Added LLM guidelines to the documentation - #2512
Conversation
felixhekhorn
left a comment
There was a problem hiding this comment.
- There is the case of dependabot - which is clearly no LLM, but also not a human. I'm still quite in favor of allowing that since it fixes known security problems with a simple (typically few line) change in a lock file. Should we mention it explicitly someway?
- There is the case of pre-commit: we already removed the
autoupdatefrom all repos (I think). Should we explicitly discourageautofix(andautoupdate)?
|
They changed the interface ever so slightly and I am now many years old, so I manage to write my reply to you as the commit msg instead of here, lol
You know my opinion on dependabot is quite different :_) But that said, it is not LLM-generated and it is not prompted by a human anyway so I think it doesn't belong here. If there were a case of "we need to bump this library because of this specific feature" I would prefer that to be in the commit rather than simply "bump to version x".
What do you mean? In general the changes should come "pre-committed" in the PRs (but again, a separate problem) |
yes, but for lazy people you can also let the CI (= a bot) do that job for you. We could add a sentence á là "We discourage the use of |
Co-authored-by: Felix Hekhorn <felixhekhorn@users.noreply.github.com>
Co-authored-by: Juan M. Cruz-Martinez <juacrumar@lairen.eu>
2022c59 to
eca8706
Compare
|
For me they are a different category as they don't add features nor they solve bugs. They increase the noise, as one has to go one commit down in the blame to find who introduced some particular change, but they don't hide attribution. I'm generally against anything that touches the code without a human at the helm, but not enough to write it in stone I think. |
As discussed in Morimondo. Let me know if you'd like to modify anything.