Now works on Debian and Windows.
Has a working moderator example, that redacts harmful (non PG14) messages from remaining in a matrix chat in about 0.5-2 seconds, and can ban them if they do it 3 times within an hour (though this at the crew level in the example).
The tool now has watch_room (wait for new messages to respond to in a room, and be alerted of message editing, etc.), ban_user, and redact_message functions, which the agent can use.
Full Changelog: v0.1.1...v0.1.2