Skip to main content
Version: 1.0.0

Documentation

Main article category

📄️ Moderation Configuration

You can set the moderation categories to exclude in the main GPT action. This is a comma-separated list of categories that you don't want the moderation filter to filter on. This is useful for cases where simple cartoon violence can be misconstrued because of very short prompts. For instance "I bonk you on the head" might get construed as violence by the moderation filter. Not having any excluded categories is a valid choice, and you would just set the variable to not have any text in it for the value.

📄️ Create GPT Bot Lore Text File

Your GPT bot needs to have a personality, backstory, likes, and dislikes. The better you build the deep lore, the more context the bot will respond to. You can “teach” the bot things by changing the wording of your lore file. You are basically digging into the AI’s brain and modifying their personality. Be creative, if you want your bot unhinged, tell them to be unhinged. You can also give negative prompting by telling them not to exhibit certain behaviors. A sample Lore file is provided below. You should save this as a text document. It is important to have a Traits section, Instructions, and a task that you expect them to do.

📄️ How to get the bot to remember things

This section will cover how to handle the bot remembering things about users, or about certain keywords. Each action being sent to ChatGPT is a brand new action where all of the context is sent each time. If the bot gave a previous answer, it doesn’t remember it right now. Implementing this is challenging, as we are using most of the context we can provide to handle the context and keywords and plan to add portions of Twitch Chat to the context in the future.