So the AutoMod feature was added, but I was sad to see it not really having 3rd party integration potential. I suggest allowing any mod to use the command that is currently twitchbot_reject (in the process, renaming it to /reject_message ?) like so: /reject_message <message_id>
This way, custom bots could flag messages for moderator attention instead of just timing them out. The way I would assume it would work is that it would temporarily remove the message, similar to how chat delay works, and wait for a moderator to allow or deny the message.
I’m not sure I follow. The initial reject/flag is generated only when a message doesn’t pass our systems. We don’t allow users to pre-blacklist things yet, so there isn’t any APIs to expose to you (yet).
In the case that we do allow channels to blacklist or whitelist things using a list, we will add an API to do so.
Because of the “your message was held” notice, the chat system needs to make the decision before sending the message to anyone (e.g. custom bots). In order to let external systems make the hold/allow decision, it would probably need to be a new, custom, separate API designed for low latency, as it would be on the critical path for chat.
I think I expressed myself very unclearly. Maybe I can rephrase my OP.
I do not seek to automatically approve or decline messages, I seek to have an alternative to /timeout that instead flags the message for moderator attention.
I am aware that this interferes with the current way things are handled, since 3rd parties can only flag messages that have already been sent out, while twitchbot acts as middleware before messages are sent out. Instead, you would need some kind of CLEARCHAT-esque command that removes the message from the chat of users until it is accepted (or removed permanently)
It was just a thought that would make this feature applicable to all channels, instead of just a small subset of channels that dont use a proper bot yet.
A command like /review or /flag would be fantastic. Basically “have a human mod approve this message.”
It would work just like /timeout, with the clients hiding the message when getting a /review, but then showing the message again when a mod approves it.
With this you could create filters with false-positives, which is something you can’t currently do.
This opens up for some nice custom things like;
“Bad user” detection (i.e. user is in many chats, has done X bad things in the last Y time - We’re not 100% sure, approve or deny this likely questionable message in chat?)
etc.