This is a submission for the Cloudflare AI Challenge.
What I Built
I built a comment system moderated by AI, eliminating the need for a human intervention and saving countless hours on reading useless comments full of spam or profanity.
Demo
https://cf-challenge-comment-system.pages.dev/
My Code
https://github.com/Xubaer/cf-challenge-comment-system
Journey
I never knew how powerful cloudflare really can be and it has pages and workers functionality.
I used cloudflare Ai to moderate comments from users using two different models, one to moderate the comments and the other for sentiment analysis which helps showing the up/down vote counts :D
I hope to make this as a white label project which can be integrated and injected into different websites and apps. It would work as plug and play.
Multiple Models and/or Triple Task Types
Used "@hf/thebloke/openhermes-2.5-mistral-7b-awq" to moderate comments and receive a json format about the comment.
"@cf/huggingface/distilbert-sst-2-int8" for sentiment analysis that helps me add count to the up/down votes for the overall comments.
Top comments (2)
I tried this comment:
My comment was rejected. I was told to communicate in a civil manner and not be insulting... when I was already doing both of these.
Just goes to show that there are many things where human judgment is needed (moderation being an obvious example), because AI understands essentially nothing.
That's true, the optimal solution is to flag such comments and then be reviewed by a human. I will improve this, since I was taking part in cloudflare ai challenge I didn't had much time adding this but I'll do that and provide an update.
Thanks for the feedback.