Rubber Duck Thursdays - Let's build with agents

Key Concepts:

• Copilot Coding Agent: An AI tool that assists developers with coding tasks.
• Custom Instructions: Guidelines provided to Copilot to tailor its behavior to a specific project or coding style.
• Copilot Setup Steps: Configuration files that ensure Copilot has the necessary dependencies and environment settings.
• GitHub Actions: A CI/CD platform to automate software workflows.
• Playwright MCP Server: A tool that allows Copilot to interact with and navigate web applications.
• Trend Radar: A visualization used for competitive analysis.
• Codebase Context: Providing Copilot with relevant code and information to improve its understanding and suggestions.
• Linting: Automated code quality checks to ensure code style and consistency.
• Dependency Management: Managing and updating project dependencies.
• Test Coverage: Measuring the extent to which code is covered by automated tests.

1. Copilot Coding Agent and Public Preview:

• Copilot Coding Agent is now available for Copilot Pro users in public preview.
• It will also be available for education students, educators, and open-source maintainers with Copilot Pro access starting July 1st.

2. GitHub Actions Fine-Grained Permissions:

• Fine-grained permissions are now generally available for custom repository roles.
• This allows for more granular permission assignments, which is beneficial for enterprise governance.

3. Deprecated Cohere Models:

• The following Cohere models are deprecated: command-r and command-r-plus.
• Users should switch to command-r-08-2024 and command-r-plus-08-2024.

4. Upgraded Llama Models:

• Metal Llama models have been upgraded.
• Users should switch to Llama 3 models.

5. Improved Pull Request File Tree Experience:

• The pull request file tree experience has been improved with accessibility enhancements, faster diff rendering, and lower memory usage.
• UX improvements include pending comments in the review submission panel and local draft comments persisting across page refreshes.

6. Breaking Changes for GitHub Actions:

• Larger hosted runners will be removed from the self-hosted API.
• The orgs actions runner API will only show self-hosted runners.

7. Copilot for Xcode Updates:

• Copilot for Xcode now supports copilot vision, custom instructions, and local response support.
• Users can upload images directly into copilot chat and receive replies that match their development language.

8. Managing Cost Centers via API:

• Cost centers can now be managed through the API for users utilizing the new billing capabilities.

9. CodeQL Support for Rust:

• CodeQL support for Rust is now in public preview.

10. Dependabot Updates:

• Dependabot supports configuration of a minimum package age.
• Single pull request per dependable is now available.
• Dependency auto submission now supports NuGet.

11. Secret Scanning and Code Scanning:

• Secret scanning adds validity checks for Doppler and defined networking.
• Delegated alert dismissal for code scanning is now generally available.

12. Enterprise-Level Access for GitHub Apps:

• Enterprise-level access is available for GitHub apps and installation automation APIs.

13. Copilot Code Review Improvements:

• Copilot code review has better handling of large pull requests.
• It can now review significantly more files in a typical pull request.

14. Copilot Coding Agent Web Browser:

• Copilot Coding Agent now has its own web browser.
• It has access to the Playwright MCP server, allowing it to navigate apps and take screenshots.

15. Fast Follow Blog Posts:

• The team is trying out fast follow blog posts to recap the streams.
• An example blog post was shared, summarizing the previous week's stream on the Nex.js app for Copilot Airways.

16. Updating Custom Copilot Instructions:

• The presenter updated the custom Copilot instructions for a GitHub action to be more clear, concise, and accurate.
• The updated instructions included guidelines for code formatting, testing, and repository structure.

17. Pre-installing Dependencies in GitHub Copilot Environment:

• The presenter created a copilot-setup-steps.yml file to pre-install dependencies in the Copilot environment.
• This ensures that Copilot has access to the necessary tools and libraries.

18. Trend Radar App UI Updates:

• The presenter demonstrated UI updates to a trend radar app using Copilot Coding Agent.
• The updates included adding the ability to place points by clicking on the diagram and drag and drop points around.
• Copilot used the Playwright MCP server to navigate the app and take screenshots.

19. Improving GitHub Action with Copilot Coding Agent:

• The presenter used Copilot Coding Agent to improve a GitHub action for validating file existence.
• Copilot identified technical debt, such as incorrect package metadata and inconsistencies in the README.
• Copilot created a pull request with fixes for these issues, including updating the package.json, README, and adding input validation.

20. Addressing Technical Debt:

• Copilot identified and addressed technical debt in the GitHub action project.
• This included updating package metadata, improving documentation consistency, and adding input validation.

21. Testing and Validation:

• Copilot added comprehensive tests for the validation logic.
• The tests verified that empty input, whitespace-only input, and input with only commas throw appropriate errors.

22. Code Coverage:

• Copilot increased the code coverage from 75% to 90%.

23. Final Summary and Pull Request Update:

• Copilot provided a final summary of the changes in the pull request body.
• The summary included details on the package name, functionality, homepage, repository URL, bug URL, version, README examples, and validation behavior.

24. Linting Errors:

• Copilot identified and fixed linting errors in the codebase.

25. Copilot Availability:

• Copilot Coding Agent is now available on Copilot Pro and Copilot Business.

Main Takeaways/Synthesis:

The video demonstrates the latest updates and capabilities of GitHub Copilot and Copilot Coding Agent. It highlights the importance of providing clear custom instructions and setting up the environment for Copilot to work effectively. The presenter showcases real-world examples of using Copilot to improve code quality, add new features, and address technical debt. The integration of the Playwright MCP server allows Copilot to interact with web applications, enabling more advanced UI testing and development workflows. The video emphasizes the iterative nature of working with Copilot, where developers can provide feedback and refine the AI's suggestions to achieve the desired results.