How to use AI models in your GitHub Actions workflows

Key Concepts:

• GitHub Models: Integration of AI models into projects and workflows.
• GitHub Actions: Automating tasks within GitHub repositories.
• AI Inference Action: Calling AI models from within GitHub Actions workflows.
• Model Catalog: A collection of over 40 leading AI models accessible through GitHub Models.
• gh-models extension: GitHub CLI extension for interacting with models.
• Project Bloom: An internal GitHub tool using Models and Actions to automate documentation for Product Managers.
• Permissions: Granting workflows access to use GitHub Models.
• Prompts: Instructions given to AI models to guide their output.
• Issue Summarization: Using AI to condense the content of GitHub issues.
• Release Note Generation: Automating the creation of release notes using AI.
• Issue Prioritization: Using AI to rank issues based on importance.

1. Using AI Models for Issue Validation

• Main Topic: Automating issue validation using AI within GitHub Actions.
• Key Points:
  • A workflow is triggered when a new issue is opened.
  • The workflow uses the models permission to access AI models.
  • The AI Inference action calls a model to determine if the issue contains enough information to reproduce the bug.
  • A system prompt instructs the model to return "pass" if sufficient detail is present, otherwise, to request more information.
  • The model value can be changed to call different models from the catalog.
  • The workflow comments on the issue only if the model's response is not "pass," enabling AI-powered branching logic.
• Example: A bug report lacking sufficient information triggers the model to generate a comment requesting more details from the issue creator.

2. Automating Release Note Generation

• Main Topic: Using GitHub CLI and AI models to automate release note creation.
• Key Points:
  • A workflow is triggered when a pull request is merged.
  • The workflow installs the gh-models extension for the GitHub CLI.
  • The workflow uses a GitHub token with permission to use models.
  • The workflow retrieves the pull request title, body, comments, and reviews.
  • The Grok model is used to generate a summary of the pull request.
  • The summary is then added to the bottom of a release issue.
• Example: Two pull requests are merged, and the workflow automatically adds two one-line items to the changelog in the release issue.

3. Weekly Issue Summarization and Prioritization

• Main Topic: Automating the summarization and prioritization of open issues using AI.
• Key Points:
  • A workflow runs weekly to summarize and prioritize open issues.
  • The workflow searches for open issues.
  • The workflow generates a summary of the issues.
  • The workflow creates a new issue containing the summary and suggested prioritization.
  • Prompts can be set directly in the workflow file or using a prompt file.
• Example: The weekly summary includes a list of new issues, general themes, and suggested prioritization.

4. Project Bloom: Automating Documentation for Product Managers

• Main Topic: Showcasing an internal GitHub tool (Project Bloom) that leverages Models and Actions to automate documentation.
• Key Points:
  • Project Bloom is triggered by new discussions in GitHub Discussions.
  • Bloom automatically creates a summary, supporting documents, an initiative brief, a changelog, and additional comments.
  • It centralizes context gathering and automatically updates documents as project details evolve.
• Example: A new discussion triggers Bloom to generate a summary, supporting documents, an initiative brief, a changelog, and additional comments.

5. Enabling Models in Actions

• Main Topic: How to enable Models in Actions.
• Key Points:
  • Models in Actions can be enabled by adding the models permission to workflows.

Synthesis/Conclusion:

GitHub Models, integrated with GitHub Actions and the GitHub CLI, provides powerful tools for automating various development workflows using AI. From validating issue details and generating release notes to summarizing and prioritizing issues and automating documentation, these tools can save developers and product managers significant time and effort. By granting workflows the models permission, users can access a catalog of AI models and leverage them through prompts and the AI Inference action. The examples provided, including Project Bloom, demonstrate the potential for AI to streamline and enhance software development processes.