Skip to content

Model Discovery & Project Switching

Jump to bottom
jorsm edited this page May 31, 2026 · 4 revisions

Model Discovery & Project Switching

This extension discovers available Vertex AI models automatically based on your configured Google Cloud project.

Table of Contents

How discovery works

When the extension starts, it probes supported GCP regions in order:

  1. global
  2. us-east5
  3. europe-west1
  4. asia-southeast1

For each region, it checks whether the models configured in models.json are available in your project. It then registers only the models that respond successfully.

Supported models

The extension includes model support for Google Gemini, Anthropic Claude, and MaaS open-weight models.

Google Gemini

  • Gemini 3.5 Flash
  • Gemini 3.5 Flash (High)
  • Gemini 3 Flash
  • Gemini 3 Flash (High)
  • Gemini 3.1 Pro
  • Gemini 3.1 Pro (High)

Anthropic Claude

  • Claude Opus 4.7
  • Claude Opus 4.6
  • Claude Sonnet 4.6
  • Claude Haiku 4.5

MaaS — Open-Weight Models

  • Qwen3 Coder 480B (Alibaba)
  • DeepSeek V3.2 (DeepSeek)
  • Kimi K2 Thinking (Moonshot AI)

See Model-as-a-Service (MaaS) for details on open-weight model support, thinking/reasoning modes, and per-model configuration.

Switching projects

The extension uses vertexAiChat.projectId to determine which GCP project to query.

  • If vertexAiChat.projectId is set in the workspace or user settings, the extension will automatically use that project for discovery and inference.
  • When the project ID changes, the extension automatically re-runs discovery and updates the available model list.

To switch projects, update your workspace settings and save. You should see a notification that discovery has restarted.

You can verify the active Project ID and its configuration source (e.g. Workspace vs Global setting) by hovering over the Status Bar.

Manual refresh

If your project permissions or model availability changes, refresh discovery manually:

  • Vertex AI Models Chat Provider: Refresh Models

This is useful after enabling new models in the Vertex AI Model Garden or changing IAM roles.

Authentication and errors

If authentication fails, the extension may prompt you to sign in with Google Cloud.

It can help you recover by running:

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

If no models are found, verify:

  • vertexAiChat.projectId is set correctly
  • The Vertex AI API is enabled
  • Your account has roles/aiplatform.user
  • The models you want are enabled in the Vertex AI Model Garden

Clone this wiki locally