Quick start with Google Cloud Vertex AI

Availability

Business Plan for the Wren AI Self-Hosted Version.

Supported regions

Ensure that your selected region supports the following Google models we're using by default:

• gemini-2.5-flash
• gemini-2.5-pro
• gemini-2.0-flash-lite-001

Refer to the official Google model endpoint locations: Google model endpoint locations.

If you encounter any issues, feel free to contact us to discuss alternative model selections for your region.

Prerequisites

1. Enable Vertex AI API Go to the Vertex AI console and enable the Vertex AI API.

   

2. A service account private key (JSON) with the role "Vertex AI User"

   How to create a service account:

   1. Go to GCP Console → IAM & Admin → Service Accounts. Click "Create service account".

   

   2. Fill in name.

   

   3. Assign permissions — choose "Vertex AI User".

   

   4. Open the service account you just created.

   

   5. In the "Keys" tab, click "Add key" → Create new key, then choose JSON key type to download the private key.

   

Steps (Quick setup)

1. During onboarding, select "Quick setup" and then "Google Gemini (Vertex AI)".

2. Enter Vertex location (region), Vertex project (your GCP project ID), then upload your service account JSON.

3. Test embedding model connection.

4. Update model configurations. There are default configs of the pre-selected models on the UI.

   

   You'll need to update the configurations for optimization. Click "Configure" for each listed model and set:

   {
     "kwargs": {
       "n": 1,
       "timeout": 60,
       "max_tokens": 4096,
       "temperature": 0
     },
     "context_window_size": 1000000
   }

5. Review the pipeline assignments.

   We provide default pipelines optimized for Gemini models. You could leave it as is.

6. Complete the setup.

   Scroll to the bottom and click "Complete setup".