Gemini 3 models fail with tool calling because of missing thoughtSignature

[krzys-h]

Describe the bug

The Gemini tinking models models require the thought signature to be passed during function calling, see https://ai.google.dev/gemini-api/docs/thought-signatures

With older models this was optional to improve performance, with Gemini 3 it's strictly required.

Currently it just returns an error after the tool calling turn every time:

Error in Agent node: [GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-3-flash-preview:streamGenerateContent?alt=sse: [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. Additional data, function call default_api:gsa , position 4. Please refer to https://ai.google.dev/gemini-api/docs/thought-signatures for more details.

To Reproduce

1. Create any agentflow which will cause an agent to call a tool. I'm using an Agent node with a Document Store, but simpler tools should trigger this as well.
2. Make it use a Gemini 3 model, e.g. gemini-3-flash-preview
3. Run it
4. See that it fails

Expected behavior

The flow runs with no error

Screenshots

No response

Flow

No response

Use Method

Docker

Flowise Version

3.0.13

Operating System

Linux

Browser

Chrome

Additional context

No response

[HenryHengZJ]

WIP: https://github.com/FlowiseAI/Flowise/tree/chore/Langchain-V1