Create AI Powered Actions

Build intelligent Javascript based GitHub Actions that leverage GitHub Models.

Welcome

• Who is this for: Developers who want to build custom AI-enhanced GitHub Actions

• What you'll learn: How to leverage GitHub Models in your Javascript actions, implement structured AI outputs with Zod schemas and create responsive workflows.

• What you'll build: A complete AI-powered Javascript based GitHub Action that analyzes and rates jokes

• Prerequisites:

  • We strongly recommend completing the write-javascript-actions exercise first, as it covers the foundational skills needed for creating JavaScript-based GitHub Actions

• How long: This exercise takes less than 1 hour to complete.

In this exercise, you will:

1. Set up a development environment and install the OpenAI SDK for GitHub Models integration
2. Create action metadata and implement AI-powered joke rating logic using GitHub Models
3. Test your action locally with debugging tools and package it for distribution
4. Author a GitHub Actions workflow that triggers on issue comments and updates them with AI analysis
5. Implement structured outputs using Zod schemas for reliable data processing and conditional workflow logic

How to start this exercise

Simply copy the exercise to your account, then give your favorite Octocat (Mona) about 20 seconds to prepare the first lesson, then refresh the page.

Having trouble? 🤷

When copying the exercise, we recommend the following settings:

• For owner, choose your personal account or an organization to host the repository.

• We recommend creating a public repository, since private repositories will use Actions minutes.

If the exercise isn't ready in 20 seconds, please check the Actions tab.

• Check to see if a job is running. Sometimes it simply takes a bit longer.

• If the page shows a failed job, please submit an issue. Nice, you found a bug! 🐛

---

© 2025 GitHub • Code of Conduct • MIT License