Auto delete message

[stnert]

Is there a feature to just delete the messages instead of banning the user, when the word selected in the spam filter?

If this feature exists, I didn't find it, could someone tell me, please?

[umputun]

Pls, use discussions for such type of questions

moved

[umputun]

There is no such mode. The closest thing to this would be "soft ban mode," which restricts users instead of a complete ban.

Adding such a mode is not a hard thing to do, but I don't really see what the use case for this would be. The goal of this bot is to detect and block spammers, so I'm not sure why you would want to let them continue to post and fight their spam repeatedly.

[stnert]

Well, I really think it would be more viable at the moment the bot erase the messages containing links automatically, they are not getting to the point of sending links with malware content for example, only potential fake news.

[umputun]

I understand your perspective, but I'm not entirely convinced this should be considered a form of soft-spam. To me, it seems more like a group policy that restricts links rather than a spam detection issue.

Regarding the technical part of this: when you said "when the word selected in the spam filter," I'm not entirely sure what you mean. This bot has several checks and filters, some based on particular "stop words," some based on links-only and picture-only detection. However, about 95% of spam is usually detected by the Bayes classifier, which weighs the tokenized message as a whole. Another filter, not based on a particular word, is a similarity check that operates on the whole message.

Another part of your suggestion is "they are not getting to the point of sending links with malware content, for example, only potential fake news." The bot doesn't check what the page link is pointing to and has no way to know if it's a malware link or something else. Usually, such messages have multiple characteristics typical of spam, and each one of them may trigger spam detection.

My guess is that you would like to have a setting per filter, allowing you to set a threshold for real spam (with full blocking) and another one for the soft-ban (removal only). If this is what you meant, I don't see how it can work in real life. For example, setting the removal threshold for the classifier to something lower (e.g., 50% probability of spam) will lead to more aggressive detection and will likely have a lot of false-positive removals. The same story applies to the similarity filter.

[stnert]

So, let's go. The purpose of the group is as follows:

If to find an open source bot (self-host) like this, it should automatically delete the comment of the user who sends an unauthorized link, with explicitly, the content related to media links, because they are usually the most sent types of links in this group generally, do you think this is possible to be configured, if so, how?

[umputun]

This answer adds more questions, like what "related to media links" means and how this should be detected. But the basic answer is still the same—I don't see it as spam protection but rather policy enforcement. I would suggest checking if there is some bot doing exactly this, as tg-spam is not the one.