Flasgger is a Flask extension toextractOpenAPI-Specificationfrom all Flask views registered in your API.
Flasgger also comes withSwaggerUIembeddedso you can access http://localhost:5000/apidocs and visualize and interact with your API resources.
Flasgger alsoprovides validationof the incoming data, using the same specification it can validates if the data received as a POST, PUT, PATCH is valid against the schema defined usingYAML,Python dictionariesorMarshmallow Schemas.
Flasgger can work with simple function views or MethodViews using docstring as specification, or using@swag_from
decorator to get specification fromYAMLordictand also providesSwaggerViewwhich can useMarshmallow Schemasas specification.
Flasgger is compatible withFlask-RESTful
so you can useResources
andswag
specifications together, take a look atrestful example.
Flasgger also supportsMarshmallow APISpec
as base template for specification, if you are using APISPec from Marshmallow take a look atapispec example.
- Top Contributors
- Examples and demo app
- Installation
- Getting started
- Use the same data to validate your API POST body.
- Get defined schemas as Python dictionaries
- HTML sanitizer
- Swagger UI and templates
- OpenAPI 3.0 Support
- Initializing Flasgger with default data.
- Customize default configurations
Created bygh-md-toc
There are someexample applicationsand you can also play with examples inFlasgger demo app
NOTE: all the examples apps are also test cases and run automatically in Travis CI to ensure quality and coverage.
The examples and demo app can also be built and run as a Docker image/container:
docker build -t flasgger.
docker run -it --rm -p 5000:5000 --name flasgger flasgger
Then access the Flasgger demo app athttp://localhost:5000.
under your virtualenv do:
Ensure you have latest setuptools
pip install -U setuptools
then install beta version (recommended)
pip install flasgger==0.9.7b2
or (latest stable for legacy apps)
pip install flasgger==0.9.5
or (dev version)
pip install https://github /flasgger/flasgger/tarball/master
NOTE: If you want to useMarshmallow Schemasyou also need to run
pip install marshmallow apispec
(You may see the command in.travis.ymlforbefore_install
part)
In your virtualenv:
pip install -r requirements.txt
pip install -r requirements-dev.txt
make test
Create a file called for examplecolors.py
fromflaskimportFlask,jsonify
fromflasggerimportSwagger
app=Flask(__name__)
swagger=Swagger(app)
@app.route('/colors/<palette>/')
defcolors(palette):
"""Example endpoint returning a list of colors by palette
This is using docstrings for specifications.
---
parameters:
- name: palette
in: path
type: string
enum: ['all', 'rgb', 'cmyk']
required: true
default: all
definitions:
Palette:
type: object
properties:
palette_name:
type: array
items:
$ref: '#/definitions/Color'
Color:
type: string
responses:
200:
description: A list of colors (may be filtered by palette)
schema:
$ref: '#/definitions/Palette'
examples:
rgb: ['red', 'green', 'blue']
"""
all_colors={
'cmyk':['cyan','magenta','yellow','black'],
'rgb':['red','green','blue']
}
ifpalette=='all':
result=all_colors
else:
result={palette:all_colors.get(palette)}
returnjsonify(result)
app.run(debug=True)
Now run:
Python colors.py
And go to:http://localhost:5000/apidocs/
You should get:
Save a new filecolors.yml
Example endpoint returning a list of colors by palette
In this example the specification is taken from external YAML file
---
parameters:
-name:palette
in:path
type:string
enum:['all', 'rgb', 'cmyk']
required:true
default:all
definitions:
Palette:
type:object
properties:
palette_name:
type:array
items:
$ref:'#/definitions/Color'
Color:
type:string
responses:
200:
description:A list of colors (may be filtered by palette)
schema:
$ref:'#/definitions/Palette'
examples:
rgb:['red', 'green', 'blue']
lets use the same example changing only the view function.
fromflasggerimportswag_from
@app.route('/colors/<palette>/')
@swag_from('colors.yml')
defcolors(palette):
...
