Python SDKSlack

VertexAI

Integrate with the VertexAI API to get access to Google models

Add your Google Cloud Platform ProjectID and location to begin using the VertexAI API.

Note: By adding this integration, your data may be sent to Google for certain actions within Arize (e.g., prompt playground) and your account may be billed for usage.

Select Google VertexAI Integration

Select the Google VertexAI Integration in the Arize Console. Settings -> Account Settings -> Integrations

Create an Arize Integration Key

To verify ownership of your project form within Arize, we need to create a project access label on your GCP project. Arize will use this to ensure that the Arize organization you are running Vertex commands from has access to your GCP project.

To create this go into your GCP and search for labels. Add a label with the key arize-integration-key and whatever value you want.

Create a label on your GCP project

Fill in your Details in the Arize Console

Add the GCP ProjectID, location, and project access label of the project you have the VertexAI API enabled in. Fill this out in the Arize Vertex integreation wizard.

Fill in your project ID, location, and Arize Integration Key

Create a Custom IAM Role Using the Code Given in the Arize Vertex Wizard

Copy the following IAM role gcloud command to run in your GCP console:

Copy the Custom IAM Role Command from the Arize Vertex Wizard

Sample commands:

You must add your project ID to the commands below where it says <my-project>. The commands pulled from the Arize UI have your project ID prefilled.

Create command (most cases):

gcloud iam roles create arizeApp --project=<my-project> --title="Arize App" --description="Custom IAM role for Arize App" --permissions=aiplatform.endpoints.predict,resourcemanager.projects.get --stage=ALPHA

Update migration (uncommon):

If you have a pre-existing integration and need to update it to add an arize-integration-key label, please change the create portion of the command to update.

gcloud iam roles update arizeApp --project=<my-project> --title="Arize App" --description="Custom IAM role for Arize App" --permissions=aiplatform.endpoints.predict,resourcemanager.projects.get --stage=ALPHA

Run the command in the GCP console:

Run the Custom IAM Role Command

Assign the Custom Role to the Arize Service Account

In order for Arize to send requests to the VertexAI API in your project, you must assign the Arize App Service Account that role. NOTE: You will be using the Arize Service Account, you do not need to create your own service account

Copy the following IAM permissions gcloud command to run in your GCP console:

Copy the Permissions Command

Run the command in the GCP console:

Run the Permissions Command

Supported Models

  • gemini-1.5-pro

  • gemini-1.5-flash

  • gemini-1.0-pro-vision

  • gemini-1.0-pro

  • gemini-pro

  • text-bison-32k

  • text-bison

Last updated

Was this helpful?