I have just created a repo, with code to create an AI which can rate how well text in commits is written (text in the form of documentation, code comment, etc.), there is also a script which can use human feedback to improve the model, so the AI matches our perspective of what well or not well-written text is. This can be integrated into the automated process on github, so for each pull request, the AI can check whether the text inside the commit is written well enough. This automation is not done in my repo and is what is still missing, the details of the AI like loss function, parameters, model, etc. are described in the read me, gladly if you find improvement you can also refine the logic. Here is the repo link: GitHub - KiyoLelou10/RIOTCommitTextScoreAI.
1 Like
Pretty cool. Any tool that makes PRs go more smoothly is a win.
Is there some way to demo this so that RIOT project maintainers can evaluate if’s usefulness?