Troubleshoot Gemini

This document shows you how to resolve setup issues with Gemini.

The Gemini icon doesn't show up on the platform bar or the Gemini pane doesn't open in the Google Cloud console

If the Gemini icon does not show on the Google Cloud console platform bar, or the Gemini pane in the Google Cloud console doesn't open, try the following to resolve the issue:

  • Verify that the ID of the Google Cloud project you're using is one of the project IDs you submitted after you were granted access to Gemini. For more information, see Apply for Gemini access.

  • Verify that the Cloud AI Companion API is enabled in your Google Cloud project. For more information, see Enable the Cloud AI Companion API.

  • Verify you have the correct Identity and Access Management (IAM) permissions. You might need to ask your administrator to verify your permissions and, if necessary, add the missing permissions. For more information, see Get required IAM permissions.

  • Check whether you're on a Google Cloud console page that doesn't support Gemini. For example, some support pages and some pages about getting started don't support Gemini.

  • If your project's ID was submitted for Gemini access, and the Cloud AI Companion API is enabled for your project, and you have the correct IAM permissions, try logging out of the Google Cloud console and then log back in.

Gemini returns an incomplete answer about another cloud service

Gemini is optimized for Google Cloud. It works well with Google Cloud products and use cases. The domain knowledge of Gemini is not optimized for products and use cases in other cloud services, such as Azure, Amazon Web Services (AWS), and Oracle Cloud Infrastructure (OCI). Because of this optimization, questions about those cloud services aren't supported.

Gemini can provide limited information about how a Google Cloud product compares to a product from another cloud provider. For example, the following prompts are supported:

  What is the equivalent product in Google Cloud to AWS Sagemaker?
  Can you help me rewrite this Azure Serverless function so that it runs in Cloud Functions?

Gemini returned an answer similar to, "Sorry, I can't help you with that"

To improve accuracy, we prevent Gemini from responding to categories of questions when there is a low level of confidence in the response. If your question is classified to be in one of those categories, then the question is blocked and you receive an answer similar to, "Sorry, I can't help you with that."