Punishment reasons, appeals and in-group commands

Automated moderation makes mistakes — that is normal. What is not normal is a person not understanding what they were punished for and having nowhere to take it. This release is about feedback in both directions.

The bot explains the reason#

A punishment now comes with a reason: which rule fired, or — if the decision came from the model — its own wording of why the message looked like a violation. The reason is translated into the chat's language, so nobody has to parse technical English.

What to show members and what to keep in the log is configurable: in some chats a public reason is welcome, in others it turns into a manual for spammers.

Appeals#

The dashboard now has an Appeals section. A member who disagrees with a punishment contests it right in their conversation with the bot; the request lands in the section, where it is shown together with the original message, the rule that fired and the member's history. The decision — lift the punishment or keep it — goes back to the member. The right to handle appeals is granted separately from the rest, so sorting out complaints can be one person's job.

Moderator review in direct messages#

Not everything needs to be dragged into the dashboard. Borderline matches are now sent to a moderator in direct messages: a card with the text, the reason and two buttons — confirm or undo. Undoing restores the message and lifts the punishment. For a chat where the moderator lives in Telegram anyway, this is the fastest route.

Commands right in the group#

A full set of commands, so you do not have to open the dashboard for a single ban. As a reply to the offender's message:

Command	What it does
/d	delete the message
/m	mute
/b	ban
/k	kick
/u	lift a ban or a mute
/s	pick a punishment from a list
/i	the member's card
/log	their recent events
/msg	a message from the bot

Members get /me for their own punishment history, /why for what their last message was deleted for, and /help.

Command replies are ephemeral: the bot answers and cleans up after itself, so housekeeping does not clutter the chat. Access to commands is checked against dashboard permissions, and the command set is switched on and off per chat — either as a whole or one command at a time.