Hello devs,
I've been using a local sqlite3 cache of all ghc tickets/mrs and
some more on top for a while now with agents as interface to the
data.
While it's a shame that this works so much better than gitlabs
search it is what it is. So I want to share what I've been using
to get
a better answer faster for questions like:
* Is there already a ticket about <thing> already.
* I remember <person> wrote a comment about <thing>,
what MR/ticket was this on.
* What ghc versions have the fix to <ticket> already.
I use a small haskell tool + skill file to create/update the data
which can be found here: https://github.com/AndreasPK/fetch-gitlab
and I want to encourage people to try it. And if they find it
worse than what they use today complain to me so I can make it
better.
I don't think my version of this is particularly perfect (It's a
lot of generated and only lightly reviewed code) . But I hope to
avoid everyone vibing their own version and instead sharing one
implementation and improving that one together.
If you want to try it clone: https://github.com/AndreasPK/fetch-gitlab.git
cabal install, ideally also install
the sqlite3 binary~/.fetch-gitlab/gitlab-tokenfetch-gitlab to fetch the latest
MRs/Ticketsskills/fetch-gitlab/ ~/.claude/skills/ to
install the claude skill (adjust if you use a different setup)