Skip to main content

​
Guide

This guide covers how to set up Google Vertex AI for integration with ASSIST AI.
Google Vertex AI and Google AI Studio serve the same models. However, Vertex AI has enterprise-grade features that may be useful for your organization.
  1. Create a Service Account for Onyx
    • Go to the Google Cloud Console Service Accounts Page
    • Select your project and click Create Service Account.
    • Give your Service Account a name and a description.
    • On the Permissions tab, grant the service account the Vertex AI Administrator role.
  2. Generate a Key for the Service Account
    • Open the service account you just created, navigate to Keys → Add Key → Create new key.
    • Choose JSON as the format and click Create. The key file will download to your device automatically.
  3. Go to the AI Model Configuration Page Open the Admin Panel via your profile icon, then navigate to Admin Panel → LLM.
  4. Set Up Google Vertex AI
    • Choose GCP Vertex AI from the list of available providers.
    • Enter a Display Name for this configuration, upload your JSON key file to the Credentials File field, and specify a Location if applicable.
  5. Set Default and Fast Models
    • The Default Model is automatically applied to new custom Agents and Chat sessions. Optionally, you can designate a Fast Model, which handles background tasks like message classification, query expansion, and session naming.
    • If you assign a Fast Model, choose something lightweight and cost-efficient, such as GPT-4.1-mini or Claude 3.7 Sonnet.
  6. Select Visible Models Under Advanced Options, you’ll find a complete list of models offered by this provider. From here, you can control which models your users can see within ASSIST AI. This is especially helpful when a provider offers multiple models or versioned variants.
  7. Define Provider Access Finally, choose whether this provider should be visible to all users in ASSIST AI or restricted. If set to private, only Admins and explicitly assigned Users will have access to the provider’s models.