Deployment

Deploy your ADK-TS agents to production using various platforms and deployment strategies. Choose the approach that best fits your infrastructure and requirements.

Platform Guides

Deploy to popular cloud platforms with step-by-step instructions:

Agent-Specific Guides

Deployment guides for specific agent types with unique requirements:

Coming Soon

We're working on guides for additional platforms:

• ☁️ Google Cloud - Cloud Run and GKE deployment
• 🌐 Netlify - Serverless functions and static site deployment
• ⚓ Kubernetes - Container orchestration for scalable deployments

Prerequisites

Before deploying your agent, ensure you have:

• Built and tested your ADK-TS agent locally
• Environment variables documented and ready to configure
• Dependencies properly defined in package.json
• Production configuration prepared (API keys, database connections, etc.)

General Best Practices

To ensure a stable and secure production environment, follow these best practices:

• Environment Variables: Use platform-specific secret management (e.g., AWS Secrets Manager, Vercel/Railway Environment Variables). Never commit .env files or sensitive data to version control.
• Resource Limits: Set appropriate memory and CPU limits to prevent resource exhaustion and manage costs.
• Logging & Monitoring: Configure production-grade logging and monitoring. Regularly review logs to catch errors and performance issues early.
• Security: Use IAM roles or service accounts instead of long-lived access keys. Implement the principle of least privilege for all permissions.
• Version Control: Use specific tags (e.g., v1.0.0) or git commit SHAs for Docker images and deployments instead of latest to ensure reproducible builds.
• Health Checks: If your agent exposes an HTTP endpoint, implement health check routes to allow platform-level monitoring and automatic restarts.

Troubleshooting Common Issues

If you encounter issues during deployment, check these common points before diving into platform-specific configurations:

• Missing Variables: The most common cause of startup failure. Verify all required environment variables are set in your platform dashboard (case-sensitive).
• Architecture Mismatch: If using Docker, ensure you are building for the target architecture (usually linux/amd64) via docker buildx build --platform linux/amd64 -t your-image:tag ..
• Timeouts: AI processing can take longer than standard API requests. Check your platform's serverless or load balancer timeout settings (e.g., Lambda's 15m limit or Vercel's maxDuration).
• Build Failures: Review build logs carefully for missing dependencies, TypeScript compilation errors, or syntax issues in your Dockerfile.
• Runtime Errors: Check your service logs for stack traces. Ensure your compiled code (usually in dist/) is being correctly referenced in the start command.

Next Steps

Choose a deployment platform from the guides above to get started, or explore our other guides for advanced configurations and features.