docs: Update contribution guidelines to include LLM usage policies - #143
docs: Update contribution guidelines to include LLM usage policies#143stv0g wants to merge 1 commit into
Conversation
n-eiling
left a comment
n-eiling
left a comment
There was a problem hiding this comment.
Good idea to add some clarity. I left some comments.
I think the biggest issue other OSS projects have with AI is getting lots of generated issues and review comments. This guideline doesn't include guidance on this. Maybe something like "GitHub is the place where humans interact. PR descriptions, comments and issues should be written by and addressed to humans"?
|
|
||
| - AI / LLM usage does not need to be declared in commit messages via `Co-authored-by`. | ||
| - AI / LLM usage should be disclosed in the pull request description. | ||
| - The use of these tools is only acceptable when done responsibly, and the resulting changes must remain reviewable by humans. |
There was a problem hiding this comment.
What does this mean? Maybe specify:
- PRs must be sized so a human can review them in one sitting.
- The submitter must understand the change well enough to explain and defend it.
| ## LLM / AI Usage Guidelines | ||
|
|
||
| - AI / LLM usage does not need to be declared in commit messages via `Co-authored-by`. | ||
| - AI / LLM usage should be disclosed in the pull request description. |
There was a problem hiding this comment.
Should? Why not must? What do you want this to include? This invites short used AI line.
There was a problem hiding this comment.
Changed to must
| - AI / LLM usage should be disclosed in the pull request description. | ||
| - The use of these tools is only acceptable when done responsibly, and the resulting changes must remain reviewable by humans. | ||
| - Review comments must be addressed properly; pull requests that cannot be reviewed by humans, or that ignore reviewer feedback, may be closed or rejected. | ||
| - Keep the DCO in mind as well, since AI-generated contributions can create copyright and provenance risks. If in doubt, do not include generated code. |
There was a problem hiding this comment.
This undercuts the point of this note. You invite AI contributions but then advise not to use them.
Maybe rephrase to something like
- Signing off means you assert you have the right to submit the code, and that assertion is yours regardless of how the code was produce.
There was a problem hiding this comment.
Yes this is true. The main point of this is to put the liability onto the contributor.
I have rephrased it slightly.
| - The use of these tools is only acceptable when done responsibly, and the resulting changes must remain reviewable by humans. | ||
| - Review comments must be addressed properly; pull requests that cannot be reviewed by humans, or that ignore reviewer feedback, may be closed or rejected. | ||
| - Keep the DCO in mind as well, since AI-generated contributions can create copyright and provenance risks. If in doubt, do not include generated code. | ||
| - All VILLAS-related commits made before June 2026 are, to the best of our knowledge, free of AI-generated code. |
There was a problem hiding this comment.
I don't think this is true. I used a lot of AI code completion even when I was working more actively on VILLAS, and I assume other contributors did as well. I'd just remove this line.
There was a problem hiding this comment.
We discussed to a have a sort of a cut-off date here. Smaller AI-enabled edits like code completion, I would neglect for now. But I would like to have a commit/date after which larger AI-based edits have entered VILLAS
Signed-off-by: Steffen Vogel <steffen.vogel@opal-rt.com>
3 participants
No description provided.