I’m a software developer, and I’m currently noodling around with the idea of a small suite of software to help me moderate a couple communities. Right now, the features I’m thinking of are:

  • Lemmy bot integration - the suite would require a bot account.
  • Some kind of content moderation (detecting things like spam, bigotry, etc.) to flag for me to review. Similar to the mod queue on that other site.
  • Automatic removal of certain kinds of content. I figure this would be configurable, like you could choose whether your community is for text posts, for links, for videos, for pics, or some combination of everything.
  • Domain blacklists and whitelists.
  • Moderation record keeping - in other words, a database to keep track of bans, warnings, reports, etc.
  • Canned automatic messages, which would be useful to send when banning, unbanning, responding to certain types of queries, and so on. So if you use the software to ban someone, they automatically receive a notification about what they were banned for, and for how long.
  • Scheduled automatic posts, which would be configured with *.md files for content. Good for things like monthly meta threads.
  • Controls for the bot via JSON objects in private messages, which would direct it to follow commands.
  • Some kind of configurable user commands for the bot. (Haven’t really figured out what this would look like yet.)
  • A simple web interface for all of this.

That’s what I’m considering so far, but I’d love to get more ideas from the broader Lemmy community, and will post it on Github if it comes together.

The technology stack I’m considering is:

  • Node/Typescript
  • The lemmy-bot library
  • Sequelize for the database
  • Nest.js for the web backend
  • [TBD] for the web frontend (I hate React, so not that. Probably either Vue or Angular)
  • Docker for containerization

So what features would other Lemmy mods like to see in a Lemmy moderation suite like this?

Edit: OK, I’ve had a chance to take a look at the code for the existing Lemmy “automoderator” bot that’s in progress, and I think I’m still leaning towards working on one of my own, partly because it looks like there’s a real desire for Lemmy bots that use Node. I think there’s room for two or more in the fediverse. And hopefully, we can all learn from each other. I know I’m bookmarking the Python bot and watching to see how it improves.

Anyway, I’d like to thank everyone for the suggestions! I’m hoping to get started on it this weekend. We’ll see how it goes.

  • @[email protected]M
    link
    fedilink
    English
    8
    edit-2
    11 months ago

    Something similar to this is already in development.

    It’s written in Python and utilizes the Pylemmy library to access the API.

    So far it:

    • Uses a deep neural network and a custom training dataset to detect toxic content (the training data out there wasn’t really very suitable)
    • Has a hard-coded regexp flagging tool. It shouldn’t be hard to make that more dynamic.
    • Has a probation list to flag users who are ‘being watched’.
    • Will report comments/post for mod action.
    • Will send a message to a selected Matrix room.
    • Keeps track of previously assessed comments/posts using SQLite…

    It’s actively being developed and I intended for it to get a lot more configurable and feature rich.

    Happy for anyone to contribute (I would actually really appreciate it).

    https://github.com/noenfugler/LemmyModBot

    PS: It’s running on this community right now.