I’ve written some posts in the last month here about this.
The secret sauce in AI dev is the INSTRUCTIONS file which the ai (hopefully) uses to craft your code. I’ve got a directory of instruction files and in my copilot-instructions.md I just point to those docs using a markup document (markup is best to use for ai apparently)
One for CFML coding standards (based on the ortus one)
One for my TeamCFML cfml framework (our 25 year old framework)
One for Testbox (for testing)
One for my database framework (i call diimes)
One for my sencha ExtJS layouts (HelpfulUI)
One for Siesta (js/ui testing framework)
One for my screenshot testing app that takes screenshots of my apps during unit testing (then hash the image) and compare image to previous version to prove interface is the same (or different with new feature in that area).
Then just tell ai what to do, and add new items along the way, or change the rules in your doc.
and I use a coldfusion port of docsify (that AI wrote in 5 seconds) to do all my docs in markdown as docsify collections, then got AI to write a documentation portal that nests all the docs from above. Docsify is super easy to use with ai also.
Copilot is amazing when you understand how each LLM works.
Check out:
Heads up : overage for github copilot is currently not charged - Tools and Libraries - Lucee Dev