[BUG] Native gemini 3-pro tool calls seem to be broken (Function call is missing a thought_signature in functionCall parts)

[dmarkey]

Problem (one or two sentences)

While trying to use the new native gemini-3-pro tool calls with the gemini provider it seems to have problems with this error being returned:

Gemini generate context stream error: {"error":{"message":"{\n "error": {\n "code": 400,\n "message": "Function call is missing a thought_signature in functionCall parts. This is required for tools to work correctly, and missing thought_signature may lead to degraded model performance. Additional data, function call default_api:ask_followup_question , position 4. Please refer to https://ai.google.dev/gemini-api/docs/thought-signatures for more details.",\n "status": "INVALID_ARGUMENT"\n }\n}\n","code":400,"status":"Bad Request"}}

I've also experienced this with the code search tool.

XML based(non native) tool calls work.

Context (who is affected and when)

Gemini 3.0-pro users that want to use native tool calls with the gemini provider.

Reproduction steps

1. Set up gemini-3-pro with the gemini provider.
2. Try to push the agent to do tool calls.

Expected result

Tool calls should consistently work.

Actual result

Tool calls inconsistently work.

Variations tried (optional)

No response

App Version

3.34.0

API Provider (optional)

Google Gemini

Model Used (optional)

Gemini-3-pro-preview

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

[roomote]

I'm exploring a potential fix for this issue. The problem appears to be that Gemini 3.0-pro requires thought_signature in all function call parts, but the current implementation only includes them when reasoning/thinking is explicitly enabled. I'll open a PR to share progress.

[dmarkey]

You're right, it doesn't seem to happen when I've "Model Reasoning Effort" set.

[roomote]

Opened a PR with my attempt at addressing this issue. The fix ensures that thought signatures are always included for gemini-3-pro models when using native tool calls, even when reasoning is not explicitly enabled.

PR: #9500

Feedback welcome!

[steenlarsen]

I have the same problem with gemini-3-flash-preview on OpenRouter

Date/time: 2025-12-23T14:41:34.649Z
Extension version: 3.37.0
Provider: openrouter
Model: google/gemini-3-flash-preview

400
OpenRouter completion error: Gemini models require OpenRouter reasoning details to be preserved in each request. Please refer to our docs: https://openrouter.ai/docs/guides/best-practices/reasoning-tokens#preserving-reasoning-blocks. Upstream error: {
"error": {
"code": 400,
"message": "Function call is missing a thought_signature in functionCall parts. This is required for tools to work correctly, and missing thought_signature may lead to degraded model performance. Additional data, function call default_api:ask_followup_question , position 2. Please refer to https://ai.google.dev/gemini-api/docs/thought-signatures for more details.",
"status": "INVALID_ARGUMENT"
}
}