lang_tools.config.sample_config

Sample config - shape-only definition of settings.

This module is the reference implementation of the Config side of the Config / Params pattern. It lives in src/lang_tools/config/ and defines only the shape of settings: typed fields, default factories, and validation. It never reads environment variables, files, or any runtime state. All of that belongs in the paired Params class (src/lang_tools/params/sample_params.py).

Pattern rules for Config models: * Extend BaseModelKwargs, not plain BaseModel. * Use SecretStr for any sensitive value (API keys, passwords, tokens). Pydantic masks the value in repr and log output automatically. * Use an optional kwargs: dict field (default empty) when the config will be forwarded to a third-party constructor via to_kw(). * Never import or call os, load_dotenv, or anything that reads runtime state.

See Also

SampleParams - the companion Params class that loads actual values. docs/guides/params_config.md - full guide with examples and rationale.

Classes:

• NestedModel –

  Nested config model.

• SampleConfig –

  Sample config - reference shape for a config model.

NestedModel

Bases: BaseModelKwargs

Nested config model.

Demonstrates that config models can be composed. A parent config holds nested models as regular Pydantic fields; they round-trip cleanly through to_kw() and JSON serialisation.

SampleConfig

Bases: BaseModelKwargs

Sample config - reference shape for a config model.

Attributes:

• some_int (int) –

  A plain integer setting. The value appropriate for each deployment environment is determined by the paired SampleParams class.

• nested_model (NestedModel) –

  Demonstrates config composition: a parent config holds a NestedModel sub-config as a typed field.

• secret_api_key (SecretStr) –

  A sensitive value stored as SecretStr. Pydantic masks it automatically in repr / str output. To obtain the raw string call config.secret_api_key.get_secret_value().

• kwargs (dict) –

  Extra arguments merged into the top-level dictionary when to_kw() is called. Use this to forward arbitrary keyword arguments to a third-party constructor without listing every one explicitly.

Example

Always produce a SampleConfig via the paired SampleParams class, never by hand-constructing it in application code::

from lang_tools.params.env_type import EnvType
from lang_tools.params.sample_params import SampleParams

params = SampleParams(env_type)
config = params.to_config()
# access the secret only when truly needed:
raw_key = config.secret_api_key.get_secret_value()