Django Ninja Service Objects

An implementation of the django-service-objects philosophy for Django Ninja with Pydantic validation.

Encapsulate your business logic in reusable, testable service classes.

Installation

pip install django-ninja-service-objects

Add to your Django settings:

# settings.py
INSTALLED_APPS = [
    ...
    'ninja_service_objects',
    ...
]

Usage

from ninja import Schema
from ninja_service_objects import Service

class CreateUserInput(Schema):
    email: str
    name: str

class CreateUserService(Service[CreateUserInput, User]):
    schema = CreateUserInput

    def process(self) -> User:
        return User.objects.create(
            email=self.cleaned_data.email,
            name=self.cleaned_data.name,
        )

    def post_process(self) -> None:
        # Called after successful transaction commit
        send_welcome_email(self.cleaned_data.email)

# In your view
user = CreateUserService.execute({"email": "test@example.com", "name": "Test"})

Using Pydantic BaseModel with Custom Validators

You can also use Pydantic's BaseModel directly for more complex validation:

from pydantic import BaseModel, EmailStr, field_validator, model_validator
from ninja_service_objects import Service

class RegisterUserInput(BaseModel):
    email: EmailStr
    password: str
    password_confirm: str

    @field_validator("password")
    @classmethod
    def password_min_length(cls, v: str) -> str:
        if len(v) < 8:
            raise ValueError("Password must be at least 8 characters")
        return v

    @model_validator(mode="after")
    def passwords_match(self) -> "RegisterUserInput":
        if self.password != self.password_confirm:
            raise ValueError("Passwords do not match")
        return self

class RegisterUserService(Service[RegisterUserInput, User]):
    schema = RegisterUserInput

    def process(self) -> User:
        return User.objects.create_user(
            email=self.cleaned_data.email,
            password=self.cleaned_data.password,
        )

Using ModelField for Django Model Instances

Use ModelField and MultipleModelField to validate Django model instances as service inputs:

from pydantic import BaseModel
from ninja_service_objects import Service, ModelField, MultipleModelField

class TransferOwnershipInput(BaseModel):
    from_user: ModelField[User]
    to_user: ModelField[User]
    posts: MultipleModelField[Post]

class TransferOwnershipService(Service[TransferOwnershipInput, None]):
    schema = TransferOwnershipInput

    def process(self) -> None:
        for post in self.cleaned_data.posts:
            post.author = self.cleaned_data.to_user
            post.save()

By default, ModelField rejects unsaved model instances (objects without a primary key). To allow unsaved instances:

from typing import Annotated

class MyInput(BaseModel):
    user: Annotated[User, ModelField(allow_unsaved=True)]
    items: Annotated[list[Item], MultipleModelField(allow_unsaved=True)]

Features

• Pydantic validation for inputs
• Automatic database transaction handling
• post_process hook for side effects (runs after commit)
• Type-safe with generics support
• ModelField and MultipleModelField for Django model instance validation

Design Decisions

Why Pydantic instead of Django Forms?

The original django-service-objects uses Django Forms for validation. Since Django Ninja already uses Pydantic for request/response schemas, this library uses Pydantic to:

• Avoid mixing two validation systems in the same project
• Reuse your existing Django Ninja schemas as service inputs
• Get better type hints and IDE support

API Compatibility

We maintain familiar patterns from django-service-objects:

• cleaned_data - Access validated input data (same naming as Django forms/original library)
• process() - Override this with your business logic
• post_process() - Runs after successful transaction commit (for emails, notifications, etc.)
• execute() - Class method entry point that handles validation and transactions

What's Different

django-service-objects	ninja-service-objects
Django Forms validation	Pydantic validation
service_clean() method	Pydantic validators
Form fields	Pydantic BaseModel
is_valid() + execute()	Single execute() call

Configuration

Transaction Control

class MyService(Service[MyInput, MyOutput]):
    schema = MyInput
    db_transaction = False  # Disable automatic transaction wrapping
    using = "other_db"      # Use a different database alias

License

MIT