Hi everyone,
I’m trying to use the new Gemini 3 Pro Preview model (gemini-3-pro-preview) with the @n8n/n8n-nodes-langchain.agent node and the Google Gemini Chat Model node.
Everything works fine for simple text generation, but as soon as I add Tools to the agent (Function Calling), I get the following error:
[GoogleGenerativeAI Error]: Error fetching from ``https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-preview:generateContent:`` [400 Bad Request] 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.
The Issue
According to the Gemini 3 documentation, this new model enforces a strict security protocol called “Thought Signatures”. When the model generates a Function Call, it includes a thoughtSignature field. The API requires the client (n8n) to send this signature back exactly as received in the next turn of the conversation history.
It seems the current lmChatGoogleGemini node implementation does not capture or pass this thoughtSignature back to the API when handling tool executions, causing the strict validation to fail with a 400 error.
Documentation Reference
Workaround Attempt
Google mentions a workaround for migrating from other systems by sending a dummy signature: "thoughtSignature": "skip_thought_signature_validator" However, there is currently no way to inject this manually into the conversation history within the standard n8n node.
Request
Could we get an update to the Google Gemini Chat Model node to support these new Thought Signatures? This is a blocker for using Gemini 3’s advanced reasoning capabilities with Agents in n8n.
Thanks!