Vertex AI

Connect Vertex AI endpoints to Evaligo with secure service accounts and regional controls. Validate quality, latency, and cost on sample datasets and roll out incrementally with monitors and dashboards.

Use Workload Identity Federation or service account keys stored in Secret Manager. Scope IAM roles to the minimum required, and enforce organization policies for data residency and access logging.

Setup

1. 1

   Select project/region Choose GCP project and region that meets latency and compliance needs.

2. 2

   Configure identity Use Workload Identity Federation or dedicated service accounts with least privilege.

3. 3

   Connect endpoint Provide endpoint and auth in Evaligo settings, then run a validation job.

4. 4

   Monitor and optimize Track request success, latency, and spend. Iterate on model and params.

from google.cloud import aiplatform
from evaligo.integrations import VertexAIIntegration

aiplatform.init(project='my-gcp-project', location='us-central1')

integration = VertexAIIntegration(endpoint='projects/123/locations/us-central1/endpoints/456')
client = integration.create_client(temperature=0.2, max_tokens=1200)
print('Vertex AI client ready:', bool(client))