[nfc] update CONTRIBUTING.md to add basic LLM policy#22591
Conversation
ferdymercury
left a comment
There was a problem hiding this comment.
Thanks!
Maybe also mention that copy pasting could lead to copyright infringement eg if LLM is copying verbatim from a GPL project into LGPL
Test Results0 tests 0 ✅ 0s ⏱️ Results for commit 3771838. ♻️ This comment has been updated with latest results. |
guitargeek
left a comment
There was a problem hiding this comment.
Very nice, I think this is well-written and captures the essential etiquette!
Going further, we might spell out precisely what it means "to disclose whether the proposed changes (code or otherwise) were automatically generated", maybe with templates for such disclosures that can be added to commit messages and PR descriptions.
|
Thanks for this initiative. Before merging, even if I do not see huge hurdles ahead, I was thinking to discuss this a bit broadly within the core developers. |
vepadulano
left a comment
There was a problem hiding this comment.
Thanks! A couple of minor suggestions from my side
| Please ping people :wave: should you not get timely feedback, for instance with `@root-project/core ping!` | ||
|
|
||
| > [!IMPORTANT] | ||
| > For compatibility with our CI, your should create your PR with a **branch name different from `master`**. |
There was a problem hiding this comment.
| > For compatibility with our CI, your should create your PR with a **branch name different from `master`**. | |
| > For compatibility with our CI, you should create your PR with a **branch name different from `master`**. |
|
|
||
| Thank you for reading this; and even more: thank you :bouquet: for considering to contribute! | ||
|
|
||
| ## LLM Disclosure |
There was a problem hiding this comment.
I don't know if this is a term already being used elsewhere, but to be more general we could think about calling it "AI-assisted coding disclosure".
| Please avoid proposing LLM-generated changes as-is without a full review and understanding, as this disrespects | ||
| the reviewers' time and may cause your future contributions to be automatically rejected if reiterated. | ||
|
|
||
| Likewise, PR comments and any sort of engagement with other contributors must not be parroting LLMs or chatbots. |
There was a problem hiding this comment.
Is "parroting" a universally understood term?
| Likewise, PR comments and any sort of engagement with other contributors must not be parroting LLMs or chatbots. | |
| Likewise, PR comments and any sort of engagement with other contributors must not be a simple copy of an LLM's or chatbot's output. |
This Pull request:
Adds a section about LLM usage to CONTRIBUTING.md. It is not meant as a fully-fledged "AI policy" but just some basic common-sense guidelines that are nonetheless good to spell out.
Also adds a section about branch names, since opening a PR with a branch named
masterconflicts with our CI.