If you do not want to use the decorator you can use the docstringfile:
shortcut.
@app.route('/colors/<palette>/')
defcolors(palette):
"""
file: colors.yml
"""
...
Create a Python dictionary as:
specs_dict={
"parameters":[
{
"name":"palette",
"in":"path",
"type":"string",
"enum":[
"all",
"rgb",
"cmyk"
],
"required":"true",
"default":"all"
}
],
"definitions":{
"Palette":{
"type":"object",
"properties":{
"palette_name":{
"type":"array",
"items":{
"$ref":"#/definitions/Color"
}
}
}
},
"Color":{
"type":"string"
}
},
"responses":{
"200":{
"description":"A list of colors (may be filtered by palette)",
"schema":{
"$ref":"#/definitions/Palette"
},
"examples":{
"rgb":[
"red",
"green",
"blue"
]
}
}
}
}
Now take the same function and use the dict in the place of YAML file.
@app.route('/colors/<palette>/')
@swag_from(specs_dict)
defcolors(palette):
"""Example endpoint returning a list of colors by palette
In this example the specification is taken from specs_dict
"""
...
FIRST:
pip install marshmallow apispec
USAGE #1:
SwaggerView
fromflaskimportFlask,jsonify
fromflasggerimportSwagger,SwaggerView,Schema,fields
classColor(Schema):
name=fields.Str()
classPalette(Schema):
pallete_name=fields.Str()
colors=fields.Nested(Color,many=True)
classPaletteView(SwaggerView):
parameters=[
{
"name":"palette",
"in":"path",
"type":"string",
"enum":["all","rgb","cmyk"],
"required":True,
"default":"all"
}
]
responses={
200:{
"description":"A list of colors (may be filtered by palette)",
"schema":Palette
}
}
defget(self,palette):
"""
Colors API using schema
This example is using marshmallow schemas
"""
all_colors={
'cmyk':['cyan','magenta','yellow','black'],
'rgb':['red','green','blue']
}
ifpalette=='all':
result=all_colors
else:
result={palette:all_colors.get(palette)}
returnjsonify(result)
app=Flask(__name__)
swagger=Swagger(app)
app.add_url_rule(
'/colors/<palette>',
view_func=PaletteView.as_view('colors'),
methods=['GET']
)
app.run(debug=True)
USAGE #2:
Custom Schema from flasgger
Body
- support all fields in marshmallowQuery
- support simple fields in marshmallow (Int, String and etc)Path
- support only int and str
fromflaskimportFlask,abort
fromflasggerimportSwagger,Schema,fields
frommarshmallow.validateimportLength,OneOf
app=Flask(__name__)
Swagger(app)
swag={"swag":True,
"tags":["demo"],
"responses":{200:{"description":"Success request"},
400:{"description":"Validation error"}}}
classBody(Schema):
color=fields.List(fields.String(),required=True,validate=Length(max=5),example=["white","blue","red"])
defswag_validation_function(self,data,main_def):
self.load(data)
defswag_validation_error_handler(self,err,data,main_def):
abort(400,err)
classQuery(Schema):
color=fields.String(required=True,validate=OneOf(["white","blue","red"]))
defswag_validation_function(self,data,main_def):
self.load(data)
defswag_validation_error_handler(self,err,data,main_def):
abort(400,err)
swag_in="query"
@app.route("/color/<id>/<name>",methods=["POST"],**swag)
defindex(body:Body,query:Query,id:int,name:str):
return{"body":body,"query":query,"id":id,"name":name}
if__name__=="__main__":
app.run(debug=True)
NOTE: take a look at
examples/validation.py
for a more complete example.
