Use Google Cloud Vertex AI with Edge Delta

Integrate Edge Delta with your Google Cloud Vertex AI platform.

  2 minute read  

Background

Edge Delta features OnCall AI to interpret anomaly patterns generated by Fleets on the edge and monitors in the Edge Delta back end. Edge Delta sends some non-sensitive data, such as log patterns, to Edge Delta’s third-party AI service providers. You can configure Edge Delta to call your Google Cloud Vertex AI platform instead.

Overview

This how to will demonstrate how to configure Vertex AI and Edge Delta so that Edge Delta calls the the Vertex AI API.

1. (Optional) Create a new Google Cloud Project

Since you will be granting Edge Delta permission to make Google Cloud Vertex AI calls on your behalf it may be difficult to differentiate between the calls made by Edge Delta from other Vertex AI calls. An optional step to separate your own Vertex AI calls from those initiated by Edge Delta is to create a new GCP account that would be governed by the same billing as your master GCP account. This provides some security and cost tracking benefits.

2. Enable Vertex AI

Follow the steps in the Google Cloud Vertex AI documentation to enable the Vertex AI API.

3. Give Access to Edge Delta’s Service Account to call Vertex AI on your behalf

Please contact support to obtain Edge Delta service account email address. Then, grant the following privileges to this service account: Vertex AI User

  1. Go to the IAM & Admin section in the Google Cloud Console.
  2. Click Grant Access.
  3. Enter the Edge Delta service account email address in the “New Principals” section.
  4. Select the role Vertex AI User in the “Select a Role” section.

Alternatively, create a new role with this permission: aiplatform.endpoints.predict and assign it to the Edge Delta service account.

4. Configure Edge Delta

In this step you specify the enabled models in Edge Delta and provide the sts:ExternalId.

  1. Click Admin and select the My Organization tab.
  2. Click Edit in your Organization pane.
  3. Select GCP in the AI provider selection section.
  4. Enter your GCP Project ID, GCP Region.
  5. Specify the names of the models you wish to use for anomaly processing.
  6. Click Save.