Diagnostics & Troubleshooting

Diagnostics & Troubleshooting

Use these steps when the extension does not behave as expected.

Table of Contents

• Diagnostics & Troubleshooting
  • Table of Contents
  • Check the output channel
  • Refresh models
  • Authentication issues
    • Standard Credentials (ADC)
    • Service Accounts
  • Common problems
  • Old settings migration
  • Debug tools

Check the output channel

Open the VS Code Output panel and select Vertex AI Models Chat Provider.

This channel logs:

• model discovery progress
• authentication errors
• tool dump output
• request and usage diagnostics

Refresh models

If model availability changes, refresh manually:

• Vertex AI Models Chat Provider: Refresh Models

If discovery fails, the extension may also show a warning or error message.

Authentication issues

Standard Credentials (ADC)

If credentials are invalid or expired, run:

gcloud auth application-default login --project YOUR_PROJECT_ID --quiet

The extension can prompt you for sign-in and attempt recovery when it detects expired credentials.

Service Accounts

If you see a warning: "Vertex AI: Could not load Service Account 'Name'...":

• You may have deleted the secret or moved the JSON file.
• Use Vertex AI: Select Authentication Method to re-select a valid key or switch back to ADC.
• You can suppress this warning for a specific project by clicking "Don't Show Again" in the notification.

Common problems

Problem	Likely cause	Fix
No models appear in Chat	vertexAiChat.projectId missing or invalid	Set vertexAiChat.projectId in VS Code settings.
Discovery returns no models	Vertex AI API not enabled, wrong region, or missing model access	Enable Vertex AI API and make sure the chosen models are available in the Model Garden.
Commit message button not visible	vscode.git not available or no Git repository	Open a Git-tracked workspace with staged changes.
Cost dashboard warning persists	hideBillingWarning unset	Set vertexAiChat.hideBillingWarning to true if you understand the estimate behavior.

Old settings migration

If you previously used the legacy vertexAnthropic namespace, the extension automatically migrates:

• vertexAnthropic.projectId → vertexAiChat.projectId
• vertexAnthropic.hideBillingWarning → vertexAiChat.hideBillingWarning

Debug tools

The extension also exposes a tool dump command for debugging installed language model tools:

• Vertex AI Models Chat Provider: Dump Installed Tools Schema

This command writes tool names, descriptions, and input schemas to the output channel.