Skip to main content
Extensions are extra packages that add functionality to a Flask application. While PyPI contains many Flask extensions, you may not find one that fits your need. If this is the case, you can create your own, and publish it for others to use as well. This guide will show how to create a Flask extension, and some of the common patterns and requirements involved. Since extensions can do anything, this guide won’t be able to cover every possibility.

​
Learning Resources

The best ways to learn about extensions are to:
  • Look at how other extensions you use are written
  • Discuss with others on Discord Chat or GitHub Discussions
  • Share your design ideas early for feedback
The best extensions share common patterns, so that anyone familiar with using one extension won’t feel completely lost with another. This can only work if collaboration happens early.

​
Naming

A Flask extension typically has flask in its name as a prefix or suffix. If it wraps another library, it should include the library name as well. This makes it easy to search for extensions, and makes their purpose clearer.

​
Naming Conventions

A general Python packaging recommendation is that the install name from the package index and the name used in import statements should be related. The import name is lowercase, with words separated by underscores (_). The install name is either lower case or title case, with words separated by dashes (-). If it wraps another library, prefer using the same case as that library’s name. Here are some example install and import names:
  • Flask-Name imported as flask_name
  • flask-name-lower imported as flask_name_lower
  • Flask-ComboName imported as flask_comboname
  • Name-Flask imported as name_flask

​
The Extension Class and Initialization

All extensions will need some entry point that initializes the extension with the application. The most common pattern is to create a class that represents the extension’s configuration and behavior, with an init_app method to apply the extension instance to the given application instance. 
class HelloExtension:
    def __init__(self, app=None):
        if app is not None:
            self.init_app(app)

    def init_app(self, app):
        app.before_request(...)

​
Important: Don’t Store the App

It is important that the app is not stored on the extension, don’t do self.app = app. The only time the extension should have direct access to an app is during init_app, otherwise it should use current_app. This allows the extension to support the application factory pattern, avoids circular import issues when importing the extension instance elsewhere in a user’s code, and makes testing with different configurations easier. 
hello = HelloExtension()

def create_app():
    app = Flask(__name__)
    hello.init_app(app)
    return app
Above, the hello extension instance exists independently of the application. This means that other modules in a user’s project can do from project import hello and use the extension in blueprints before the app exists.

​
Storing Extension State

The Flask.extensions dict can be used to store a reference to the extension on the application, or some other state specific to the application. Be aware that this is a single namespace, so use a name unique to your extension, such as the extension’s name without the “flask” prefix.

​
Adding Behavior

There are many ways that an extension can add behavior. Any setup methods that are available on the Flask object can be used during an extension’s init_app method.

​
Request Lifecycle Hooks

A common pattern is to use before_request to initialize some data or a connection at the beginning of each request, then teardown_request to clean it up at the end. This can be stored on g. 
def init_app(self, app):
    app.before_request(self._before_request)
    app.teardown_request(self._teardown_request)

def _before_request(self):
    g._myext_connection = create_connection()

def _teardown_request(self, exception):
    connection = g.pop('_myext_connection', None)
    if connection is not None:
        connection.close()

​
Lazy Initialization

A more lazy approach is to provide a method that initializes and caches the data or connection. For example, a ext.get_db method could create a database connection the first time it’s called, so that a view that doesn’t use the database doesn’t create a connection.

​
Adding Views with Blueprints

Besides doing something before and after every view, your extension might want to add some specific views as well. In this case, you could define a Blueprint, then call register_blueprint during init_app to add the blueprint to the app. 
def init_app(self, app):
    from .views import bp
    app.register_blueprint(bp)

​
Configuration Techniques

There can be multiple levels and sources of configuration for an extension. You should consider what parts of your extension fall into each one.

​
Configuration Levels

  1. Per Application Instance - Through app.config values. This is configuration that could reasonably change for each deployment of an application. A common example is a URL to an external resource, such as a database. Configuration keys should start with the extension’s name so that they don’t interfere with other extensions.
  2. Per Extension Instance - Through __init__ arguments. This configuration usually affects how the extension is used, such that it wouldn’t make sense to change it per deployment.
  3. Instance Attributes and Decorators - It might be more ergonomic to assign to ext.value, or use a @ext.register decorator to register a function, after the extension instance has been created.
  4. Global Class Attributes - Changing a class attribute like Ext.connection_class can customize default behavior without making a subclass. This could be combined with per-extension configuration to override defaults.
  5. Subclassing - Making the API of the extension itself something that can be overridden provides a very powerful tool for advanced customization.
The Flask object itself uses all of these techniques.

