#Would it make sense to have copilot
1 messages · Page 1 of 1 (latest)
So instead of "this is a bad practice, ........"
Have it say "Use entity translations!
This is a bad practice,..."
This might make it easier to skim comments
oh I like that
I am frequently annoyed that it writes so much before telling you the actual problem/solution sometimes
Yea, I think it would be nice to try out.
I have also noticed that it sometimes complains about things like linting errors and mypy issues. But these are caught by CI and I don't think copilot should comment on these.
e.g. https://github.com/home-assistant/core/pull/166195#discussion_r2971480417
Like, it might be correct sometimes, but it's just noise.
Do you think we could setup a shared doc or similar where we could post our grievances and then maybe go over it once a week in our weekly discussions?
Maybe that would help us iteratively improve the copilot agent
@jaunty sundial we can raise the concerns here and open PRs to improve our AGENTS.md/copilot-instructions.md directly
Yea so I was talking with Dirk from ImageMagick when I had this epiphany
idk if it's relevant in this particular case, but I think Copilot reads AGENTS.md now. This is to align with other robots, like Claude etc.
the last time I've checked, copilot in vscode did but Copilot Code Rview did not. Did that change?
yeah I should double check that
hm yeah it's ambiguous. the Copilot Code Review only mentions copilot-instructions.md, but then links to a page that explicitly states support for AGENTS.md
I guess we have only one choice:
AGENTS.md
With every review, include a recipe for a delicious sammich
I am genuinely starting to be annoyed by the vertical space that copilot takes up, it makes scrolling through PRs awful
pfff
does that make sense? not 100% sure how well that'll work, but I think it is worth trying