Lemmy Mod Bot

A Lemmy bot written in Python that allows you to automatically moderate communities from toxicity, duplicate posts, and CSAM content.

This bot is a work in progress and is expected to expand and grow.

Running the bot

A Docker image is provided for the purposes of easily running the moderation bot in a containerised environment. To run a bot with just the toxicity detection an example docker-compose file has been provided. To set up further modules (as detailed further below), mount a replacement main.py file at /app/main.py.

The bot can also be run un-containerised, either by cloning the repo, or by using the pip package.

⚠️ Compatibility ⚠️

Version 0.19.0+ of Lemmy incompatibally updates the method through which clients interact with the API. By default, this project will work with these newer versions. If your community is hosted on an older instance, the following steps are necessary for the bot to function:

If you are using docker, ensure your container uses versions prefixed with compat-0.18-.

If you are using the package hosted on pypi, explicitly declare your plemmy version as 0.3.11.

If you are running from source, update the plemmy version in Pipfile.

Modules

Different aspects of moderation are divided into "Processors". These scan and report content for a single kind of violation, and can be configured individually. Currently, there are eight different processors:

• BlacklistProcessor - This processor allows moderators to provide a list of blacklisted words. If they are encountered in any comment or post, a report is generated.
• MimeWhitelistProcessor - This processor whitelists MIME-types of files that are permitted in the community. If a comment or post where this doesn't hold is encountered, a report is generated.
• MimeBlacklistProcessor - Same as the previously discussed whitelist, except a blacklist
• PhashProcessor - Calculates and tracks a perceptual hash (phash) for each post. If the image has been uploaded before, the bot posts a comment linking to the duplicates.
• TitleConformityProcessor - Enforces a certain title structure (specified using Regex) and warns a poster if their title does not match.
• UserProcessor - Automatically reports content posted by a listed user.
• ToxicityProcessor - Assesses the toxicity of any post/comment and automatically files a report if it is above a specified threshold.
• PhotoDNAProcessor (untested) - Given a Microsoft PhotoDNA API Key, this processor checks posts for Child Sexual Abuse Material (CSAM). If any violating content is found, the post is automatically deleted. Reports of this nature are sent to the provided Matrix server. This processor is untested as I am unable to obtain an API key.
• SpamImageProcessor - Given a collection of phashes (which can be calculated using the included tool phash_tool.py), will scan through all posts and comments in the community and delete any containing the blacklisted images.