stream-django is a Django client for Stream, it supports Django from 1.11 up to and including 4.0 using Python 2.7 and 3.4, 3.5, 3.6+.
You can sign up for a Stream account at https://getstream.io/get_started.
Note there is also a lower level Python - Stream integration library which is suitable for all Python applications.
💡 This is a library for the Feeds product. The Chat SDKs can be found here.
You can build:
- Activity streams such as seen on Github
- A twitter style newsfeed
- A feed like instagram/ pinterest
- Facebook style newsfeeds
- A notification system
You can check out our example apps built using this library (you can deploy them directly to Heroku with 1 click):
- Build activity streams & news feeds
- Example apps
- Table of Contents
- Installation
- Model integration
- Feed manager
- Showing the newsfeed
- Settings
- Temporarily disabling the signals
- Customizing enrichment
- Full documentation and Low level APIs access
- Copyright and License Information
Install stream_django package with pip:
pip install stream_django
add stream_django to your INSTALLED_APPS
INSTALLED_APPS = [
...
'stream_django'
]
STREAM_API_KEY = 'my_api_key'
STREAM_API_SECRET = 'my_api_secret_key'Login with Github on getstream.io and add
STREAM_API_KEY and STREAM_API_SECRET to your Django settings module (you can find them in the dashboard).
Stream Django can automatically publish new activities to your feed. Simple mixin the Activity class on the models you want to publish.
from stream_django.activity import Activity
class Tweet(models.Model, Activity):
...
class Like(models.Model, Activity):
...Every time a Tweet is created it will be added to the user's feed. Users which follow the given user will also automatically get the new tweet in their feeds.
Models are stored in feeds as activities. An activity is composed of at least the following fields: actor, verb, object, time. You can also add more custom data if needed. The Activity mixin will try to set things up automatically:
object is a reference to the model instance actor is a reference to the user attribute of the instance verb is a string representation of the class name
By default the actor field will look for an attribute called user or actor and a field called created_at to track creation time. If you're user field is called differently you'll need to tell us where to look for it. Below shows an example how to set things up if your user field is called author.
class Tweet(models.Model, Activity):
created_at = models.DateTimeField(auto_now_add=True)
author = models.ForeignKey(settings.AUTH_USER_MODEL, on_delete=models.CASCADE)
@property
def activity_actor_attr(self):
return self.authorOften you'll want to store more data than just the basic fields. You achieve this by implementing the extra_activity_data method in the model.
NOTE: you should only return data that json.dumps can handle (datetime instances are supported too).
class Tweet(models.Model, Activity):
@property
def extra_activity_data(self):
return {'is_retweet': self.is_retweet }Django Stream comes with a feed_manager class that helps with all common feed operations.
To get you started the manager has 4 feeds pre configured. You can add more feeds if your application needs it. The three feeds are divided in three categories.
The user feed stores all activities for a user. Think of it as your personal Facebook page. You can easily get this feed from the manager.
from stream_django.feed_manager import feed_manager
feed_manager.get_user_feed(user_id)The news feeds (or timelines) store the activities from the people you follow. There is both a simple timeline newsfeed (similar to twitter) and an aggregated version (like facebook).
timeline = feed_manager.get_news_feeds(user_id)['timeline']
timeline_aggregated = feed_manager.get_news_feeds(user_id)['timeline_aggregated']The notification feed can be used to build notification functionality.
Below we show an example of how you can read the notification feed.
notification_feed = feed_manager.get_notification_feed(user_id)By default the notification feed will be empty. You can specify which users to notify when your model gets created. In the case of a retweet you probably want to notify the user of the parent tweet.
class Tweet(models.Model, Activity):
@property
def activity_notify(self):
if self.is_retweet and self.parent_tweet.author != self.author:
target_feed = feed_manager.get_notification_feed(self.parent_tweet.author_id)
return [target_feed]Another example would be following a user. You would commonly want to notify the user which is being followed.
class Follow(models.Model, Activity):
@property
def activity_notify(self):
return [feed_manager.get_notification_feed(self.target_user.id)]To create the newsfeeds you need to notify the system about follow relationships. The manager comes with APIs to let a user's news feeds follow another user's feed. This code lets the current user's timeline and timeline_aggregated feeds follow the target_user's personal feed.
feed_manager.follow_user(request.user.id, target_user)When you read data from feeds, a like activity will look like this:
{'actor': 'core.User:1', 'verb': 'like', 'object': 'core.Like:42'}This is far from ready for usage in your template. We call the process of loading the references from the database enrichment. An example is shown below:
from stream_django.enrich import Enrich
enricher = Enrich()
feed = feed_manager.get_feed('timeline', request.user.id)
activities = feed.get(limit=25)['results']
enriched_activities = enricher.enrich_activities(activities)Now that you've enriched the activities you can render the template. For convenience we include the render activity template tag:
{% load activity_tags %}
{% for activity in activities %}
{% render_activity activity %}
{% endfor %}
The render_activity template tag will render the template activity/[aggregated]/[verb].html with the activity as context.
