Cookiecutter template for flask restful, including blueprints, application factory, and more
This cookie cutter is a very simple boilerplate for starting a REST api using Flask, flask-restful, marshmallow, SQLAlchemy and jwt. It comes with basic project structure and configuration, including blueprints, application factory and basics unit tests.
Features
- Simple flask application using application factory, blueprints
- Flask command line interface integration
- Simple cli implementation with basics commands (init, run, etc.)
- Flask Migrate included in entry point
- Authentication using Flask-JWT-Extended including access token and refresh token management
- Simple pagination utils
- Unit tests using pytest and factoryboy
- Configuration using environment variables
- OpenAPI json file and swagger UI
Used packages :
- Flask
- Flask-RESTful
- Flask-Migrate
- Flask-SQLAlchemy
- Flask-Marshmallow
- Flask-JWT-Extended
- marshmallow-sqlalchemy
- passlib
- tox
- pytest
- factoryboy
- dotenv
- apispec
- Installation
- Configuration
- Authentication
- Running tests
- WSGI Server
- Flask CLI
- Using Celery
- Using Docker
- Makefile
- APISpec, swagger, and redoc
- Changelog
Make sure you have cookiecutter installed in your local machine.
You can install it using this command : pip install cookiecutter
Starting a new project is as easy as running this command at the command line. No need to create a directory first, the cookiecutter will do it for you.
To create a project run the following command and follow the prompt
cookiecutter https://github.com/karec/cookiecutter-flask-restful
Let's say you named your app myapi
and your project myproject
You can install it using pip :
cd myproject
pip install -r requirements.txt
pip install -e .
You now have access to cli commands and can init your project
flask db init
flask db migrate
flask db upgrade
flask init # creates admin user
To list all commands
flask --help
Configuration is handled by environment variables, for development purpose you just
need to update / add entries in .flaskenv
file.
It's filled by default with following content:
FLASK_ENV=development
FLASK_APP="myapp.app:create_app"
SECRET_KEY=changeme
DATABASE_URI="sqlite:///myapp.db"
CELERY_BROKER_URL=amqp://guest:guest@localhost/ # only present when celery is enabled
CELERY_RESULT_BACKEND_URL=amqp://guest:guest@localhost/ # only present when celery is enabled
Avaible configuration keys:
FLASK_ENV
: flask configuration key, enablesDEBUG
if set todevelopment
SECREY_KEY
: your application secret keyDATABASE_URI
: SQLAlchemy connection stringCELERY_BROKER_URL
: URL to use for celery broker, only when you enabled celeryCELERY_RESULT_BACKEND_URL
: URL to use for celery result backend (e.g:redis://localhost
)
To access protected resources, you will need an access token. You can generate
an access and a refresh token using /auth/login
endpoint, example using curl
curl -X POST -H "Content-Type: application/json" -d '{"username": "admin", "password": "admin"}' http://localhost:5000/auth/login
This will return something like this
{
"access_token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ0eXBlIjoiYWNjZXNzIiwiaWRlbnRpdHkiOjEsImlhdCI6MTUxMDAwMDQ0MSwiZnJlc2giOmZhbHNlLCJqdGkiOiI2OTg0MjZiYi00ZjJjLTQ5MWItYjE5YS0zZTEzYjU3MzFhMTYiLCJuYmYiOjE1MTAwMDA0NDEsImV4cCI6MTUxMDAwMTM0MX0.P-USaEIs35CSVKyEow5UeXWzTQTrrPS_YjVsltqi7N4",
"refresh_token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpZGVudGl0eSI6MSwiaWF0IjoxNTEwMDAwNDQxLCJ0eXBlIjoicmVmcmVzaCIsImp0aSI6IjRmMjgxOTQxLTlmMWYtNGNiNi05YmI1LWI1ZjZhMjRjMmU0ZSIsIm5iZiI6MTUxMDAwMDQ0MSwiZXhwIjoxNTEyNTkyNDQxfQ.SJPsFPgWpZqZpHTc4L5lG_4aEKXVVpLLSW1LO7g4iU0"
}
You can use access_token to access protected endpoints :
curl -X GET -H "Content-Type: application/json" -H "Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ0eXBlIjoiYWNjZXNzIiwiaWRlbnRpdHkiOjEsImlhdCI6MTUxMDAwMDQ0MSwiZnJlc2giOmZhbHNlLCJqdGkiOiI2OTg0MjZiYi00ZjJjLTQ5MWItYjE5YS0zZTEzYjU3MzFhMTYiLCJuYmYiOjE1MTAwMDA0NDEsImV4cCI6MTUxMDAwMTM0MX0.P-USaEIs35CSVKyEow5UeXWzTQTrrPS_YjVsltqi7N4" http://127.0.0.1:5000/api/v1/users
You can use refresh token to retreive a new access_token using the endpoint /auth/refresh
