Django-SpamBayes includes a few utility apps for Django that offer an easy to use interface to the SpamBayes statistical anti-spam filter allowing filtering and moderation (with training) of comments, contact forms and other publicly submitted data.

Note: These apps do not depend on any of the e-mail client plugins (i.e.: Outlook plugin) available from the SpamBayes project. However, the core spambayes python package needs to be installed on your PYTHONPATH. It is recommended that you download and install a recent "source release" of SpamBayes appropriate for your platform.

Apps:djangobayesThe "djangobayes" app is the workhorse that gets Django and SpamBayes to talk to each other. It contains a Django model for SpamBayes to store its 'learned' data in and a SpamBayes Classifier subclass that wraps the Django model. This is all wrapped up in a SpamBayes Hammie instance (djangobayes.filter.filter) which contains the public API accessed by other apps.

Basic Usage:

>>> from djangobayes.filter import filter
>>> score = filter.score('some text')
>>> if score < 0.3:
... # do something with spam
... elif score > 0.8:
... # do something with ham
... else:
... # unsure - do something elseOf course, this will only work if SpamBayes has been trained. To train do:

>>> filter.train('some spam', True)
>>> filter.store()
>>> filter.train('some ham', False)
>>> filter.store()In some situations you may want to filter based on more than a single body of text. For example, an author's name, e-mail address, IP address, etc. As SpamBayes is intended for filtering e-mail, you can pass in a message object rather than a plain string. djangobayes provides a handy shortcut, build_msg which excepts a body and any keywords you provide to build a message. For example:

>>> from djangobayes.filter import build_msg
>>> msg = build_msg('some text', author='Bob', email='bob@example.com')
>>> score = filter.score(msg)
>>> # or ...
>>> msg = build_msg('body text', **{'name' : 'Sue', 'ip' : '127.0.0.1'})
>>> filter.train(msg)
>>> filter.store()As far as I can tell, SpamBayes doesn't really care what the names of the keywords are as long as they are consistent throughout your app. It simply sees them as e-mail headers and doesn't give any particular headers more weight than others. That's why the primary text must be passed in separately.

As this is a SpamBayes API, see the SpamBayes documentation (and read the code) for more information.

commentbayesAn add-on to the django.contrib.comments app that adds SpamBayes filtering. Requires "djangobayes" to be installed in the project. commentbayes simply uses signals to add scoring and training to django.contrib.comments moderation.

commentbayes supports two optional settings to indicate the ham and spam thresholds. Each should be a floating point number between 0 and 1:

HAM_LEVEL (default is 0.3) SPAM_LEVEL (default is 0.8)

Assuming you have already installed django.contrib.comments, add commentbayes and djangobayes to INSTALLED_APPS in your settings.py file, run manage.py syncdb (to add the djangobayes table) and your all set. All future comments will be run through the SpamBayes filter. Any comment with a score greater than SPAM_LEVEL will be marked for deletion (comment.is_removed = True), any comment with a score equal to or between HAM_LEVEL and SPAM_LEVEL will be marked for moderation (comment.is_public = False), and any comment with a score less than HAM_LEVEL will be marked as ham.

To train SpamBayes, point your browser at http://yourdomain.com/comments/moderate/ (an undocumented part of django.contrib.comments) and either 'approve' or 'remove' the comments awaiting moderation. In a short time and with good training, you will be enjoying spam-free comments!

You may want to set up a cron job to remove comments marked for deletion (comment.is_removed = True) on occasion.

sb_demoThe "sb_demo" app is a very simple demonstration that provides a single page of comments with feedback on the SpamBayes scoring. This app does not use Django's forms or various other mechanisms that would make is useful for others. It's simply a demonstration that filtering works - mostly.

It also adds a very basic template for flatpages with comments enabled which provides a demo of commentbayes.

GAE BranchThe gae branch in the repo is an early semi-working version of the code for Google App Engine and BigTable. The file sb.py is essentially the same as the sbayes app above, but for BigTable. I'm not currently using or maintaining this branch, but it's available for anyone who wants it.