Need Your Thoughts on AI-Powered Mod Tool for Flarum

forumcopilot

Hey everyone,

I'm working on an AI-powered add-on for Flarum to help with moderating forums, starting with tackling spam and NSFW content. I know tools like Akismet already exist, but I think there's always room to make things better.

I'm a long-time forum user and believe AI can really improve moderation. This project isn't commercial—at least not at the start. I just want to create something useful for our community.

I'd love to hear your thoughts:

Do you have issues with spam and NSFW content?
What features would you love in an AI mod tool?
Any specific moderation pain points you think AI could help with?
How happy are you with your current spam/NSFW solutions?
Interested in beta testing?
Also, I’m not into AI-generated content like reply bots. I just want to make moderation easier. I've looked into Discourse AI, but it's super pricey for Enterprise users. I think it should be more affordable.

Your feedback would be super helpful! Looking forward to hearing from you.

Thanks!

luceos

forumcopilot there are a few initiatives that use ai inside an extension, for instance:

Each implementation is different though and so I think you need to figure out what approach you will take that suits your needs or the needs of a community.

Would love to see what you build 👍️

Chope

A good feature should be flagging post if

It contains link.
Contains certain keywords.

And the user

Is in certain usergroups
Registered account in < days old.

What to do

Flag post and put it in moderation (ie hiding it from users views)
Suspend users.

forumcopilot

Chope Thanks this is exactly what I am thinking to work on. I will probably release a beta extension which simply flag posts if it is a spam. And/or check and reject user uploading inappropriate image.

One thing I notice is that the AI with the proper prompt, is incredibily smart and accurate to find out whether a post is a spam or not, not just to rely on certain keywords or have a link etc. Same for the vision ability to identify NSFW images.

I will post something hopefully soon and you guys can give more feedback!

treyb

AI is smart. I think it could easily flag post that are not relevant to the topic or overall thread discussion.

Dr_Sommer

Its a pitty that you dont like o AI-generated content & reply bots....
This would be so great for Flarum.... 👍️
For example: you Ask the AI with "@AI : what should I know about xxxx?"
And after sending, the AI reply would be in the next post...

OrdinaryJellyfish

I have actually created a premium extension using Google's Natural Language and Vision APIs to do this. It flags unsafe images and posts that relate to certain topics etc. It works great for NSFW content. I've not messed with it in a few months, so I'm sure you have some great ideas that I haven't implemented! I'd love to see you take it further as I'm not super well-versed in AI so I'm not sure how to make mine better, so please feel free to reach out if you'd like to use my extension as inspiration or work together!

Edit to add: if you want to reach out, my Discord is OrdinaryJellyfish!

luceos

OrdinaryJellyfish I have actually created a premium extension using Google's Natural Language and Vision APIs to do this.

So forgot about that one 🙈 https://flarum.org/extension/ordinaryjellyfish/automod

forumcopilot

OrdinaryJellyfish Thanks! I did try the Google Vision API but it seems the GPT-4o is much better. I am going to check out your extension!

OrdinaryJellyfish

forumcopilot The main reason I went with Google’s APIs over OpenAI’s is Google offers free quotas, which is good for keeping costs low for smaller communities. I do agree OpenAI might be a bit better! I have also been looking into Gemini as they also offer free quotas but apparently I suck at it 😅

forumcopilot

OrdinaryJellyfish

Thanks for the insight. If being free is absolutely important, we can make use of the OpenAI Moderation API which is free, able to detect problematic content such as hate / harassment etc, although it doesn't work with spam detection.

We will get something out in next two weeks once I can put my hands on, and hope to get some feedback from you guys!