Add learning feature like in rspamd

[patefoniq]

Hi,

in short.

Is there any possibility to add one feature like in rspamd? I mean manual learning from raw e-mail:


It's the best way to check if the rule works properly.

 

[danielb]

You can hook this with some scripts and configuration. Here's what I'm doing, working with Zimbra as backend email server :

1. In Zimbra, there's always specials spam and ham users, where emails are fowarded when a user mark a mail as spam or as ham
2. In Zimbra, I create an IMAP account for PMG, which has access to both the spam and ham accounts (their INBOX appear as IMAP folders)
3. Now on PMG, add a script like this one : https://git.fws.fr/fws/ansible-roles/src/branch/master/roles/pmg/files/imap-sa-learn

The script will open the PMG imap account and feeds the baysian filtering with what's in there (either as spam, or as ham, based on the IMAP folder where the emails are). You just have to set the approriate env vars, and setup a cron job. Or a systemd service + timer (see https://git.fws.fr/fws/ansible-roles/src/branch/master/roles/pmg/templates/imap-sa-learn.service.j2 and https://git.fws.fr/fws/ansible-roles/src/branch/master/roles/pmg/templates/imap-sa-learn.timer.j2)

Hope it helps.
Regards, Daniel

 

[patefoniq]

Sounds fine, but I'll be very happy seeing this as a native option in pmg