For example activity/tweet.html will be used to render an normal activity with verb tweet
{{ activity.actor.username }} said "{{ activity.object.body }} {{ activity.created_at|timesince }} ago"
and activity/aggregated/like.html for an aggregated activity with verb like
{{ activity.actor_count }} user{{ activity.actor_count|pluralize }} liked {% render_activity activity.activities.0 %}
If you need to support different kind of templates for the same activity, you can send a third parameter to change the template selection.
The example below will use the template activity/[aggregated]/homepage_%(verb)s.html
{% render_activity activity 'homepage' %}
STREAM_API_KEY
Your stream site api key. Default ''
STREAM_API_SECRET
Your stream site api key secret. Default ''
STREAM_LOCATION
The location API endpoint the client will connect to. Eg: STREAM_LOCATION='us-east'
STREAM_TIMEOUT
The connection timeout (in seconds) for the API client. Default 6.0
STREAM_FEED_MANAGER_CLASS
The path to the feed manager class. Default 'stream_django.managers.FeedManager'
STREAM_USER_FEED
The name of the feed (as it is configured in your GetStream.io Dasboard) where activities are stored. Default 'user'
STREAM_NEWS_FEEDS
The name of the news feed (as they are configured in your GetStream.io Dasboard) where activities from followed feeds are stored. Default {'timeline':'timeline', 'timeline_aggregated':'timeline_aggregated'}
STREAM_NOTIFICATION_FEED
The name of the feed (as it is configured in your GetStream.io Dasboard) where activity notifications are stored. Default 'notification'
STREAM_DISABLE_MODEL_TRACKING
Disable automatic tracking of Activity models. Default False
Model synchronization is disabled during schema/data migrations runs, syncdb and fixture loading (and during django test runs).
You can completely disable feed publishing via the STREAM_DISABLE_MODEL_TRACKING django setting.
Sometimes you'll want to customize how enrichment works. The documentation will show you several common options.
If you store references to model instances in the activity extra_data you can use the Enrich class to take care of it for you
from stream_django.activity import create_model_reference
class Tweet(models.Model, Activity):
@property
def extra_activity_data(self):
ref = create_model_reference(self.parent_tweet)
return {'parent_tweet': ref }
# instruct the enricher to enrich actor, object and parent_tweet fields
enricher = Enrich(fields=['actor', 'object', 'parent_tweet'])
feed = feed_manager.get_feed('timeline', request.user.id)
activities = feed.get(limit=25)['results']
enriched_activities = enricher.enrich_activities(activities)The enrich class that comes with the packages tries to minimise the amount of database queries. The models are grouped by their class and then retrieved with a pk__in query. You can implement a different approach to retrieve the instances of a model subclassing the stream_django.enrich.Enrich class.
To change the retrieval for every model you should override the fetch_model_instances method; in alternative you can change how certain models' are retrieved by implementing the hook function fetch_<model_name>_instances
class MyEnrich(Enrich):
'''
Overwrites how model instances are fetched from the database
'''
def fetch_model_instances(self, modelClass, pks):
'''
returns a dict {id:modelInstance} with instances of model modelClass
and pk in pks
'''
...
class AnotherEnrich(Enrich):
'''
Overwrites how Likes instances are fetched from the database
'''
def fetch_like_instances(self, pks):
return {l.id: l for l in Like.objects.cached_likes(ids)}You will commonly access related objects such as activity['object'].user. To prevent your newsfeed to run N queries you can instruct the manager to load related objects. The manager will use Django's select_related functionality. (https://docs.djangoproject.com/en/dev/ref/models/querysets/#select-related).
class Tweet(models.Model, Activity):
@classmethod
def activity_related_models(cls):
return ['user']When needed you can also use the low level Python API directly. Documentation is available at the Stream website.
from stream_django.client import stream_client
special_feed = stream_client.feed('special:42')
special_feed.follow('timeline:60')Copyright (c) 2014-2022 Stream.io Inc, and individual contributors. All rights reserved.
See the file "LICENSE" for information on the history of this software, terms & conditions for usage, and a DISCLAIMER OF ALL WARRANTIES.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent