the new version no longer includes it. This causes errors when using it.

Context (who is affected and when)

When I use Roo Code, it throws the following error:
Roo says
[Tool Use: list_files]
Model response incomplete
The model did not use any tools in its response. This usually happens when the model only provides text/reasoning without invoking the tools required to complete the task.

Desired behavior (conceptual, not technical)

Provide a solution

Constraints / preferences (optional)

No response

Request checklist

• I've searched existing Issues and Discussions for duplicates
• This describes a specific problem with clear context and impact

Roo Code Task Links (optional)

No response

Acceptance criteria (optional)

No response

Proposed approach (optional)

No response

Trade-offs / risks (optional)

No response

[roomote]

I'd like to help with this issue, but I need some clarification to ensure I implement the right solution. Could you please provide more details on the following:

1. Which API provider and model are you using? (e.g., OpenAI/GPT-4, Anthropic/Claude, OpenRouter, Ollama, etc.)

2. Is this happening with a specific model, or all models you've tried?

3. Did the tool protocol setting (when it existed) help resolve similar issues for you in the past? If so, which protocol setting worked better for your use case (XML or native)?

Context on what I've found:
The "Tool Call Protocol" dropdown was intentionally removed in a recent update (PR #10281) that deprecated XML tool protocol. The system now automatically uses native tool calling for models that support it. However, if your model doesn't handle native tools well despite being marked as supporting them, this could be causing the "Model response incomplete" error.

Understanding your specific provider and model will help determine whether:

• The model configuration needs to be updated (disable supportsNativeTools for that model)
• We need to re-add a way for users to override the protocol for edge cases
• There's another underlying issue

Please tag @roomote in your response with the answers, and I'll be happy to investigate further and propose a solution.

[lftx1986]

@roomote I tried both the Qwen3 and Kimi2 models, but both threw errors. In an older version of Roo Code, I could set it to use the XLM tool protocol and everything worked fine. I’d like to know how to enable the XLM protocol in the current version—do I have to stick with the historical release?

[roomote]

Thanks for the additional details! Qwen3 and Kimi2 are OpenAI-compatible models that may not handle native tool calling well despite the API supporting it.

I'll look into adding a way for users to override the tool protocol back to XML for cases like this. Will open a PR with my attempt at addressing this shortly.

[roomote]

Opened a PR with my attempt at addressing this: #10460

The PR restores the Tool Call Protocol setting to the Advanced Settings section. Once merged, you can go to Settings > Providers > Advanced Settings and select "XML" for the Tool Call Protocol to use the XML protocol with your Qwen3 and Kimi2 models.

Feedback welcome!

[mini2s]

@roomote I tried both the Qwen3 and Kimi2 models, but both threw errors. In an older version of Roo Code, I could set it to use the XLM tool protocol and everything worked fine. I’d like to know how to enable the XLM protocol in the current version—do I have to stick with the historical release?

Our solution is to support rollback XML -> use costrict

But it is still recommended to use the native protocol

[roomote]

Thanks for sharing your solution with costrict! You're right that using the native protocol is the recommended approach when possible.

PR #10460 restores the Tool Call Protocol setting in Advanced Settings, which will allow users like @lftx1986 to switch to XML for models that don't handle native tool calling well (like Qwen3 and Kimi2).

Once merged, users can access this via: Settings > Providers > Advanced Settings > Tool Call Protocol > XML