Skip to content

Issue with Gemini 3.0 (gemini-3-pro-preview & gemini-3-flash-preview) & Tools: “Function call is missing a thought_signature” #23798

Description

@hugelm

Bug Description

When using the gemini-3-pro-preview or gemini-3-flash-preview models in n8n via Vertex AI Node, the workflow crashes during Function Calling.

While standard chat interactions work, any attempt by the model to use a connected Tool/Agent results in a Bad request Error regarding missing thought_signatures.

Error message:

Unable to submit request because function call default_api: in the 2. content block is missing a thought_signature. Learn more: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/thought-signatures

This does not occur with other Gemini models like gemini-2.5-flash.

The issue was also reported in some earlier versions of n8n:
https://community.n8n.io/t/issue-with-gemini-3-0-gemini-3-pro-preview-tools-function-call-is-missing-a-thought-signature/223824

To Reproduce

Create a new workflow with an AI Agent node.

Select the Google Vertex AI Node.

Set the Model ID to gemini-3-pro-preview or gemini-3-flash-preview.

Connect a Tool.

Send a prompt that forces the agent to use the tool.

The node fails immediately after the model attempts to generate the function call.

Expected behavior

The agent should be able to process the "thought_signature" generated by Gemini 3 and pass it back to the API along with the function call.

https://docs.cloud.google.com/vertex-ai/generative-ai/docs/thought-signatures

Debug Info

Debug info

core

  • n8nVersion: 2.0.3
  • platform: docker (self-hosted)
  • nodeJsVersion: 22.21.0
  • nodeEnv: production
  • database: sqlite
  • executionMode: regular
  • concurrency: -1
  • license: enterprise (production)
  • consumerId: 2c7cd698-b0f6-46b3-b8fa-dcfc11deb037

storage

  • success: all
  • error: all
  • progress: false
  • manual: true
  • binaryMode: filesystem

pruning

  • enabled: true
  • maxAge: 336 hours
  • maxCount: 10000 executions

client

  • userAgent: mozilla/5.0 (windows nt 10.0; win64; x64) applewebkit/537.36 (khtml, like gecko) chrome/143.0.0.0 safari/537.36
  • isTouchDevice: false

Generated at: 2026-01-02T16:40:13.114Z

Operating System

Raspberry Pi OS

n8n Version

2.0.3

Node.js Version

22.21.0

Database

SQLite (default)

Execution mode

main (default)

Hosting

self hosted

Metadata

Metadata

Assignees

No one assigned

    Labels

    status:in-linearIssue or PR is now in Linearstatus:team-assignedA team has been assigned the issue or PRteam:aiIssue is with the AI teamtriage:completeTriage has been completed and issue is ready for internal teams

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions