Suggestion: carapace autocomplete definitions for Dorothy commands

[balupton]

This is a suggestion from @molleweide

We could build a parser for the help documentation, or the options parsing, or even better, use a LLM trained on Dorothy's source files, to generate the carapace definitions.

As for which AI to use, it would need to be something that can be trained/loaded with Dorothy's source files, in which a GitHub Copilot alternative seems suitable, in which a quick search reveals these (there could be more):

• https://stackoverflow.com/q/76741410/130638
• https://codeium.com - not local, doesn't seem to have an api, is a cloud solution
• https://supermaven.com - not local, doesn't seem to have an api, is an IDE
• https://github.com/fauxpilot/fauxpilot - not really any docs, not sure if it supports bash
• https://github.com/TabbyML/tabby - doesn't support bash: https://tabby.tabbyml.com/docs/references/programming-languages/
• https://github.com/slashml/awesome-coding-copilot?tab=readme-ov-file#models - many outdated

[molleweide]

Since dorothy commands follow posix more or less it shouldnt be hard to put together something basic that just adds options and actions to the yaml configs.

The first iteration doesnt even have to be perfect, just a dwmo to see how it would be..