This tutorial will teach you how to use AI to generate and execute workflows from natural language descriptions. No workflow coding required!
Prerequisites
Before starting, ensure you have:1
Kubiya SDK Installed
2
API Keys Ready
- Kubiya API Key: Get from compose.kubiya.ai
- Together AI Key: Get from together.ai
3
Environment Setup
Your First AI Workflow
Let’s start with a simple example - creating a workflow that checks system health:Understanding AI Generation
How It Works
The AI provider follows this process:What the AI Considers
Available Resources
Runners, integrations, and tools in your environment
Best Practices
Optimal patterns for reliability and performance
Error Handling
Automatic retry logic and failure recovery
Dependencies
Proper step ordering and parallelization
Streaming Generation
Watch the AI work in real-time:Generate and Execute
Go from idea to execution in one step:Advanced Examples
Multi-Service Deployment
Deploy a complete application stack:Incident Response Automation
Create an intelligent incident response workflow:Data Pipeline Automation
Build complex data pipelines with natural language:Working with Context
Provide context to generate better workflows:Iterative Refinement
Improve workflows through conversation:Error Handling
The AI handles errors intelligently:Production Tips
Be Specific
Be Specific
The more detailed your description, the better the result:
Test First
Test First
Always use plan mode first:
Monitor Generation
Monitor Generation
Track AI performance:
Handle Failures
Handle Failures
Implement proper error handling:
Next Steps
ADK Deep Dive
Learn more about the ADK provider
Streaming Guide
Master real-time streaming
Examples
Explore more workflow examples
API Reference
Complete API documentation
Summary
You’ve learned how to:- ✅ Generate workflows from natural language
- ✅ Stream generation progress in real-time
- ✅ Execute workflows immediately after generation
- ✅ Provide context for better results
- ✅ Refine workflows iteratively
- ✅ Handle errors gracefully