7 Mailvisa is a spam filter along the lines of Paul Graham's "A Plan for
8 Spam". It classifies messages by comparing the words in them to known
9 spam words. Memory usage, filtering speed, and filtering accuracy are
10 comparable to other bayesian filters. Mailvisa is written in Ruby, and
11 runs on any Unix platform that Ruby runs on.
16 See the file INSTALL for building and installation instructions. In
17 short, it boils down to
26 A little setup is required before using mailvisa. It looks for configuration
27 files in $HOME/settings/mailvisa, so you need to create that directory.
29 Then, you can train mailvisa by teaching it what good and bad messages look
30 like (this assumes you have your messages stored in directories, one message
31 per file. If you store your messages in an mbox file, use a tool like mailconv
32 to convert it to MH or maildir):
34 mailvisa add good ~/mail/good_messages/*
35 mailvisa add bad ~/mail/bad_messages/*
37 Then, calculate spam scores for each word:
41 Mailvisa uses a daemon, so that the score file doesn't have to be
42 re-read on every invocation. This means you must start the daemon before
43 you can start filtering spam:
47 Then, you can process a message by issuing:
49 mailvisa check < message
51 This will output the message, with an X-Spam: header prepended that indicates
52 whether mailvisa thinks your message is spam or not. It will also
53 indicate its verdict in the exit code: 0 means non-spam, 160 means spam.
54 Anything else means an error occured.
56 All mailvisa commands can be passed a -h option to get help on using
57 that command. See the manpage (mailvisa.1) for more details.