Deploy to Vertex AI Agent Engine¶

Push your Regulus-wrapped ADK agent to Vertex AI Agent Engine as a managed runtime, with audit going to Kafka and residency pinned to UK or EU.

Prerequisites¶

Vertex AI Agent Engine enabled on your GCP project.
Cloud KMS key in the same region as your Vertex resources, exposed via REGULUS_CMEK_KEY.
A Kafka cluster (Confluent / Managed Kafka) reachable from Agent Engine.
A container image — we use Jib.

1. Production configuration¶

regulus:
  compliance:
    profiles: [eu-ai-act, uk-gdpr, fca-sysc, pra-ss1-23]
  adk:
    name: production-agent
    session-service:
      kind: vertex-ai
      project-id: ${GOOGLE_CLOUD_PROJECT}
      location: europe-west2
      cmek-key-name: ${REGULUS_CMEK_KEY}
    audit:
      sink: kafka
      kafka-topic: audit.regulus.v1
    kill-switch:
      enabled: true
      dual-control: true
    residency:
      allowed-regions: [europe-west2]
      require-cmek: true
    model-risk:
      tenant-tier: REGULATED

2. Build the container¶

./gradlew :examples:adk-vertex-agent-engine-deploy:jib

Jib pushes ghcr.io/neul-labs/regulus-adk-demo:<version> (or wherever you configure).

3. Run the build-time doctor¶

./gradlew regulusAdkDoctor

This checks ADK + Regulus version compatibility, residency wiring, signing config, and refuses to proceed on misconfiguration. Designed to fail your CI before adk deploy ever runs against a misconfigured service.

4. Deploy¶

adk deploy --image ghcr.io/neul-labs/regulus-adk-demo:0.1.0 \
           --location europe-west2 \
           --project $GOOGLE_CLOUD_PROJECT

5. Verify in Agent Engine¶

In the Vertex console:

Confirm the agent's region is europe-west2.
Inspect the Cloud Logging stream — Regulus audit events appear under the Kafka producer's path, plus startup logs from each plugin.
Send a sample request via the Agent Engine REST API. Confirm the audit topic in Kafka receives the expected events.