NOTE: when catching arguments in path rule always use explicit types, bad:
/api/<username>
good:/api/<string:username>
Flasgger is compatible with Flask-RESTful you only need to installpip install flask-restful
and then:
fromflaskimportFlask
fromflasggerimportSwagger
fromflask_restfulimportApi,Resource
app=Flask(__name__)
api=Api(app)
swagger=Swagger(app)
classUsername(Resource):
defget(self,username):
"""
This examples uses FlaskRESTful Resource
It works also with swag_from, schemas and spec_dict
---
parameters:
- in: path
name: username
type: string
required: true
responses:
200:
description: A single user item
schema:
id: User
properties:
username:
type: string
description: The name of the user
default: Steven Wilson
"""
return{'username':username},200
api.add_resource(Username,'/username/<username>')
app.run(debug=True)
Flasgger can be configured to auto-parse external YAML API docs.Set adoc_dir
in yourapp.config['SWAGGER']
and Swagger will load API docs by looking indoc_dir
for YAML files stored by endpoint-name and method-name. For example,'doc_dir': './examples/docs/'
and a file./examples/docs/items/get.yml
will provide a Swagger doc forItemsView
methodget
.
Additionally, when usingFlask RESTfulper above, by passingparse=True
when constructingSwagger
,Flasgger will useflask_restful.reqparse.RequestParser
,locate allMethodView
s and parsed and validated data will be stored inflask.request.parsed_data
.
You can separate specifications by endpoint or methods
fromflasgger.utilsimportswag_from
@app.route('/api/<string:username>',endpoint='with_user_name',methods=['PUT','GET'])
@app.route('/api/',endpoint='without_user_name')
@swag_from('path/to/external_file.yml',endpoint='with_user_name')
@swag_from('path/to/external_file_no_user_get.yml',endpoint='without_user_name',methods=['GET'])
@swag_from('path/to/external_file_no_user_put.yml',endpoint='without_user_name',methods=['PUT'])
deffromfile_decorated(username=None):
ifnotusername:
return"No user!"
returnjsonify({'username':username})
And the same can be achieved with multiple methods in aMethodView
orSwaggerView
by
registering theurl_rule
many times. Take a look atexamples/example_app
Settingswag_from
'svalidationparameter toTrue
will validate incoming data automatically:
fromflasggerimportswag_from
@swag_from('defs.yml',validation=True)
defpost():
# if not validate returns ValidationError response with status 400
# also returns the validation message.
Usingswagger.validate
annotation is also possible:
fromflasggerimportSwagger
swagger=Swagger(app)
@swagger.validate('UserSchema')
defpost():
'''
file: defs.yml
'''
# if not validate returns ValidationError response with status 400
# also returns the validation message.
Yet you can callvalidate
manually:
fromflasggerimportswag_from,validate
@swag_from('defs.yml')
defpost():
validate(request.json,'UserSchema','defs.yml')
# if not validate returns ValidationError response with status 400
# also returns the validation message.
It is also possible to definevalidation=True
inSwaggerView
and also use
specs_dict
for validation.
Take a look atexamples/validation.py
for more information.
All validation options can be found athttp://json-schema.org/latest/json-schema-validation.html
By default Flasgger will usePython -jsonschema to perform validation.
Custom validation functions are supported as long as they meet the requirements:
- take two, and only two, positional arguments:
- the data to be validated as the first; and
- the schema to validate against as the second argument
- raise any kind of exception when validation fails.
Any return value is discarded.
Providing the function to the Swagger instance will make it the default:
fromflasggerimportSwagger
swagger=Swagger(app,validation_function=my_validation_function)
Providing the function as parameter ofswag_from
orswagger.validate
annotations or directly to thevalidate
function will force it's use
over the default validation function for Swagger:
fromflasggerimportswag_from
@swag_from('spec.yml',validation=True,validation_function=my_function)
...
fromflasggerimportSwagger
swagger=Swagger(app)
@swagger.validate('Pet',validation_function=my_function)
...
fromflasggerimportvalidate
...
validate(
request.json,'Pet','defs.yml',validation_function=my_function)
By default Flasgger will handle validation errors by aborting the request with a 400 BAD REQUEST response with the error message.
