Pydantic interop¶

The didactic-pydantic distribution provides two adapters for moving Models between didactic and Pydantic v2.

Installation¶

pip install didactic-pydantic

The package contributes didactic.pydantic to the namespace.

from_pydantic¶

Convert a pydantic.BaseModel subclass into a dx.Model subclass:

from pydantic import BaseModel, Field
from didactic.pydantic import from_pydantic


class PydUser(BaseModel):
    id: str
    email: str = Field(description="primary contact")


User = from_pydantic(PydUser)        # dx.Model subclass
u = User(id="u1", email="a@b.c")

Metadata that survives:

field annotations, including Annotated[T, ...] constraint metadata (Ge, Le, MinLen, ...)
defaults and factories
aliases (alias, validation_alias, serialization_alias collapse to a single alias)
descriptions, examples
the deprecated flag
json_schema_extra lands under dx.field(extras=...)

Pydantic features that do not flow across:

@field_validator and @model_validator methods. Re-author with @dx.validates (per-field) or axioms (cross-field).
@computed_field. Re-author with @dx.computed or @dx.derived.
Discriminated unions. Re-author as dx.TaggedUnion.

to_pydantic¶

The inverse direction. Use it to expose a dx.Model to FastAPI, OpenAPI generators, or any other Pydantic-shaped tool:

import didactic.api as dx
from didactic.pydantic import to_pydantic


class User(dx.Model):
    id: str
    email: str = dx.field(description="primary contact")


PydUser = to_pydantic(User)          # pydantic.BaseModel subclass
PydUser(id="u1", email="a@b.c")

The conversion is cached: a second call with the same input class returns the same Pydantic class.

Round-tripping¶

from_pydantic(to_pydantic(M)) and to_pydantic(from_pydantic(P)) both round-trip the cross-format-stable metadata exactly. Computed fields and discriminated unions are dropped on the dx.Model -> Pydantic direction; everything else survives.