Chat commands and their system prompts & ai configs are defined in the toml config files. To get started you can copy config.toml to prod.toml and edit it, then just run ./dave prod.toml
You first start off a chat session with one of the commands you defined in the config.
<knivey> -dave how are you today?
<dave> I am an AI and do not have emotions, so I am unable to feel any particular way. Is there anything I can assist you with?
If you would like to reply to the AI you just use the bots nick, replyhere
Keep in mind only the chat endpoints support replying, the completions api doesn't
<knivey> dave_bird: that's a lame answer
<dave> Oh, sorry about that! Let me try again: I'm feeling as bold and vibrant as a chat bot can get! How about you?
- option to trim last line if it was truncated by max tokens
- optionally remove thinking block from chat history contexts sent back to server
- have a way to view current chat context history
- save all chat history for possible fine-tuning data in the future
- maybe add a way to rank the last response by users
- more general chatbot with context shared by entire channel would be interesting, would involve a long running history, mechanizm to look up past logs?
The goal for this project is to remain a dead simple IRC bot that can interface with openai apis to allow running prompts in the chatroom
Originally based off an early version of birdneststream/aibird however the two projects completely differ now