umputun / tg-spam / 36768223717

84%

Build Type

push

github

Committed by web-flow

Commit Message

feat(jev): add opt-in gibberish check (#468)

* feat(jev): add opt-in gibberish check

the first copy of "dsfg dfgh ewrt xczv dsfh" cleared production checks. Jev
scored it 0.12-0.15 under criteria that count only direct promotion as spam.

--jev.gibberish-threshold (0 = off) adds a second question to the same jev
call: is the message random letter groups that are not words in any language.
At or above the threshold jev reports spam; below it the spam question
decides as before, and with the check off the request is unchanged.

the question and criteria are fixed, since 0.50 was measured against that
wording. The gibberish question scored the post 0.87, and no other input
crossed 0.50 among 1,061 distinct 15+ character inputs rebuilt from six days
of one chat. The check requires jev history 0, because history is folded into
the judged message.

* fix(jev): skip gibberish question for quoted or replied-to text

bot.OnMessage appends quoted or replied-to text to the message, so the
gibberish question judged text the sender did not write: a caption-less reply
to a gibberish post was flagged as gibberish. The question is now skipped for
messages with quoted or replied-to text; the spam question still runs on them.

also names the question in every jev answer error, makes the makeDetector test
fail when the threshold does not reach the checker, and corrects the README
limits for quoted text, min-msg-len and veto mode with short messages.

Coverage Stats

70 of 70 new or added lines in 6 files covered. (100.0%)

2 existing lines in 1 file now uncovered.

9625 of 11305 relevant lines covered (85.14%)

231.56 hits per line