curl -X POST -H "Content-Type: application/json" -H "Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpZGVudGl0eSI6MSwiaWF0IjoxNTEwMDAwNDQxLCJ0eXBlIjoicmVmcmVzaCIsImp0aSI6IjRmMjgxOTQxLTlmMWYtNGNiNi05YmI1LWI1ZjZhMjRjMmU0ZSIsIm5iZiI6MTUxMDAwMDQ0MSwiZXhwIjoxNTEyNTkyNDQxfQ.SJPsFPgWpZqZpHTc4L5lG_4aEKXVVpLLSW1LO7g4iU0" http://127.0.0.1:5000/auth/refresh
This will only return a new access token
{
"access_token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ0eXBlIjoiYWNjZXNzIiwiaWRlbnRpdHkiOjEsImlhdCI6MTUxMDAwMDYxOCwiZnJlc2giOmZhbHNlLCJqdGkiOiIzODcxMzg4Ni0zNGJjLTRhOWQtYmFlYS04MmZiNmQwZjEyNjAiLCJuYmYiOjE1MTAwMDA2MTgsImV4cCI6MTUxMDAwMTUxOH0.cHuNf-GxVFJnUZ_k9ycoMMb-zvZ10Y4qbrW8WkXdlpw"
}
Simplest way to run tests is to use tox, it will create a virtualenv for tests, install all dependencies and run pytest
tox
If you just want to run pytest and avoid linters you can use
tox -e test
If you want to run pytest manually without using tox you'll need to install some dependencies before
pip install pytest pytest-runner pytest-flask pytest-factoryboy pytest-celery factory_boy
Then you can invoke pytest
pytest
Note that tox is setting environment variables for you when testing, but when using pytest directly that's not the case. To avoid setting up env variables each time you run pytest, this cookiecutter provide a .testenv
file that contains default configuration for testing. Don't forget to update it if your local env doesn't match those defaults.
Testing with docker is another great option, since it take cares of everything and spawn required services for you. To run tests within docker containers, you can use the provided Makefile:
Build images:
make build
Running tox with flake8, black and pytest:
make tox
Running tox with pytest only:
make test
Testing celery require at least a rabbitMQ (or any other compatible broker) running. By default, when you use tox or the .testenv
file, celery broker and result backend are configured as follow:
CELERY_BROKER_URL=amqp://guest:guest@localhost/
CELERY_RESULT_BACKEND_URL=amqp://guest:guest@localhost/
Meaning that it will try to connect to a local rabbitMQ server using guest user. Don't forget to update those settings if your configuration doesn't match.
If you can't / don't want to install a local rabbitMQ server or any other available celery broker, you have 2 options:
- Use docker
You can use docker-compose to run tests, as it will spawn a rabbitMQ and a redis servera and set correct env variables for configuration. All tests commands are available inside the Makefile to simplify this process.
- Update the tests to use eager mode
NOTE this is not recommanded by celery: https://docs.celeryproject.org/en/stable/userguide/testing.html
Alternatively, if you don't have a local broker and can't use docker, you can update unit tests to run them using the task_always_eager
celery setting. This will actually run all tasks locally by blocking until tasks return (see https://docs.celeryproject.org/en/stable/userguide/configuration.html#std:setting-task_always_eager for more details).
Example of test_celery.py
file that use task_always_eager
import pytest
from myapi.app import init_celery
from myapi.tasks.example import dummy_task
@pytest.fixture(scope="session")
def celery_session_app(celery_session_app, app):
celery = init_celery(app)
celery_session_app.conf = celery.conf
celery_session_app.conf.task_always_eager = True
celery_session_app.Task = celery_session_app.Task
yield celery_session_app
def test_example(celery_session_app):
"""Simply test our dummy task using celery"""
res = dummy_task.delay()
assert res.get() == "OK"
This project provide a simple wsgi entry point to run gunicorn or uwsgi for example.
For gunicorn you only need to run the following commands
pip install gunicorn
gunicorn myapi.wsgi:app
And that's it ! Gunicorn is running on port 8000
If you chose gunicorn as your wsgi server, the proper commands should be in your docker-compose file.
Pretty much the same as gunicorn here
pip install uwsgi
uwsgi --http 127.0.0.1:5000 --module myapi.wsgi:app
And that's it ! Uwsgi is running on port 5000
If you chose uwsgi as your wsgi server, the proper commands should be in your docker-compose file.
This cookiecutter is fully compatible with default flask CLI and use a .flaskenv
file to set correct env variables to bind the application factory.
Note that we also set FLASK_ENV
to development
to enable debugger.
