where is an OK place to experiment with bots?

maengd@sh.itjust.works to Asklemmy@lemmy.ml – 51 points –

I used code from a github repo to make a lemmy repost bot from a reddit sub.

I tested it out, it seemed OK. So I let it run over night.

When I got back I found out it had been posting the same thing over and over again every few minutes. The account was banned for spam. But in the meantime it was very annoying to people. Also, there are a bunch of posts that can't be removed because it's impossible to remove federated stuff.

Is there a responsible way to test this stuff?

I don't want to make spam, be annoying, etc. I feel bad about the spam bot.

24

Yes. Host your own Lemmy instance. It's not that hard and can be run locally with no federation.

Technical problems aside, no one wants repost bots from Reddit. Nothing makes people unsubscribe faster from a sub than a ton of bot posts without comments.

I thought these bots would be useful to quickly steal and or crosspost the content. But crossposting is... imperfect, and the bots arr far more annoying than useful. Oof.

Ya I planned for it to be in a separate community from the native fediverse. Idea was to allow people to sub to the reddit repost if they wanted.

It is lucky that I did that because otherwise this oopsie spambot thing might have got the native community in trouble, people unsub, reported, banned etc.

How is that any different from organic posts without comments, though?

Ideally posts and comments are somewhat in a relationship. For a community with little traffic, posts will occur scarcely, and few subscribers will add and read comments in their time.

If the community has few subscribers but many posts, any commented posts will be scrolled down so much that engagements stagnates.

So yes, you are right. If a small community has a single human poster who posts all the time without commenters engagement, it’s the same problem.

But a bot almost guarantees that any community it posts runs into that situation. You can’t automate human engagement.

And if you comment on a organic post, at least the original author is probably gonna read it. Engaging with a bot post is just wasted time typing something out that noone will ever read.

There are communities specifically for testing. Maybe one of them would be willing to have you? Or make your own testing community?

!test@sh.itjust.works

!test@lemmy.ca

!test@lemmy.ml

It won't fix the spam problem if stuff goes south though with things like repost bots. That's a lot of excess traffic. OP definitely should host their own instance.

Or just ask an admin of your instance for permission. Regarding traffic, Lemmy has rate limiting. Ask me how I know.

As an old programmer, always build in checks for your systems. Keep a cache of posted articles and check it before posting so you know you haven't posted this one yet.

When you let something run overnight, that's going to go south somehow. If running overnight for the first time, throttle it to one post per hour. And not the same post. I'm the morning you check if it successfully posted a new article once per hour. Next let it post a little more frequently. Ease into your desired frequency once you have figured out all your edge cases and scale issues.

And so forth.

Also, don't let it run against the live system the first time. Implement a "dry run" option which doesn't post to Lemmy, but writes to a file or something similar. This way you can make sure nothing goes wrong in the parts aside from the Lemmy integration.

I was going to add this but figured I had already given too many points. This is the way, though.

This is great advice, and to the OP, don't feel bad. You're really not an IT person of any caliber until you have experienced when I like to call the "Production Incident Experience", or PIE. IT work is a job with unforseen consequences and hurdles, and we've all run into them at one point or another.

This being a learning experience, do what we've all done and learn from it. Now you can set up logging, whatif, sandbox instance, whatever you have to do.

You're on the road to becoming a good programmer - just learn from your mistakes, do your research on best practices, ask intelligent questions, and in no time at all you'll be writing one of these posts yourself.

I think https://enterprise.lemmy.ml is specifically for testing. It's not typically federated with other instances, so it won't be a problem if your bot goes crazy again.

this page also mentions https://voyager.lemmy.ml/ and https://ds9.lemmy.ml/

voyager and ds9 are front ends for lemmy

Lemmy has an HTTP API for clients and frontends.

For others, see https://github.com/dbeley/awesome-lemmy

I don't think they help anything here. But thanks :)

edit: I can't tell what enterprise.lemmy.ml is for? anyone else?

Bots are clients that use the HTTP API

Voyager was set up to test the app, but that doesn't mean other clients can't use it.

Enterprise is full of random test communities that have many been populated by bots. I don't understand how something like https://enterprise.lemmy.ml/c/mels_test (to pick a random one) isn't useful for what you're trying to do.

These instances are specifically made for people testing code that interfaces with Lemmy. This includes front ends and bots.

You should not use any communities on real instances since people will see your testing on their local page.

good call asking for a proper venue to test this, but how do you mean you can't remove federated stuff? i was under the impression (from lemmy's homepage) that one of the features is 100% complete deletion by replacing post/comment content with 'removed by user'. is this not the case?

1 more...

stub out collaborating servers and test it in isolation. Run it on a vm with limited network access.

Can you make your own community? If no one is subscribed to it, it shouldn't federate anywhere, right?