​
Configuration Best Practices

Configuration should not be changed after the application setup phase is complete and the server begins handling requests. Configuration is global, any changes to it are not guaranteed to be visible to other workers.

​
Data During a Request

When writing a Flask application, the g object is used to store information during a request. Extensions can also use this, with some care. Since g is a single global namespace, extensions must use unique names that won’t collide with user data. For example, use the extension name as a prefix, or as a namespace. 
# an internal prefix with the extension name
g._hello_user_id = 2

# or an internal prefix as a namespace
from types import SimpleNamespace
g._hello = SimpleNamespace()
g._hello.user_id = 2

​
Application Context Lifecycle

The data in g lasts for an application context. An application context is active during a request, CLI command, or with app.app_context() block.
  • If you’re storing something that should be closed, use teardown_appcontext to ensure that it gets closed when the app context ends
  • If it should only be valid during a request, or would not be used in the CLI outside a request, use teardown_request

​
Views and Models

Your extension views might want to interact with specific models in your database, or some other extension or data connected to your application.

​
Example: Blog Extension with Models

For example, let’s consider a Flask-SimpleBlog extension that works with Flask-SQLAlchemy to provide a Post model and views to write and read posts. The Post model needs to subclass the Flask-SQLAlchemy db.Model object, but that’s only available once you’ve created an instance of that extension, not when your extension is defining its views.

​
Using MethodView

One method could be to use class-based views. During __init__, create the model, then create the views by passing the model to the view class’s as_view method. 
class PostAPI(MethodView):
    def __init__(self, model):
        self.model = model

    def get(self, id):
        post = self.model.query.get(id)
        return jsonify(post.to_json())

class BlogExtension:
    def __init__(self, db):
        class Post(db.Model):
            id = db.Column(primary_key=True)
            title = db.Column(db.String, nullable=False)

        self.post_model = Post

    def init_app(self, app):
        api_view = PostAPI.as_view(model=self.post_model)

db = SQLAlchemy()
blog = BlogExtension(db)
db.init_app(app)
blog.init_app(app)

​
Using app.extensions

Another technique could be to use an attribute on the extension, such as self.post_model from above. Add the extension to app.extensions in init_app, then access current_app.extensions["simple_blog"].post_model from views.

​
Providing Base Classes

You may also want to provide base classes so that users can provide their own Post model that conforms to the API your extension expects. So they could implement class Post(blog.BasePost), then set it as blog.post_model. As you can see, this can get a bit complex. Unfortunately, there’s no perfect solution here, only different strategies and tradeoffs depending on your needs and how much customization you want to offer. Flask previously had the concept of “approved extensions”, where the Flask maintainers evaluated the quality, support, and compatibility of the extensions before listing them. While the list became too difficult to maintain over time, the guidelines are still relevant to all extensions maintained and developed today, as they help the Flask ecosystem remain consistent and compatible.

​
1. Maintenance

An extension requires a maintainer. In the event an extension author would like to move beyond the project, the project should find a new maintainer and transfer access to the repository, documentation, PyPI, and any other services. The Pallets-Eco organization on GitHub allows for community maintenance with oversight from the Pallets maintainers.

​
2. Naming Scheme

The naming scheme is Flask-ExtensionName or ExtensionName-Flask. It must provide exactly one package or module named flask_extension_name.

​
3. Open Source License

The extension must use an open source license. The Python web ecosystem tends to prefer BSD or MIT. It must be open source and publicly available.

​
4. API Requirements

The extension’s API must have the following characteristics:
  • It must support multiple applications running in the same Python process. Use current_app instead of self.app, store configuration and state per application instance.
  • It must be possible to use the factory pattern for creating applications. Use the ext.init_app() pattern.

​
5. Installation

From a clone of the repository, an extension with its dependencies must be installable in editable mode with pip install -e ..

​
6. Testing

It must ship tests that can be invoked with a common tool like tox -e py, nox -s test or pytest. If not using tox, the test dependencies should be specified in a requirements file. The tests must be part of the sdist distribution.

​
7. Documentation

A link to the documentation or project website must be in the PyPI metadata or the readme. The documentation should use the Flask theme from the Official Pallets Themes.

​
8. Dependency Management

The extension’s dependencies should not use upper bounds or assume any particular version scheme, but should use lower bounds to indicate minimum compatibility support. For example, sqlalchemy>=1.4.

​
9. Python Version Support

Indicate the versions of Python supported using python_requires=">=version". Flask and Pallets policy is to support all Python versions that are not within six months of end of life (EOL). See Python’s EOL calendar for timing.

Build docs developers (and LLMs) love

Get started for freeTalk to us