Pydantic Generator

Overview

The Pydantic generator emits Pydantic v2 models (BaseModel subclasses) for each @Schema class and its variants. It honors PydanticMeta inner classes on both @Schema classes and Enum subclasses, and wires top-level Config.pydantic settings into the generated model_config = ConfigDict(...) block.

Config.pydantic options

Top-level Config.pydantic keys are now threaded into every generated model's ConfigDict. Only the keys below are honored — any unrecognized keys are silently ignored so typos don't spuriously enable model_config on otherwise default schemas.

Key	Type	Description
extra	str	Pydantic's extra-field policy. Typical values: "forbid", "allow", "ignore".
validate_assignment	bool	Re-run validation when an attribute is assigned after construction.
frozen	bool	Make models immutable (__hash__ auto-generated).
strict	bool	Enable strict-mode coercion rules.
str_strip_whitespace	bool	Auto-strip whitespace from all string fields.
populate_by_name	bool	Allow populating fields by their declared name even when an alias is set.

String values are properly repr()-escaped, so extra="forbid" emits extra='forbid' in the generated ConfigDict(...).

from schema_gen import Config

config = Config(
    targets=["pydantic"],
    pydantic={
        "extra": "forbid",
        "validate_assignment": True,
        "frozen": False,
        "strict": True,
        "str_strip_whitespace": True,
        "populate_by_name": True,
    },
)

The generator combines these with its default from_attributes=True, producing e.g.:

class User(BaseModel):
    model_config = ConfigDict(
        from_attributes=True,
        extra='forbid',
        validate_assignment=True,
        strict=True,
        str_strip_whitespace=True,
        populate_by_name=True,
    )
    ...

PydanticMeta on schemas

PydanticMeta inner classes inject custom validators, instance methods, and imports into the generated base model (variants never inherit custom code). See the Schema Format document for the full syntax.

Supported attributes:

Attribute	Description
imports	Extra import lines emitted at the top of the file.
raw_code	Verbatim code appended inside the class body (typically @field_validator blocks).
methods	Instance methods appended inside the class body.
validators	Reserved for future validator-specific emission.

PydanticMeta on enums

PydanticMeta can also be attached to Enum subclasses. Today the generator honors the methods attribute, which injects domain methods directly into the generated Python enum class body. This keeps Python and Rust side-by-side in sync when you use enum-level SerdeMeta for the Rust target.

Python 3.12+: Inner classes inside str, Enum or StrEnum are treated as enum members, which causes errors. Wrap with enum.nonmember() to prevent this.

from enum import Enum, nonmember

class OrderStatus(str, Enum):
    PENDING = "pending"
    FILLED = "filled"
    CANCELLED = "cancelled"

    PydanticMeta = nonmember(
        type(
            "PydanticMeta",
            (),
            {
                "methods": """
    def is_terminal(self) -> bool:
        return self in (OrderStatus.FILLED, OrderStatus.CANCELLED)
"""
            },
        )
    )

Generated:

class OrderStatus(Enum):
    PENDING = "pending"
    FILLED = "filled"
    CANCELLED = "cancelled"

    def is_terminal(self) -> bool:
        return self in (OrderStatus.FILLED, OrderStatus.CANCELLED)