Settings and configuration

A database URL, an OpenAI key, the list of origins allowed to call your API: these are inputs to your program, they arrive from outside as strings, and a wrong one should fail loudly. That is the same problem Pydantic solves for request bodies, and pydantic-settings applies the same tool to the environment. You declare a Settings class, it reads environment variables and a .env file, converts them to typed values, and refuses to start if something required is missing.

The alternative, os.environ["DATABASE_URL"] scattered through modules, fails at whatever moment that line first runs, which in a web service is often the first request after a deploy. A settings class moves that failure to process start, where it belongs.