ADR 0002: Use Pydantic Settings for Configuration¶
Status¶
Accepted
Date¶
2026-10-13
Context¶
The server needs configuration from multiple sources:
- Environment variables (12-factor app)
- .env files for local development
- Sensible defaults for optional settings
- Type validation and IDE support
Decision¶
Use pydantic-settings with BaseSettings for all configuration.
Consequences¶
Positive¶
- Type-safe configuration with validation
- Automatic
.envfile loading - IDE autocomplete via Pydantic models
lru_cachefor singleton settings instanceextra="ignore"prevents crashes from unknown env vars
Negative¶
- Pydantic dependency (already required for tool schemas)
- Settings are global singleton (harder to test with different configs)
Alternatives Considered¶
- python-dotenv + os.environ — No type safety, manual validation
- dynaconf — More features but heavier, less Pythonic
- Custom config class — Reinventing the wheel