Is there any plans for dovecot and rspam integration? For example, when in the email box configuration in ispconfig, the move spam to junk box is ticked, does it automatically activate something where the user moves the spam from the junk box to the inbox, it will teach rspamd to learn that is ham. Another one, to teach rspamd to lower the score when the email is from a bank. Something similar like that.
I don't know of plans to set it up automatically, but you could log a feature request if you want. It's not hard to do in just rspamd with the docs available; I also use an external script to train from specific folders. I would recommend setting up a second redis instance for Bayes which limits memory use.
Thank you very much for the reply. Now I know the feature is not in. I will check the rspamd documents on this. Thank you for the external script. Will check it out. Will try and figure out how to log a feature request.
This guide might help you (it's in German, but using Google Translate might probably work): https://www.allerstorfer.at/rspamd-dovecot-ispconfig-automatic-spam-ham-training/
It's in english for me too. Great tutorial. Very helpful. I will try it out. Thank you very much for helping with the information.
Just a brief note, you might decide if you want to use that script to train from folders which I linked to above, and follow the instructions in that readme if so (ie. skip the other tutorial, as the steps are already covered in the readme). Using just training by moving to folders as that other tutorial sets up does work quite well, the extra functionality just allows an admin to moderate the training decisions that users make, eg. they will file something they don't want as spam, even though it may be from a legitimate source, and the moderator can override that so it doesn't affect the scanning of other users . (It's probably not very relevant for systems with per-user Bayes databases.)
