AI-Assisted Development

Modern AI tools can significantly accelerate Lumi plug-in development by acting as a collaborative coding partner.

VS Code in Agent Mode

Using Visual Studio Code with an AI assistant in Agent mode (such as GitHub Copilot’s Agent mode or other tool-enabled assistants) allows you to perform complex, multi-step tasks using natural language.

Instead of just completing a single line of code, an Agent can:

• Read your entire workspace to understand context.
• Create new files and directories.
• Run terminal commands to test or validate scripts.
• Search for existing patterns in your codebase.

Repository Access

AI assistance is most effective when the Agent has access to the lumi-dev or your specific project repository. With visibility into the existing codebase, the Agent can:

• Use the Utility Libraries as a reference for helper functions.
• Follow existing patterns for GEGL operations and layer management.
• Re-use boilerplate code from established plug-ins.

Example Workflow

You can directly ask the Agent to generate a full plug-in by describing the desired functional outcome:

“Using the available Scheme utilities and examples in the workspace, write a new plug-in that creates a 50% horizontal guide on the active image and names it ‘Center Guide’.”

The Agent will search for how to create guides, identify the correct utility function (like lumi-image-add-hguide-percent from common.scm), and generate the complete .scm file with the correct registration boilerplate.

Best Practices

• Be Specific: Describe exactly what you want the plug-in to do.
• Reference Utilities: Encourage the Agent to look at the share/lumi/scripts/ directory for high-level helpers.
• Review and Test: Always test the plug-in generated by the AI, it’s often an iterative and creative process.