Skip to content

Response Schema

Django Ninja allows you to define the schema of your responses both for validation and documentation purposes.

Imagine you need to create an API operation that creates a user. The input parameter would be username+password, but output of this operation should be id+username (without the password).

Let's create the input schema:

from ninja import Schema

class UserIn(Schema):
    username: str
    password: str"/users/")
def create_user(request, data: UserIn):
    user = User(username=data.username) # User is django auth.User
    # ... return ?

Now let's define the output schema, and pass it as a response argument to the decorator:

from ninja import Schema

class UserIn(Schema):
    username: str
    password: str

class UserOut(Schema):
    id: int
    username: str"/users/", response=UserOut)
def create_user(request, data: UserIn):
    user = User(username=data.username)
    return user

Django Ninja will use this response schema to:

  • convert the output data to declared schema
  • validate the data
  • add an OpenAPI schema definition
  • it will be used by the automatic documentation systems
  • and, most importantly, it will limit the output data only to the fields only defined in the schema.

Nested objects

There is also often a need to return responses with some nested/child objects.

Imagine we have a Task Django model with a User ForeignKey:

from django.db import models

class Task(models.Model):
    title = models.CharField(max_length=200)
    is_completed = models.BooleanField(default=False)
    owner = models.ForeignKey("auth.User", null=True, blank=True)

Now let's output all tasks, and for each task, output some fields about the user.

from typing import List
from ninja import Schema

class UserSchema(Schema):
    id: int
    first_name: str
    last_name: str

class TaskSchema(Schema):
    id: int
    title: str
    is_completed: bool
    owner: UserSchema = None  # ! None - to mark it as optional

@api.get("/tasks", response=List[TaskSchema])
def tasks(request):
    queryset = Task.objects.select_related("owner")
    return list(queryset)

If you execute this operation, you should get a response like this:

        "id": 1, 
        "title": "Task 1",
        "is_completed": false,
        "owner": {
            "id": 1,
            "first_name": "John",
            "last_name": "Doe",
        "id": 2, 
        "title": "Task 2",
        "is_completed": false,
        "owner": null


Instead of a nested response, you may want to just flatten the response output. The Ninja Schema object extends Pydantic's Field(..., alias="") format to work with dotted responses.

Using the models from above, let's make a schema that just includes the task owner's first name inline, and also uses completed rather than is_completed:

from ninja import Field, Schema

class TaskSchema(Schema):
    id: int
    title: str
    # The first Field param is the default, use ... for required fields.
    completed: bool = Field(..., alias="is_completed")
    owner_first_name: str = Field(None, alias="owner.first_name")

Aliases also support django template syntax variables access:

class TaskSchema(Schema):
    last_message: str = Field(None, alias="message_set.0.text")
class TaskSchema(Schema):
    type: str = Field(None)
    type_display: str = Field(None, alias="get_type_display") # callable will be executed


You can also create calculated fields via resolve methods based on the field name.

The method must accept a single argument, which will be the object the schema is resolving against.

When creating a resolver as a standard method, self gives you access to other validated and formatted attributes in the schema.

class TaskSchema(Schema):
    id: int
    title: str
    is_completed: bool
    owner: Optional[str] = None
    lower_title: str

    def resolve_owner(obj):
        if not obj.owner:
        return f"{obj.owner.first_name} {obj.owner.last_name}"

    def resolve_lower_title(self, obj):
        return self.title.lower()

Accessing extra context

Pydantic v2 allows you to process an extra context that is passed to the serializer. In the following example you can have resolver that gets request object from passed context argument:

class Data(Schema):
    a: int
    path: str = ""

    def resolve_path(obj, context):
        request = context["request"]
        return request.path

if you use this schema for incoming requests - the request object will be automatically passed to context.

You can as well pass your own context:

data = Data.model_validate({'some': 1}, context={'request': MyRequest()})

Returning querysets

In the previous example we specifically converted a queryset into a list (and executed the SQL query during evaluation).

You can avoid that and return a queryset as a result, and it will be automatically evaluated to List:

@api.get("/tasks", response=List[TaskSchema])
def tasks(request):
    return Task.objects.all()


If your operation is async, this example will not work because the ORM query needs to be called safely.

@api.get("/tasks", response=List[TaskSchema])
async def tasks(request):
    return Task.objects.all()

See the async support guide for more information.

FileField and ImageField

Django Ninja by default converts files and images (declared with FileField or ImageField) to string URL's.