A custom validation error handling function can be provided to supersede default behavior as long as it meets the requirements:
- take three, and only three, positional arguments:
- the error raised as the first;
- the data which failed validation as the second; and
- the schema used in to validate as the third argument
Providing the function to the Swagger instance will make it the default:
fromflasggerimportSwagger
swagger=Swagger(app,validation_error_handler=my_handler)
Providing the function as parameter ofswag_from
orswagger.validate
annotations or directly to thevalidate
function will force it's use
over the default validation function for Swagger:
fromflasggerimportswag_from
@swag_from(
'spec.yml',validation=True,validation_error_handler=my_handler)
...
fromflasggerimportSwagger
swagger=Swagger(app)
@swagger.validate('Pet',validation_error_handler=my_handler)
...
fromflasggerimportvalidate
...
validate(
request.json,'Pet','defs.yml',
validation_error_handler=my_handler)
Examples of use of a custom validation error handler function can be found atexample validation_error_handler.py
You may wish to use schemas you defined in your Swagger specs as dictionaries
without replicating the specification. For that you can use theget_schema
method:
fromflaskimportFlask,jsonify
fromflasggerimportSwagger,swag_from
app=Flask(__name__)
swagger=Swagger(app)
@swagger.validate('Product')
defpost():
"""
post endpoint
---
tags:
- products
parameters:
- name: body
in: body
required: true
schema:
id: Product
required:
- name
properties:
name:
type: string
description: The product's name.
default: "Guarana"
responses:
200:
description: The product inserted in the database
schema:
$ref: '#/definitions/Product'
"""
rv=db.insert(request.json)
returnjsonify(rv)
...
product_schema=swagger.get_schema('product')
This method returns a dictionary which contains the Flasgger schema id, all defined parameters and a list of required parameters.
By default Flasgger will try to sanitize the content in YAML definitions
replacing every\n
with<br>
but you can change this behaviour
setting another kind of sanitizer.
fromflasggerimportSwagger,NO_SANITIZER
app=Flask()
swagger=Swagger(app,sanitizer=NO_SANITIZER)
You can write your own sanitizer
swagger=Swagger(app,sanitizer=lambdatext:do_anything_with(text))
There is also a Markdown parser available, if you want to be able to render Markdown in your specs description useMK_SANITIZER
You can override thetemplates/flasgger/index.html
in your application and
this template will be theindex.html
for SwaggerUI. Useflasgger/ui2/templates/index.html
as base for your customization.
Flasgger supports Swagger UI versions 2 and 3, The version 3 is still experimental but you
can try settingapp.config['SWAGGER']['uiversion']
.
app=Flask(__name__)
app.config['SWAGGER']={
'title':'My API',
'uiversion':3
}
swagger=Swagger(app)
There is experimental support for OpenAPI 3.0 that should work when using SwaggerUI 3. To use OpenAPI 3.0, setapp.config['SWAGGER']['openapi']
to a version that the current SwaggerUI 3 supports such as'3.0.2'
.
For an example of this that usescallbacks
andrequestBody
,see thecallbacks example.
Starting with Flasgger 0.9.2 you can specify external URL locations for loading the JavaScript and CSS for the Swagger and jQuery libraries loaded in the Flasgger default templates. If the configuration properties below are omitted, Flasgger will serve static versions it includes - these versions may be older than the current Swagger UI v2 or v3 releases.
