What problem does it solve? Litestar applications need a consistent, typed way to load environment variables into configuration without scattering os.environ reads across handlers or re-parsing env vars on every request. This Skill provides the patterns for dataclass-based or pydantic-settings-based config with cached factories and app-state wiring. ## Core Features & Use Cases - Two settings patterns: Choose between zero-dependency @dataclass(frozen=True) + get_env() for fresh projects, or pydantic_settings.BaseSettings when Pydantic is already in the dependency graph. - Cached settings factories: Use @lru_cache(maxsize=1) on get_settings() so configuration is evaluated once per process and shared across all call sites. - Lazy materialization with PEP 562: Defer DB pools, Redis clients, and Channels backends until first attribute access so Docker builds, CLI commands, and test suites do not trigger import-time connections. - Use Case: You are bootstrapping a new Litestar API and need DATABASE_URL, REDIS_URL, and APP_SECRET_KEY loaded as typed, immutable settings with test-friendly overrides via get_settings.cache_clear(). ## Quick Start Ask the AI to create a typed Litestar settings module with a cached get_settings factory that loads DATABASE_URL and REDIS_URL from environment variables.
