AI assistant and image analysis
The artificial intelligence module covers two distinct functions, enabled separately: an assistant that answers when a member mentions the bot, and analysis of posted images meant to flag explicit content.
Both are disabled by default and only act once an administrator enables them on the server.
The conversational assistant
Mentioning the bot in a message triggers an answer generated from the text following the mention. There is no command to remember: the mention is enough.
The text is sent to the model provider for the time it takes to produce the answer, with retention disabled, and is then kept nowhere. It is not used to train a model.
Explicit image detection
When the module is active, images attached to messages are analysed, and those identified as explicit raise an alert in the configured moderation channel.
It is decision support, not a verdict: automatic image classification produces both false positives and false negatives, and the sanction stays with a moderator.
What the module does not do
- It does not read messages that do not mention the bot.
- It keeps no conversation history from one message to the next.
- It takes no sanction of its own following an image analysis.