[Rspamd-Users] Disable autolearn per user for spam-reporting mailbox

[Gerald Galster]

Hi Stefan,

> thank you for your time. Are you sure that
> "With prefilter=true; and action = "accept"; no filters will be processed."
> Will not still LEARN ham/spam to bayes?
> 
> I dont think, this is a filter as you described.


I'm not 100% sure but documentation suggests pre-filter is executed before classifiers:
https://rspamd.com/doc/faq.html#what-are-filters-pre-filters-and-post-filters

If all those checks are skipped I can't imagine why a mail would be auto_learned,
a lot of spam/ham hints are missing then.

To be sure, you could enable debugging of bayes module in logging and see what happens
http://rspamd.com/doc/configuration/logging.html (Module: bayes)


IIRC a message is auto_learned when the score has reached reject level or is negative.
You could try to override e.g. reject level in settings.conf for your rcpt like you
did with groups_enabled in your last mail.

http://rspamd.com/doc/configuration/settings.html
(See example at Settings structure)


Best regards
Gerald