This cookiecutter has an optional Celery integration that let you choose if you want to use it or not in your project. If you choose to use Celery, additionnal code and files will be generated to get started with it.
This code will include a dummy task located in yourproject/yourapp/tasks/example.py
that only return "OK"
and a celery_app
file used to your celery workers.
In your project path, once dependencies are installed, you can just run
celery -A myapi.celery_app:app worker --loglevel=info
If you have updated your configuration for broker / result backend your workers should start and you should see the example task avaible
[tasks]
. myapi.tasks.example.dummy_task
To run a task you can either import it and call it
>>> from myapi.tasks.example import dummy_task
>>> result = dummy_task.delay()
>>> result.get()
'OK'
Or use the celery extension
>>> from myapi.extensions import celery
>>> celery.send_task('myapi.tasks.example.dummy_task').get()
'OK'
WARNING both Dockerfile and docker-compose.yml
are NOT suited for production, use them for development only or as a starting point.
This template offer simple docker support to help you get started and it comes with both Dockerfile and a docker-compose.yml
. Please note that docker-compose is mostly useful when using celery
since it takes care of running rabbitmq, redis, your web API and celery workers at the same time, but it also work if you don't use celery at all.
Dockerfile has intentionally no entrypoint to allow you to run any command from it (server, shell, init, celery, ...)
Note that you still need to init your app on first start, even when using compose.
docker build -t myapp .
...
docker run --env-file=.flaskenv myapp init
docker run --env-file=.flaskenv -p 5000:5000 myapp run -h 0.0.0.0
* Serving Flask app "myapi.app:create_app" (lazy loading)
* Environment: development
* Debug mode: on
* Running on http://0.0.0.0:5000/ (Press CTRL+C to quit)
* Restarting with stat
* Debugger is active!
* Debugger PIN: 214-619-010
With compose
docker-compose up
...
docker exec -it <container_id> flask init
With docker-compose and the Makefile
make init
Initizalize the environment
make init
Build the containers
make build
Run the containers
make run
Create migrations folder and database
make db-init
Create new database migration
make db-migrate
Apply database migrations
make db-upgrade
Run tests inside containers
make test
This boilerplate comes with pre-configured APISpec and swagger endpoints. Using default configuration you have four endpoints available:
/swagger.json
: return OpenAPI specification file in json format/swagger-ui
: Swagger UI configured to hit OpenAPI json file/openapi.yaml
: return OpenAPI specification file in yaml format/redoc-ui
: ReDoc UI configured to hit OpenAPI yaml file
This come with a very simple extension that allow you to override basic settings of APISpec using your config.py
file:
APISPEC_TITLE
: title for your spec, default to{{cookiecutter.project_name}}
APISPEC_VERSION
: version of your API, default to1.0.0
OPENAPI_VERSION
: OpenAPI version of your spec, default to3.0.2
SWAGGER_JSON_URL
: Url for your JSON specifications, default to/swagger.json
SWAGGER_UI_URL
: Url for swagger-ui, default to/swagger-ui
OPENAPI_YAML_URL
: Url for your YAML specifications, default to/openapi.yaml
REDOC_UI_URL
: Url for redoc-ui, default to/redoc-ui
SWAGGER_URL_PREFIX
: URL prefix to use for swagger blueprint, default toNone
- Updated readme makefile calls
- Fixed Makefile
- Removed entrypoint from setup to use flask default CLI
- Re-format apispec to fit black specs
- Fixed CLI to work with Flask 2.0's built-in CLI
- Added ReDoc UI and YAML OpenAPI Spec routes
- Updated Swagger UI version to fix previously-distorted version
- Updated README to reflect new CLI and ReDoc information
- Updated README for tests and celery
- Added a
.testenv
file to avoid needing to set env variables when running pytest manually - Updated celery fixtures to use session fixtures (for worker and app)
- Replaced
prefork
incelery_worker_pool
bysolo
(#41)
- Added python 3.8 support
- Upgraded to marshmallow 3
- Added
lint
andtests
envs to tox - Added black support
- Improved travis tests
- Updated Makefile to handle tests with celery
- Updated tox to handle env variables for celery when runing tests
- Added initial db migration instead of relying on
db.create_all()
- Added new step to create database in README
- Various cleanup
- Added apispec dependencies
- Registered
users
endpoints into swagger - New
apispec
extension - Added two new routes
/swagger.json
and/swagger-ui
(configurable urls) - Added swagger html template
- Add travis file
- Added docker and docker-compose support
- Update configuration to only use env variables,
.flaskenv
has been updated too - Add unit tests for celery
- Add flake8 to tox
- Configuration file cannot be overridden by
MYAPP CONFIG
env variable anymore - various cleanups (unused imports, removed
configtest.py
file, flake8 errors)