ml_switcheroo.config¶

Runtime Configuration Store. Supports Dynamic Defaults, TOML loading, and Framework Flavours.

This module resolves default Source and Target frameworks by querying the ui_priority of registered adapters, ensuring the logic is completely agnostic to the specific libraries installed.

Attributes¶

`tomllib`
`T`

Classes¶

RuntimeConfig

Global configuration container for the translation engine.

Functions¶

`get_framework_priority_order`(→ List[str])	Returns a list of framework keys sorted by UI Priority.
`parse_cli_key_values`(→ Dict[str, Any])	Parses a list of 'key=value' strings into a dictionary with type inference.

Module Contents¶

ml_switcheroo.config.tomllib = None¶

ml_switcheroo.config.T¶

ml_switcheroo.config.get_framework_priority_order() → List[str][source]¶

Returns a list of framework keys sorted by UI Priority.

The sort order is determined by the ui_priority attribute of the registered FrameworkAdapter. Lower numbers appear first (Source default).

If no priority is defined, defaults to 999.

Returns:: Sorted list of framework identifiers (e.g. [‘torch’, ‘jax’]).
Return type:: List[str]

class ml_switcheroo.config.RuntimeConfig(/, **data: Any)[source]¶

Bases: pydantic.BaseModel

Global configuration container for the translation engine.

source_framework: str = None¶

target_framework: str = None¶

source_flavour: str | None = None¶

target_flavour: str | None = None¶

strict_mode: bool = None¶

intermediate: str | None = None¶

enable_graph_optimization: bool = None¶

enable_import_fixer: bool = None¶

plugin_settings: Dict[str, Any] = None¶

plugin_paths: List[pathlib.Path] = None¶

validation_report: pathlib.Path | None = None¶

classmethod validate_framework(v: str) → str[source]¶

Ensures the framework is registered in the system.

Parameters:: v (str) – The framework key to validate.
Returns:: The normalized framework key.
Return type:: str
Raises:: ValueError – If the framework is not registered and not a placeholder.

property effective_source: str¶

Returns the resolved source framework key.

Prioritizes source_flavour if present, otherwise returns source_framework.

Returns:: The active source framework key.
Return type:: str

property effective_target: str¶

Returns the resolved target framework key.

Prioritizes target_flavour if present, otherwise returns target_framework.

Returns:: The active target framework key.
Return type:: str

parse_plugin_settings(schema: Type[T]) → T[source]¶

Validates plugin settings against a Pydantic model.

Parameters:: schema (Type[T]) – The Pydantic model class defining the settings schema.
Returns:: An instance of the schema populated with validated settings.
Return type:: T
Raises:: ValueError – If validation against the schema fails.

Loads configuration from pyproject.toml, overriding with CLI arguments. Defaults are calculated dynamically via factory methods if not found.

Parameters:

source (Optional[str]) – Override source framework.
target (Optional[str]) – Override target framework.
source_flavour (Optional[str]) – Override source flavour.
target_flavour (Optional[str]) – Override target flavour.
strict_mode (Optional[bool]) – Override strict mode setting.
intermediate (Optional[str]) – Override intermediate representation mode.
enable_graph_optimization (Optional[bool]) – Override fusion/optimization setting.
plugin_settings (Optional[Dict]) – Additional plugin settings.
validation_report (Optional[Path]) – Override validation report path.
search_path (Optional[Path]) – Directory path to search for config file.

Returns:

The fully resolved configuration object.

Return type:

RuntimeConfig

ml_switcheroo.config.parse_cli_key_values(items: List[str] | None) → Dict[str, Any][source]¶

Parses a list of ‘key=value’ strings into a dictionary with type inference.

Parameters:: items (Optional[List[str]]) – List of strings in “key=value” format.
Returns:: Dictionary with parsed values (bool, int, float, or str).
Return type:: Dict[str, Any]