Skip to content

Config files

Environment variables

Compas lets you forget loading your .env files. All main entry points (mainFn, mainTestFn and mainBenchFn) will load both the .env and .env.local files automatically. The loader that is used is provided by the dotenv package.

Compas advises to check in a ready to go .env.example in your version control or preferably when using all Compas provided tooling, just a .env file without any actual production secrets. This way new contributors can just fire up some commands and verify that they can run your project locally instead of trying to figure out what they need to configure where.

The .env file loader skips keys already present in your environment. This is also used by the .env.local file. It is loaded first so any environment variables present in both .env.local and .env will have the value defined by the former.

To speed up accessing environment variables the @compas/stdlib package provides an exported object environment which is an exact copy of process.env before the callback is called in mainFn. If you use process.env to set values, please don't, you can use refreshEnvironmentCache to make a new copy.


Never check in production secrets in to your version control.

Config loader

Compas also comes with a configuration file loader in @compas/stdlib. This will be used by other Compas features, but also allows you to have a project specific configuration file. You can use this to validate and convert your environment variables for example doing the following transformation:



// config/my-project.js

export function config() {
  return {
    postgres: {
      user: environment.POSTGRES_USER,
      password: environment.POSTGRES_PASSWORD,

Or it can be used by for example Compas to let you specify your migrations directory to make compas migrate more configurable.

The configuration files can be in 2 formats: JSON or ES modules (both .js and .mjs extensions are supported). The ES module should export a synchronous function returning a config data object. The configuration files can be loaded from the project root, the project config directory, or from the OS and user specific config directory.


The config file load function, it caches results based on the provided name and key. It returns promise resolving to an object with the following properties:

const result = await configLoaderGet({ name: "example", location: "project" });

// always equals to the provided name, 'example';
// always equal to the provided location, 'project'

// object, does not exist if no config file is found
// The directory where the config file is found;
// Filename with extension, `example.json` or `example.js` or `example.mjs`

// Object, either the object returned from a config function or a parsed json file.
// This is an empty object if no config file is found;


Delete the config loader cache. This can be used if your tool overwrites the config, or when writing some reload / watch behaviour.