perchance @ perchance @lemmy.world

Posts

25

Comments

440

Joined

2 yr. ago

Moderating

---

Thanks! Still a lot of work to do, but I'm getting there

Save issue should be solved now 👍

It was a different users ID

Are you saying that Veg (while on mobile) never had/used the 5OND-5a260453c6f17f5a07dd tag. If so, are you confident about that?

The user has same tag on mobile and desktop, one is banned, one is not.

The system is simply detecting that their mobile device was previously using a tag that you banned. You should unban that other tag, and then their mobile device will not be banned anymore.

The system works this way because otherwise people can evade bans by switching IP addresses, which is bad.

@RudBo@RudBo@lemmy.world Should be fixed now - thank you for providing the example URLs! Seems like Cloudflare added a buggy rule to their firewall.

CC @Ruud@ruud@lemmy.world Heads up that currently Lemmy comments can't be edited, and I think it's might be caused by the same issue [Security > Security rules > Managed rules > Cloudflare Managed Ruleset] though I'm unsure which specific rule - I couldn't find an ordered-by-date-added list.

Sorry about this - some sort of strange Cloudflare issue. I think it's related to the reason comments on Lemmy (which also uses Cloudflare) can't be edited right now - same 403 Cloudflare response. Will comment here again when I've fixed it

And yeah I'm getting the 403 whenever I try to edit a Lemmy comment right now. Seems to be a temporary Cloudflare issue. Unsure if it's related to the Perchance save issue - looking into that now.

This might be what you're looking for? https://perchance.org/text-editor-plugin-v1#edit

From the top comment in that channel:

User Veg Is banned when using his mobile but not his computer. I unbanned the ID 5OND-5a260453c6f17f5a07dd and that fixed the issue of his mobile being banned.

Am I understanding correctly that the issue has been resolved? I.e. 5a260453c6f17f5a07dd was a previous alias of that user, which you found and unbanned? Or are you saying they're claiming to not have had that tag before? All I'll say is that I've checked on my end and things seem to be functioning correctly. It's possible they shared an IP with someone who got banned - e.g. sibling, school/college/work wifi. I may need to adjust it so that aliases aren't so 'sticky' - although this would kinda be the same as mods just using a lower ban duration. I'll have a think about this once I get back to comment plugin stuff.

Short answer: Bans target not just the tag that you see, but known 'aliases' of that tag. I suspect you've previously banned that user while they were talking with a different tag.

If the user is adamant that they haven't been banned under an alias, then you can ask them to comment in a fresh/unique channel on their phone and DM me that channel and I can take a look.

I need to do a lot more work on the comments plugin to make it easier for mods to understand this stuff, but I'm too busy right now with text gen stuff.

@RudBo@lemmy.world @kasprzakjoseph24@lemmy.world Sorry for the delay here! Fixed.

Please revert it my device won’t let me enable them

Which device do you have, and why won't it let you enable them? Have you tried asking the person who has administrator access to the machine? There are a lot of web pages which use the confirm popup - it's a very basic web browser feature. If your admin has broken this feature, then they've probably broken many thousands of websites.

Thanks everyone for the reports, and sorry for the trouble there. I think I've fixed at least some of the issues mentioned in this thread. If you find any example prompts/keywords that still cause the deletion issue, then if you could share them here or via DM that would be super helpful and I'll fix asap.

@RudBo@lemmy.world @gr3en@lemmy.world @squonk2112@lemmy.world @Mack2@lemmy.world

Thanks! Should be fixed now. Please ping/post again if not.

Thanks! Fixed.

I think what you want is just to put this on a new line in your lists editor:

    
hugeEmojiList = {import:huge-emoji-list}


  

and then add it like this:

 js

    
    let mainChatPlugin = commentsPlugin({
      channel: "qine-main-chat", 
      width: "100%", 
      height: "350px",
      adminPasswordHash: "4b7a8a671d949baf24756f731091d13018de5c2ab4dd2cc1c1612a6643986933",
      customEmojis: hugeEmojiList, // <-- Add this line
      adminFlair: "👑 Creator",
      // ...
    });

  

Perchance does have IP-based abuse detection/mitigation in place for some aspects of the site, but it's not as easy as you might think, because some IPs are shared by hundreds of people (e.g. some university campuses), and others may even be shared by thousands of people (e.g. browser VPNs).

The proper solution here is to just make the moderation system bullet proof, so it's not possible to generate illegal content in the first place. I've temporarily paused work on text gen upgrade for a few days, and have been working on this today actually.

There's no simple fix, so all the issues won't go away over night, but it will iteratively improve over the next few weeks/months as I make improvements, catch edge cases, etc. Same with trolls on the comments plugin.

Note: If perchance were a big company with lots of revenue then this would be easy. There is no company, no subscription revenue, no plural developers - it's just one person running this whole site. So please keep that in mind when comparing perchance to other websites/companies who might be able to fix issues faster by throwing a few spare devs at it, or spend more money on a smarter AI model for the moderation system, etc. I will fix all this - it'll just take me a bit longer than it would for a company.

I'm assuming that you're talking about this as a non-generator-author. If you are an author you can ban people (see comments plugin page).

If you are not the author of the generator, it should ideally only take a few people in the chat to report them and they'll be banned (for an amount of time that depends on severity, 'recidivism', etc.).

That said, improving the comments plugin moderation system is pretty high on my todo list. Probably not in the next few weeks, but ideally within the next couple of months. I think it can be significantly improved by rewriting some of the core detection logic and upgrading the AI model that it uses.

Ah found the bug - thanks to you both! Fixed. You'll need to re-save your generator. It only affected generators that were saved in the last day or so.

@qinerous@lemmy.world

Perchance doesn't block any traffic other than malicious/DDoS type stuff, so I'm not sure what's happening here. It sounds like your ISP may be having issues, or may be explicitly blocking Perchance for some reason.

Do you know the specific dates that issues started to occur? I can look for blips in the traffic statistics and try to investigate if it's a widespread problem.