litestar-dto（莱特星 DTO）

DTO

Use this skill when request and response payloads need explicit shape control, field-level policy, or DTO factory customization.

Execution Workflow

1. Start with the handler contract: define inbound (dto) and outbound (return_dto) behavior explicitly.
2. Choose DTO factory type for your model ecosystem (for example, DataclassDTO, SQLAlchemyDTO, or plugin-provided DTOs).
3. Define policy with DTOConfig (exclude, rename, nesting depth, unknown-field handling, PATCH behavior).
4. Apply DTOs at the correct layer (handler, controller, router, or app) and rely on closest-layer precedence.
5. Use DTOData[T] in write/update handlers when you need controlled instantiation and patching semantics.
6. Validate wrappers and envelopes (Response[T], pagination, generic wrappers) to ensure DTO transformation is still applied.
7. Implement a custom AbstractDTO only when factory DTOs cannot express required behavior.

Implementation Rules

• Keep read and write DTOs separate when policies differ.
• Keep DTO configuration explicit and reviewable; avoid implicit field exposure.
• Treat nested serialization depth as an API and performance control, not a default.
• Reject unknown fields (forbid_unknown_fields=True) in stricter API surfaces.
• Use DTOs to enforce immutability and server-owned fields (id, audit fields).
• Keep handler logic focused on business behavior, not ad-hoc payload transformation.

DTO Basics: Layering and Parameters

Litestar exposes two DTO parameters on each app layer:

• dto: inbound parsing for handler data; if return_dto is not set, this is also used for outbound serialization.
• return_dto: outbound serialization policy only.

DTOs can be declared on handler, controller, router, or application. The DTO closest to the handler in the ownership chain applies.

Common pattern:

• dto=WriteDTO to parse inbound payloads.
• return_dto=ReadDTO to serialize outbound payloads.
• return_dto=None on handlers where DTO serialization should be disabled (for example, DELETE -> None).

DTOConfig Policy Guide

Use DTOConfig to define stable contract rules:

• exclude={...}: remove fields (supports nested paths, including list-item paths such as "pets.0.id").
• rename_fields={...}: rename specific fields.
• rename_strategy="camel" (or callback): apply systematic renaming.
• max_nested_depth: control nested parsing/serialization depth (1 default shown in docs).
• forbid_unknown_fields=True: reject extra payload fields instead of silently ignoring them.
• partial=True: PATCH-friendly DTO behavior.
• leading_underscore_private=False only when you intentionally want underscore-prefixed fields treated as public.
• experimental_codegen_backend=False to selectively disable codegen backend behavior.

Important behavior:

• Explicit rename_fields mappings are not further transformed by rename_strategy.

Field Marking and Access Control

Use dto_field(...) metadata to mark model fields with DTO semantics:

• "private": not parsed from inbound data and never serialized in outbound data.
• "read-only": not parsed from inbound data.

Also note:

• Fields with leading underscores are implicitly private by default.

DTOData[T]: Controlled Create and Update Flows

Use DTOData[T] for create/update handlers that need controlled model mutation:

• data.create_instance(...) creates model instances from validated input and can inject server-side values.
• Nested values can be supplied with double-underscore paths (for example, address__id=...) when excluded nested fields must still be set.
• data.update_instance(existing) supports partial updates cleanly for PATCH workflows when partial=True.

PATCH pattern:

• Exclude immutable/server-owned fields (for example, id) in DTO config.
• Use DTOData.update_instance() to apply only submitted fields.

Wrapper and Envelope Handling

DTO factory types can transform data inside supported wrappers:

• Generic wrapper dataclasses (for example, WithCount[T]) when one type parameter maps to DTO-supported model data.
• Litestar pagination wrappers such as ClassicPagination[T] (DTO applies to items).
• Response[T] wrappers (DTO applies to content).

Performance Notes

DTO codegen backend:

• Introduced in 2.2.0, stabilized and enabled by default in 2.8.0.
• Can be toggled per DTO with DTOConfig(experimental_codegen_backend=...).

Use selective override only when debugging compatibility or behavior differences.

AbstractDTO and Custom DTO Classes

Use built-in DTO factories first. Create custom DTO classes only when required behavior is not expressible via existing DTO factories + DTOConfig.

When implementing custom DTOs, you must implement AbstractDTO methods:

• generate_field_definitions(model_type): yield DTOFieldDefinition objects for DTO-visible fields.
• detect_nested_field(field_definition): return whether the field represents nested model data.

This is the minimal protocol Litestar requires for custom DTO implementations.

Example Patterns

from litestar import get, patch
from litestar.dto import DTOConfig, DTOData, DataclassDTO

class UserReadDTO(DataclassDTO[User]):
    config = DTOConfig(exclude={"password_hash"})

class UserPatchDTO(DataclassDTO[User]):
    # Patch DTO excludes immutable id and accepts partial payloads.
    config = DTOConfig(exclude={"id"}, partial=True)

@get("/users/{user_id:int}", return_dto=UserReadDTO)
async def get_user(user_id: int) -> User:
    return ...

@patch("/users/{user_id:int}", dto=UserPatchDTO, return_dto=UserReadDTO)
async def patch_user(user_id: int, data: DTOData[User]) -> User:
    user = ...
    return data.update_instance(user)

from litestar.dto import DTOConfig
from litestar.plugins.sqlalchemy import SQLAlchemyDTO

class PublicUserDTO(SQLAlchemyDTO[User]):
    config = DTOConfig(exclude={"password_hash"})

Validation Checklist

• Confirm dto and return_dto are intentionally selected (or explicitly disabled) per handler.
• Confirm layer precedence does not accidentally override handler-level DTO intent.
• Confirm exclude/rename/nesting config generates expected payload shape.
• Confirm unknown field behavior matches API strictness (forbid_unknown_fields).
• Confirm private/read-only/underscore fields are not writable by clients.
• Confirm PATCH handlers use partial=True + DTOData.update_instance() where appropriate.
• Confirm wrapper responses (Response, pagination, generic envelopes) still apply DTO transformations.
• Confirm OpenAPI output matches the DTO-shaped contract.

Cross-Skill Handoffs

• Use litestar-dataclasses, litestar-plugins, or litestar-custom-types for model ecosystem specifics.
• Use litestar-databases when DTO behavior couples to SQLAlchemy or Piccolo persistence design.
• Use litestar-openapi to verify generated schema accuracy after DTO changes.
• Use litestar-requests and litestar-responses for transport-level behavior outside DTO transformation.

Litestar References

• https://docs.litestar.dev/2/usage/dto/index.html
• https://docs.litestar.dev/2/usage/dto/0-basic-use.html
• https://docs.litestar.dev/2/usage/dto/1-abstract-dto.html
• https://docs.litestar.dev/2/usage/dto/2-creating-custom-dto-classes.html