Post

asa
@asap.systems

relatedly, I've needed to set up CI and a binary artifact cache recently for a team (with private code hosted on uni github enterprise without actions), and it's quite the landscape

asa@asap.systems

why am I getting emails about my "ai software factory" man I just wanted good CI

asa
@asap.systems

most "CI" services/frameworks seem to be handling a variety of - integration with code forges for e.g. merge gates - environment setup - task runner - status reporting

0 reposts0 likes

asa
@asap.systems

we use nix, which solves setup and task orchestrator, meaning something like jenkins is massive overkill. there's buildbot-nix but it doesn't seem to integrate into random github enterprise instances... ...in the end I had glm write a several hundred line python script to handle everything

asa
@asap.systems

the builders, cache, and ci service are running in microvms a desktop pc with credentials handled by agenix and it's super straightforward

asa
@asap.systems

ideally this would run on some cloud service instead of a random pc for reliability, but vps prices are slightly ridiculous and "I just want to spin up a build vm occasionally without bankrupting myself" is a _very_ daunting task

Marco Z
@ocramz.bsky.social

serverless? eg gcp cloud run has very long request timeouts (I think an hour nowadays), and with some incantations podman-in-docker can run inside that