Your personal operating system for career transition.

Build a Type-Safe Config Loader from Environment Variables

AI services are configured through environment variables: API keys, model names, timeout values, rate limits, feature flags. The naive approach — calling os.environ.get("MODEL") scattered throughout the codebase — creates invisible coupling between config layout and application code, and lets the service start with missing or invalid config that will fail on the first request.

What to build

Implement a ServiceConfig Pydantic model and a load_config() factory function that:

1. Reads all config from environment variables — use pydantic-settings BaseSettings or standard BaseModel with model_validator.
2. Validates types and constraints — max_tokens must be between 1 and 8192, temperature between 0.0 and 2.0, api_key must be non-empty.
3. Provides defaults for non-critical settings — timeout_s defaults to 30.0, max_concurrent defaults to 5.
4. Raises a descriptive error on startup if required values are missing — do not let the service boot with api_key = None.
5. Supports a from_dict class method for testing — so tests can pass config without touching os.environ.

Config fields to implement

Why this matters

Missing or malformed config is one of the most common production failure modes. A service that validates its entire config at boot time fails fast with a clear error — not 10 minutes into the first request with a cryptic AttributeError. The from_dict method makes tests config-hermetic: no monkeypatching environment variables.

Constraints

• Use Pydantic v2. Do not use os.environ inside the model — inject values via from_dict for tests.
• Raise a ValueError with a clear message if api_key is missing or too short.