Pydantic V1迁移V2遇配置错误，替换Config无效求助

核心问题分析 #

你遇到的「Valid config keys have changed in V2」警告，本质是V2彻底移除了V1中class Config的配置方式，同时部分核心类（比如BaseConfig）、配置项的定义逻辑也发生了变化。下面分模块给出精准修复代码：

1. 配置类（Settings）修复 #

V2中不再使用BaseConfig，读取环境变量需要继承BaseSettings，且环境文件配置直接通过model_config声明：

from pydantic_settings import BaseSettings  # 注意导入路径变更

class Settings(BaseSettings):
    DATABASE_URL: str
    MONGO_INITDB_DATABASE: str

    ACCESS_TOKEN_EXPIRES_IN: int
    REFRESH_TOKEN_EXPIRES_IN: int
    JWT_ALGORITHM: str

    CLIENT_ORIGIN: str

    JWT_PRIVATE_KEY: str
    JWT_PUBLIC_KEY: str

    model_config = {
        "env_file": "./env",
        "extra": "ignore"  # 可选，忽略.env中未定义的变量
    }

settings = Settings()

注意：需要单独安装pydantic-settings包，执行pip install pydantic-settings

2. 模型类修复（含json_encoders替代方案） #

V2中class Config必须替换为model_config（用ConfigDict或字典格式），且原json_encoders需要移到model_config内。以下提供两种适配方案：

方案一：沿用类级别的序列化配置（对应原json_encoders逻辑）

from typing import List
from bson import ObjectId
from pydantic import BaseModel, Field, ConfigDict


class OID(str):
    @classmethod
    def __get_validators__(cls):
        yield cls.validate

    @classmethod
    def validate(cls, v):
        try:
            return ObjectId(str(v))
        except Exception:
            raise ValueError("not a valid ObjectId")


class MyBaseModel(BaseModel):
    def __hash__(self):
        return hash((type(self),) + tuple(self.__dict__.values()))


class Projects(MyBaseModel):
    model_config = ConfigDict(
        arbitrary_types_allowed=True,
        populate_by_name=True,
        json_encoders={ObjectId: lambda oid: str(oid)}  # 原json_encoders移至此处
    )

    title: str = Field(description='titulo', alias='title')
    imagePath: str = Field(description='path image', alias='imagePath')
    url_project: str = Field(default='', description='path image', alias='url')


class User(BaseModel):
    model_config = ConfigDict(
        arbitrary_types_allowed=True,
        populate_by_name=True,
        json_encoders={ObjectId: lambda oid: str(oid)}
    )

    name: str
    login: str
    projects: List[Projects] = Field(alias='projects')


class CreateUserSchema(User):
    password: str
    passwordConfirm: str
    verified: bool = False


class LoginUserSchema(BaseModel):
    name: str
    password: str


class UserResponseSchema(User):
    id: str


class UserResponse(BaseModel):
    status: str
    user: UserResponseSchema

方案二：使用V2推荐的@field_serializer装饰器（更灵活）

如果只需针对特定字段做序列化，推荐用这种方式替代类级别的json_encoders：

from pydantic import field_serializer

# 在需要序列化ObjectId的模型类中添加如下代码
class User(BaseModel):
    model_config = ConfigDict(
        arbitrary_types_allowed=True,
        populate_by_name=True
    )

    id_user: ObjectId = Field(description='user id', alias='_id')
    name: str
    login: str
    projects: List[Projects] = Field(alias='projects')

    @field_serializer('id_user')
    def serialize_oid(self, value: ObjectId) -> str:
        return str(value)

3. 关键注意事项 #

• 必须删除所有遗留的class Config(BaseConfig)代码，这是警告的直接来源
• V2中BaseConfig已被移除，所有模型配置都通过model_config声明
• 若使用alias字段，需确保model_config中开启populate_by_name=True，否则别名映射会失效

内容的提问来源于stack exchange，提问作者learn.boy