pattern-agentic-settings 2.0.0.dev0

Pydantic-powered config management

Pattern Agentic Settings

A settings class based on pydantic-settings that facilitates:

• printing all settings on startup (redacts senstive attrs)
• easy loading of .env files via an env variable
• optionally load app version from importlib (sourced from pyproject.toml)
• optional hot reload using watchfiles

Installation

pip install pattern_agentic_settings[hotreload]

Note that without the hotreload extension:

pip install pattern_agentic_settings

watchfiles is not pulled in, and dot envs will not auto-reload.

Example

class Settings(PABaseSettings):
    model_config = SettingsConfigDict(
        env_prefix="MYAPP_"
    )
    worker_count: int

# expects MYAPP_WORKER_COUNT 
# if MYAPP_DOT_ENV points to a file, will try to load vars from it
# throws an error if not defined
settings = Settings.load('my_app')

# This will watch the env file on disk and reload values if it changes
# This must be called in the context of an existing async runloop
# it also requires the hotreload extension (see Installation)

settings.watch_env_file()

Exepcted output:

My App v1.0.0
Configuration:
  WORKER_COUNT: 3

And on failure:

RuntimeError: Configuration validation failed:

      Missing required configuration fields:
        - worker_count

Tests

To run the tests the package must be installed in edit mode:

uv pip install -e .

# or to test hotreload
uv pip install -e .[hotreload]

After that:

# Base test
uv run pytest tests/test_base.py

# All tests
uv run pytest