I am currently creating a lemmy automod bot. And i would like to know what features you want or like in a automod here?
It is open source too: https://github.com/hdevelopments/lemmy-automod-bot So that everyone can self host itself :)
Development test thread: https://lemmy.world/post/685919 There you will see what i work on :)
The already done/planed features:
- Autocomment
- Automod ( Profanity filter => remove, “Hardprofanity” filter => ban )
- Commands ( This is a bit developer heavy )
- Multi Community Support
- Keyword response
- Discord Logs ( On Comment creation, On Post creation, On report TBD )
- Discord Commands
- Matrix Logs (TBD)
- Matrix commands (TBD)
If you have any suggestions or ideas please let me know :)
Which features are planned and which ones are done? Is the source code available?
I updated the post already :) But i can see that it didnt updated on your side. See https://lemmy.world/post/840745
Everything with a TBD is planned / is already in the doings. And everything without TBD is implemented.
(TBD = To Be Done)
The source code is in the updated post too and you can see it here https://github.com/hdevelopments/lemmy-automod-bot
The config has to be edited to your needs if you want to set it up for you. If you have any further question tell me :)
I would like automod to respond to comments mentioning suicide by providing the user with the national suicide hotline which is simply 988 and can be called or texted.
Here is a list of international hotlines for our brothers and sisters in other countries.
Even when we have been sustained struggling with suicidal ideation, the act itself is often spontaneous and very commonly influenced by substance use (primarily alcohol). I’ve talked people thru/out of suicide attempts, who then improved upon their life and mental health drastically. Sometimes we need to hit rock bottom before we really comprehend the need/have the drive to commit to the hard changes required. Sometimes we just need one person to show they care enough for us to change our minds in our most vulnerable and desperate state. People have reported a stranger simply smiling at them or saying hello was enough to stop them from jumping off a bridge.
Please include a feature like this getting people in touch with support when they desperately need it the most.
Would you think this should be globally active? Or only on the “active” communities?
I think there should be no bots with global actions (as in, interacting with posts and comments that were created on federated instances). Bots should only be allowed to interact with their home instance. Bots on Reddit were really annoying for this, as you would always get replies from the global grammar correction bots that added nothing to the conversation. Besides, because each instance admin will probably be hosting their own version of the bot, if global actions are allowed and several instances enable them, one single comment that contains the trigger can get spammed by several bots.
I meant site wide we meant instance locally only.
Ideally I would like to see it active site-wide. Sometimes small communities are where people feel safe confiding in one another.
But I may not be totally comprehending your question. I’m pretty illiterate when it comes to how automod tools work…
Yeah that what i thought too site wide but some communities doesnt want ( somehow ??? ). But i will deal with it later :)
I will implement that!
Awesome! Thank you so much!
undefined> with the national suicide hotline which is simply 988 and can be called or texted.
ehm, we dont all live in the same country
Can it share rss from our chosen news sources?
What do you mean? Like you choose a RSS and it automatically post it to lemmy? As a post?
Can it post the news? actually this moderation bot is not expected to have such a function but maybe you have another project.
It currently can not do that. But i can implement that!
Thanks! Federation is quite annoying like that. I’m working on a Lemmy AutoModerator in Python that uses the YAML configuration like Reddit’s AutoModerator but it doesn’t take commands from any external source, so it’s nice to have similar projects as a reference. My focus is specifically on the post scheduling module right now because what my communities need most are weekly threads, so that would be a great feature to see in this bot.
(This is a reply to this comment. When I try to reply to your comment directly, it doesn’t sync over to lemmy.world. Your post with the links to the source code isn’t updating on my instance either)
Scheduling modules is a good thing that i could doo too ;)
You guys should join forces
I am sorry to tell you, that i am not going to use python. ;(
Because of personal trauma/opinion/experience.
Cougar: Being honest, we really don’t trust auto-mods.
Because of that i do that open source ;)
Cougar: It’s not that? It’s just that a dumb automaton making judgement calls about moderation rings all sorts of alarm bells for us.
Why are you then on the internet?
It wont do like “hey screw you an AI decided that your comment is bad” rather only if the comment swears hard like N-Word, Hard Insults etc. he steps in hard and do something.
Automod is one of the best tools for such things because it is not biased of anything it just has that insult list and filters it out.
Cougar: It just seems way too easy to game by malicious actors, from our perspective.
Cow: Why do you put “Cougar:” in front of your text? Are you a bot ? XD
And btw i dont mean bad actor bots, it auto marks itself as bot so if he does something he has definitly the bot tag
Cougar: I’m just indicating who’s talking; I share the account with the twelve other people rattling around in this noggin, so disambiguating makes sense.
Pearl: And our concern isn’t someone spamming the bot, our concern is someone outsmarting it. For example, figuring out the list of “tripwires” for a given server and baiting otherwise innocent posters into getting bonked.
It would probably not be 1 Bot for 99999 Comms rather 1 for 5 communities ( pre selected ) and those are hosted by someone who is in their communities mod. And they can just edit the bot to add or remove bad words or filter out other factors. And how do you share an account AND a comment? Or do you share it on 1 pc too?
If a malicious actor wants to spam this. I am sorry to tell you he doesnt need a bot. Lemmy has no real anti bot measures ( other than ratelimit ) So just a simple bot net or just proxies and spam the heck out of a lemmy instance. The lemmy instance is going to have some spam.
I am totally new to being a mod so take my suggestion with a grain of salt. Would it be possible to remove images or content based on a database of known offending content? I grew up in the days of goatse (please please do NOT Google that) and would like such things to be auto removed. I will survive the occasional false negative, but minimizing such things would be appreciated. I know large tech companies have something similar to remove abuse photos.
Yeah it is possible. But it would be a bit difficult and performance heavy.