The cache feature sounds interesting but I’m not sure how to use it
I’m dealing with an Elixir app that has some very happy setup on weak machines: Install Erlang, install Elixir, download all deps, build the PLT (mix dialyzer.build). PLT alone on Codex VMs can take 5-10 minutes, so the full setup often takes 13 minutes or longer. The PLT is needed to do statical analysis across the app, so for the agent to do things like lint.
In Jules or GitHub based runners once setup it caches the entire environment state so the work, I was hoping it would do the same on Codex but even with the new caching, my tasks still take 15 minutes for a few line changes due to the heavy setup
Can we do something for Codex to properly cache certain things? What exactly is cached from the setup script, and how can I avoid Codex running the heavy bits again and again?
