https://gitlab.synchro.net/main/sbbs/-/commit/8fe77e0ee2e083c5f8736be9
Added Files:
exec/spamlearn.js
Modified Files:
docs/v322_new.md exec/load/salib.js
Log Message:
Add spamlearn.js mail processor to train SpamAssassin Bayes from e-mail

New exec/spamlearn.js consumes a (redirected) message and feeds it to a listening spamd via the TELL command to train the Bayes classifier as
spam or ham. Intended as a mailproc.ini script: redirect spam to a
"spamlearn" address and false-positives to a "hamlearn" address, gated
to the authenticated sysop (AccessRequirements = user equal 1) with passthru=false so the message is learned, not delivered, and outsiders
can't poison Bayes. Requires spamd started with --allow-tell.

salib.js gains a learn(msgclass) method (Message_DoLearn) implementing
the TELL protocol, mirroring the existing DoCommand convention. Unlike
the scoring path it injects no synthetic Received/Return-Path (the
original sender's headers train Bayes) and sends no User header, so it
trains spamd's own run-as Bayes database - the same one used to score
inbound mail.

Validated end-to-end against the live spamd: a fresh message returns learned:true (spamd "Setting local"); a repeat returns learned:false
(already known).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---
■ Synchronet ■ Vertrauen ■ Home of Synchronet ■ [vert/cvs/bbs].synchro.net