How do you prefer to load configuration values in your backend development?

The Config class singleton approach is a solid choice for standardized runtime access. It’s easier than having process.env calls all over your code, which complicates testing. But let's not forget the challenge of how those values get onto the machine. If you're still using a .env file that gets passed around, you could run into secrets management issues. A secrets manager that populates the environment before your app starts could solve that. This setup, like what keyenv.dev offers, injects per-environment secrets at runtime, making everything cleaner.

For maximum flexibility, I recommend implementing overrides. Have hardcoded values as a fallback, static configs as default, and environment variables take precedence but can be overridden by other sources like a database or web services. Just make sure you include validation for those values. It can get complicated, especially with regional and specific country settings, so having a well-structured config system is essential.

Check out varlock.dev! It's a comprehensive configuration toolkit that comes with a lot of useful features to streamline your configuration management.

Most applications load their config only once at startup from environment variables, files, or databases, and then export a typed object. It's best to avoid scattering process.env calls throughout the code. Instead, keep it all in a dedicated configuration service—static values can stay as constants, while dynamic ones come from that config service.

Using environment variables is great for simplifying deployments across different environments. They help manage configurations like service URLs and async pool sizes. On the other hand, pulling configurations from a database is powerful for instance-specific settings that can evolve over time. In short, use environment variables for system bootstrapping and database values for customer-specific configurations. Hardcoding values? Major architectural no-no.