The following example loads Swagger UI and jQuery versions from unpkg:
swagger_config = Swagger.DEFAULT_CONFIG
swagger_config['swagger_ui_bundle_js'] = '//unpkg /swagger-ui-dist@3/swagger-ui-bundle.js'
swagger_config['swagger_ui_standalone_preset_js'] = '//unpkg /swagger-ui-dist@3/swagger-ui-standalone-preset.js'
swagger_config['jquery_js'] = '//unpkg /[email protected]/dist/jquery.min.js'
swagger_config['swagger_ui_css'] = '//unpkg /swagger-ui-dist@3/swagger-ui.css'
Swagger(app, config=swagger_config)
You can start your Swagger spec with any default data providing a template:
template={
"swagger":"2.0",
"info":{
"title":"My API",
"description":"API for my data",
"contact":{
"responsibleOrganization":"ME",
"responsibleDeveloper":"Me",
"email":"me@me",
"url":"me",
},
"termsOfService":"http://me /terms",
"version":"0.0.1"
},
"host":"mysite",# overrides localhost:500
"basePath":"/api",# base bash for blueprint registration
"schemes":[
"http",
"https"
],
"operationId":"getmyData"
}
swagger=Swagger(app,template=template)
And then the template is the default data unless some view changes it. You can also provide all your specs as template and have no views. Or views in external APP.
Sometimes you need to get some data at runtime depending on dynamic values ex: you want to checkrequest.is_secure
to decide ifschemes
will behttps
you can do that by usingLazyString
.
fromflaskimportFlask
fromflasggerimport,Swagger,LazyString,LazyJSONEncoder
app=Flask(__init__)
# Set the custom Encoder (Inherit it if you need to customize)
app.json_encoder=LazyJSONEncoder
template=dict(
info={
'title':LazyString(lambda:'Lazy Title'),
'version':LazyString(lambda:'99.9.9'),
'description':LazyString(lambda:'Hello Lazy World'),
'termsOfService':LazyString(lambda:'/there_is_no_tos')
},
host=LazyString(lambda:request.host),
schemes=[LazyString(lambda:'https'ifrequest.is_secureelse'http')],
foo=LazyString(lambda:"Bar")
)
Swagger(app,template=template)
TheLazyString
values will be evaluated only whenjsonify
encodes the value at runtime, so you have access to Flaskrequest, session, g, etc..
and also may want to access a database.
Sometimes you're serving your swagger docs behind an reverse proxy (e.g. NGINX). When following theFlask guidance, the swagger docs will load correctly, but the "Try it Out" button points to the wrong place. This can be fixed with the following code:
fromflaskimportFlask,request
fromflasggerimportSwagger,LazyString,LazyJSONEncoder
app=Flask(__name__)
app.json_encoder=LazyJSONEncoder
template=dict(swaggerUiPrefix=LazyString(lambda:request.environ.get('HTTP_X_SCRIPT_NAME','')))
swagger=Swagger(app,template=template)
Custom configurations such as a different specs route or disabling Swagger UI can be provided to Flasgger:
swagger_config={
"headers":[
],
"specs":[
{
"endpoint":'apispec_1',
"route":'/apispec_1.json',
"rule_filter":lambdarule:True,# all in
"model_filter":lambdatag:True,# all in
}
],
"static_url_path":"/flasgger_static",
# "static_folder": "static", # must be set by user
"swagger_ui":True,
"specs_route":"/apidocs/"
}
swagger=Swagger(app,config=swagger_config)
Definitions can be extracted whenid
is found in spec, example:
fromflaskimportFlask,jsonify
fromflasggerimportSwagger
app=Flask(__name__)
swagger=Swagger(app)
@app.route('/colors/<palette>/')
defcolors(palette):
"""Example endpoint returning a list of colors by palette
---
parameters:
- name: palette
in: path
type: string
enum: ['all', 'rgb', 'cmyk']
required: true
default: all
responses:
200:
description: A list of colors (may be filtered by palette)
schema:
id: Palette
type: object
properties:
palette_name:
type: array
items:
schema:
id: Color
type: string
examples:
rgb: ['red', 'green', 'blue']
"""
all_colors={
'cmyk':['cyan','magenta','yellow','black'],
'rgb':['red','green','blue']
}
ifpalette=='all':
result=all_colors
else:
result={palette:all_colors.get(palette)}
returnjsonify(result)
app.run(debug=True)
In this example you do not have to passdefinitions
but need to addid
to
your schemas.
Version0.9.5.*
will be the last version that supports Python2.
Please direct discussions to#399.