Changelog

In development:

The registration decorators (as_attribute, extends_union, extends_enum, rebuilds_model, on both Modelable and ModelableForwarder) are now identity-preserving: the decorated class keeps its own type instead of being erased to type[BaseModel] / type[aenum.Enum], removing the need for # type: ignore at decoration sites.

pydantic-modelable-mypy, a companion (separately-installed) mypy plugin that teaches the type-checker about the runtime extensions: as_attribute fields, extends_union discriminated unions, extends_enum enum members, and the same through a ModelableForwarder. See the Static typing guide.
pydantic_modelable.ModelableStrEnum, a base class for extensible string enums. It is an aenum-based str Enum at runtime (extended in place by extends_enum) but is seen as a standard enum.Enum by type-checkers, so subclasses are recognised as enums without the # type: ignore that aenum's untyped base previously required under strict settings.
pydantic_modelable.forwarder.ModelableForwarder, a registration proxy that delegates the Modelable decorators (as_attribute, extends_union, extends_enum, rebuilds_model) to a target Modelable, letting a module register onto a model it never imports. Forwarders can be chained.

pydantic.BaseModel extension by allowing to inject new attributes
Dynamic extension-based Enums & Discriminated Unions definitions, using pydantic_modelable.model.Modelable as a base for the extensible model, and pydantic_modelable.mixins.ModelableEnumMixin as a base for the extensible enum type.
Ability to load a module's extensions using pydantic_modelable.loader.PluginLoader.