An example:

class Picture(models.Model):
    title = models.CharField(max_length=100)
    image = models.ImageField(upload_to='images')

If you need to output to response image field, declare a schema for it as follows:

class PictureSchema(Schema):
    title: str
    image: str

Once you output this to a response, the URL will be automatically generated for each object:

    "title": "Zebra",
    "image": "/static/images/zebra.jpg"

Multiple Response Schemas

Sometimes you need to define more than response schemas. In case of authentication, for example, you can return:

  • 200 successful -> token
  • 401 -> Unauthorized
  • 402 -> Payment required
  • etc..

In fact, the OpenAPI specification allows you to pass multiple response schemas.

You can pass to a response argument a dictionary where:

  • key is a response code
  • value is a schema for that code

Also, when you return the result - you have to also pass a status code to tell Django Ninja which schema should be used for validation and serialization.

An example:

class Token(Schema):
    token: str
    expires: date

class Message(Schema):
    message: str'/login', response={200: Token, 401: Message, 402: Message})
def login(request, payload: Auth):
    if auth_not_valid:
        return 401, {'message': 'Unauthorized'}
    if negative_balance:
        return 402, {'message': 'Insufficient balance amount. Please proceed to a payment page.'}
    return 200, {'token': xxx, ...}

Multiple response codes

In the previous example you saw that we basically repeated the Message schema twice:

...401: Message, 402: Message}

To avoid this duplication you can use multiple response codes for a schema:

from ninja.responses import codes_4xx'/login', response={200: Token, codes_4xx: Message})
def login(request, payload: Auth):
    if auth_not_valid:
        return 401, {'message': 'Unauthorized'}
    if negative_balance:
        return 402, {'message': 'Insufficient balance amount. Please proceed to a payment page.'}
    return 200, {'token': xxx, ...}

Django Ninja comes with the following HTTP codes:

from ninja.responses import codes_1xx
from ninja.responses import codes_2xx
from ninja.responses import codes_3xx
from ninja.responses import codes_4xx
from ninja.responses import codes_5xx

You can also create your own range using a frozenset:

my_codes = frozenset({416, 418, 425, 429, 451})
...'/login', response={200: Token, my_codes: Message})

Empty responses

Some responses, such as 204 No Content, have no body. To indicate the response body is empty mark response argument with None instead of Schema:"/no_content", response={204: None})
def no_content(request):
    return 204, None

Self-referencing schemes

Sometimes you need to create a schema that has reference to itself, or tree-structure objects.

To do that you need:

  • set a type of your schema in quotes
  • use update_forward_refs method to apply self referencing types
class Organization(Schema):
    title: str
    part_of: 'Organization' = None     #!! note the type in quotes here !!

Organization.update_forward_refs()  # !!! this is important

@api.get('/organizations', response=List[Organization])
def list_organizations(request):

Self-referencing schemes from create_schema()

To be able to use the method update_forward_refs() from a schema generated via create_schema(), the "name" of the class needs to be in our namespace. In this case it is very important to pass the name parameter to create_schema()

UserSchema = create_schema(
    name='UserSchema',  # !!! this is important for update_forward_refs()  
    fields=['id', 'username']
        ('manager', 'UserSchema', None),

Serializing Outside of Views

Serialization of your objects can be done directly in code through the use of the .from_orm() method on the schema object.

Consider the following model:

class Person(models.Model):
    name = models.CharField(max_length=50)

Which can be accessed using this schema:

class PersonSchema(Schema):
    name: str

Direct serialization can be performed using the .from_orm() method on the schema. Once you have an instance of the schema object, the .dict() and .json() methods allow you to get at both dictionary output and string JSON versions.

>>> person = Person.objects.get(id=1)
>>> data = PersonSchema.from_orm(person)
>>> data
PersonSchema(id=1, name='Mr. Smith')
>>> data.dict()
{'id':1, 'name':'Mr. Smith'}
>>> data.json()
'{"id":1, "name":"Mr. Smith"}'

Multiple Items: or a queryset (or list)

>>> persons = Person.objects.all()
>>> data = [PersonSchema.from_orm(i).dict() for i in persons]
[{'id':1, 'name':'Mr. Smith'},{'id': 2, 'name': 'Mrs. Smith'}...]

Django HTTP responses

It is also possible to return regular django http responses:

from django.http import HttpResponse
from django.shortcuts import redirect

def result_django(request):
    return HttpResponse('some data')   # !!!!

def some_redirect(request):
    return redirect("/some-path")  # !!!!