Just an addendum to this older post: In runtime config (config/runtime.exs), you can use your modules, as they have already been compiled. No applications are running at that point, but for example you can write config helpers like this:

• config/runtime.exs · b1cf53462a3fa34369eaa06494754c7ae38aed2a · CodeStats / code-stats · GitLab
• lib/code_stats/config_helpers.ex · b1cf53462a3fa34369eaa06494754c7ae38aed2a · CodeStats / code-stats · GitLab

Your modules are compiled using the generated config, not the other way around. So when the config is being processed, you cannot use the modules because they have not been compiled. Otherwise there would be cyclic dependencies.

What you can do is use MFA (module, function, arguments) syntax to refer to functions that you will later run at runtime. So for example callback_function: {MyApp.MyModule, :function, [1, "foo"]} is fine.

As for the release config files that are evaluated during runtime at system startup, I think @LostKobrakai’s post covers that case. But I’m not an expert on them.

Both answers come down to: Your App depends on your configuration, so your configuration cannot depend on your app.