如何向Pydantic模型传递泛型类型定义，兼顾运行与代码补全？

我使用Python 3.12和Pydantic 2.6.4编写了如下概念验证代码：

from abc import ABC, abstractmethod
from typing import Type
from pydantic import BaseModel
from pydantic_settings import BaseSettings


class Settings[T: BaseModel](BaseSettings):
    name: str
    job: T


class BaseJob[T: BaseModel](ABC):
    def __init__(self, name: str, **kwargs) -> None:
        # self.settings = Settings[T](name=name, **kwargs)  # Happy pylance, but raises: AttributeError: 'BaseModel' object has no attribute '__private_attributes__'
        self.settings = Settings[self.settings_model](name=name, **kwargs)  # Works but pylance raises: Cannot access member "workers" for type "BaseModel"

    @abstractmethod
    def run(self) -> None:
        pass

    @property
    @abstractmethod
    def settings_model(self) -> Type[BaseModel]:
        pass


class JobSettings(BaseModel):
    workers: int = 2


class Job(BaseJob[JobSettings]):
    def __init__(self, name: str, **kwargs) -> None:
        super().__init__(name, job={"workers": 4})

    @property
    def settings_model(self) -> Type[BaseModel]:
        return JobSettings

    def run(self) -> None:
        print(self.settings.name)
        print(self.settings.job.workers)


if __name__ == "__main__":
    Job("A generic job").run()

遇到的矛盾问题：

• 使用Settings[T](name=name, **kwargs)初始化settings时，Pylance代码补全功能正常，但运行程序会抛出AttributeError: 'BaseModel' object has no attribute '__private_attributes__'错误。
• 使用Settings[self.settings_model](name=name, **kwargs)初始化时，程序可正常运行，但Pylance无法识别job的workers成员，提示无法访问BaseModel类型的workers属性。

请问兼顾程序正常运行与代码补全的正确实现方式是什么？是否存在根本性错误？

问题根源 #

泛型类型参数T在BaseJob抽象类中仅作为类型标识，运行时无法直接用于实例化Settings泛型类；而直接用self.settings_model实例化虽然能运行，但丢失了类型推导信息，导致Pylance无法识别具体字段。另外原代码中JobSettings的继承关系写错了，它应该继承BaseModel而非BaseJob，这属于根本性类结构错误，必须修正。

正确实现方式 #

通过为BaseJob的settings属性添加精确类型注解，同时强化settings_model的返回类型约束，即可兼顾运行正确性与代码补全：

from abc import ABC, abstractmethod
from typing import Type
from pydantic import BaseModel
from pydantic_settings import BaseSettings


class Settings[T: BaseModel](BaseSettings):
    name: str
    job: T


class BaseJob[T: BaseModel](ABC):
    def __init__(self, name: str, **kwargs) -> None:
        # 为settings属性添加泛型类型注解，给Pylance提供类型信息
        self.settings: Settings[T] = Settings[self.settings_model](name=name, **kwargs)

    @abstractmethod
    def run(self) -> None:
        pass

    @property
    @abstractmethod
    def settings_model(self) -> Type[T]:  # 将返回类型从Type[BaseModel]改为Type[T]，强化类型约束
        pass


# 修正JobSettings的继承关系
class JobSettings(BaseModel):
    workers: int = 2


class Job(BaseJob[JobSettings]):
    def __init__(self, name: str, **kwargs) -> None:
        super().__init__(name, job={"workers": 4})

    @property
    def settings_model(self) -> Type[JobSettings]:  # 这里可以直接返回具体类型，进一步明确
        return JobSettings

    def run(self) -> None:
        print(self.settings.name)
        print(self.settings.job.workers)


if __name__ == "__main__":
    Job("A generic job").run()

关键修改点 #

1. 给self.settings添加Settings[T]类型注解，让Pylance明确该属性的具体泛型类型，从而正确识别job字段的成员。
2. 将settings_model的返回类型从Type[BaseModel]改为Type[T]，让类型检查器明确该属性返回的就是当前泛型参数对应的模型类。
3. 修正JobSettings的继承关系，确保它继承自BaseModel，符合Pydantic模型的定义规范。

内容的提问来源于stack exchange，提问作者ub-