/deploy

Arguments: $ARGUMENTS

Goal: Deploy Google ADK agents to Google Cloud platforms (Vertex AI, Cloud Run, or GKE) with proper configuration and validation

Core Principles:

• Detect project structure and deployment requirements
• Validate credentials and prerequisites before deployment
• Support multiple deployment targets
• Provide clear deployment status and URLs

Phase 1: Discovery Goal: Understand project structure and deployment target

Actions:

• Parse $ARGUMENTS for deployment target (vertex-ai, cloud-run, gke, or auto-detect)
• Check for Google ADK project files
• Example: !{bash ls -la adk.yaml package.json pyproject.toml go.mod pom.xml 2>/dev/null}
• Detect project language and framework
• Load deployment configuration if exists
• Example: @adk.yaml

Phase 2: Validation Goal: Verify prerequisites and credentials

Actions:

• Check if gcloud CLI is installed and authenticated
• Example: !{bash gcloud auth list --filter=status:ACTIVE --format="value(account)" 2>/dev/null}
• Verify Google Cloud project is set
• Example: !{bash gcloud config get-value project 2>/dev/null}
• Check for required deployment files (Dockerfile, k8s manifests, etc.)
• Validate ADK configuration is complete
• Load project context for agent

Phase 3: Planning Goal: Determine deployment strategy

Actions:

• Based on $ARGUMENTS and detected configuration:
  • If "vertex-ai": Deploy as Vertex AI agent
  • If "cloud-run": Deploy as Cloud Run service
  • If "gke": Deploy to Google Kubernetes Engine
  • If no target specified: Recommend based on project structure
• Identify required Google Cloud APIs to enable
• Check if deployment artifacts exist (container image, etc.)
• Present deployment plan to user

Phase 4: Deployment Goal: Execute deployment with specialist agent

Actions:

Task(description="Deploy Google ADK agent", subagent_type="google-adk-deployment-specialist", prompt="You are the google-adk-deployment-specialist agent. Deploy the Google ADK agent for $ARGUMENTS.

Project Context:

• Project structure detected in Phase 1
• Deployment target: [vertex-ai/cloud-run/gke]
• Google Cloud project: [from gcloud config]
• Language/framework: [detected language]

Requirements:

• Build container image if needed
• Push to Google Container Registry or Artifact Registry
• Configure deployment settings (scaling, resources, environment variables)
• Deploy to target platform
• Set up monitoring and logging
• Configure IAM permissions if needed

Deliverable:

• Deployment status (success/failure)
• Service URL or endpoint
• Deployment configuration summary
• Next steps for testing and monitoring")

Phase 5: Verification Goal: Confirm deployment succeeded

Actions:

• Check deployment status
• Verify service is running
• Example: !{bash gcloud run services list --filter="metadata.name:adk-agent" --format="value(status.url)" 2>/dev/null}
• Test health endpoint if available
• Display service logs for verification

Phase 6: Summary Goal: Report deployment results and next steps

Actions:

• Display deployment summary:
  • Platform deployed to
  • Service URL/endpoint
  • Configuration applied
  • Resource allocation
• Show monitoring and logging commands
• Suggest next steps:
  • Test the deployed agent
  • Set up custom domain
  • Configure CI/CD pipeline
  • Enable additional monitoring