- Documentation
- Moderation
Auto-moderation
The nine rules, and what Bao does when they fire.
Bao's auto-moderation is built on rules, not on an AI. It is predictable: the same situation always produces the same decision.
The possible actions
Each rule has an action:
| Action | Effect |
|---|---|
| Disabled | The rule does not apply |
| Delete | The message is deleted, with no infraction |
| Warn | An infraction is counted, the message stays |
| Delete and warn | Both |
Only an action that warns counts an infraction, and therefore feeds the escalating sanctions.
The rules
| Rule | Settings |
|---|---|
| Banned words | Your word list (1000 maximum) |
| Repeated text | Number of repetitions (2 to 50) over a window (1 to 300 s) |
| Discord invites | Invite links to other servers |
| External links | Any outbound URL |
| Excessive caps | Percentage of capitals (10 to 100%) beyond a minimum length |
| Excessive emoji | Maximum per message (1 to 100) |
| Excessive spoilers | Maximum per message |
| Excessive mentions | Maximum mentions per message |
| Spam | Number of messages (2 to 50) over a window (1 to 300 s) |
Global settings
- Ignore bots: do not moderate other bots.
- Notify by direct message: the sanctioned member receives an explanation.
- Log channel: where to record moderation actions.
- Exempt channels and exempt roles: never moderated.
- Infraction expiry: after how many days an infraction stops counting. 0 = never.
A word on tuning
Start with Delete alone on the noisy rules (caps, emoji), and reserve Delete and warn for the behaviour you genuinely want to sanction (spam, invites, banned words). Otherwise your members pile up infractions over trifles and your sanctions lose their meaning.