Discord Repeat-Offender Daily Pattern Digest

On a daily schedule, this scans the Postgres offender ledger and surfaces the users who matter: those crossing a strike threshold, or whose offense rate is climbing day over day. An LLM turns the raw rows into a readable narrative per user, and the bot posts a ranked digest so the mod team starts the day knowing exactly who to watch.

Use it when individual offenses are already logged but nobody reviews the aggregate. Single warnings rarely reveal a pattern; this catches the slow-burn repeat offender before they require an emergency ban.

1. 1A daily schedule trigger fires at your chosen hour.
2. 2A Postgres query pulls every user with offenses in the trailing 7 days plus their prior-week counts.
3. 3A logic step keeps only users above the strike threshold or showing an upward trend.
4. 4An OpenAI call writes a short pattern summary for each flagged user, citing their worst categories.
5. 5The bot posts a ranked digest (highest risk first) to the moderator channel for the morning review.