MiladyGPT

A minimalist discord bot built using gpt-3.5-turbo and langchain.

You'll need to setup a Discord bot and get an API key from OpenAI to get started. You can put them in a .env file in the root directory of the project for local development/utilization or use environment variables in production.

Getting Started

Local

pip install -r requirements.txt
python bot.py

Docker

docker build -t miladygpt .
docker run -d miladygpt --env-file .env

Usage

To get a response from the bot, simply mention it in a channel and it will respond with a generated response. There are also some additional commands to let you tweak the bot's behavior at runtime:

• $show_prompt: Show the current prefix prompt used by the chatbot
• $update_prompt <prompt>: Set the prefix prompt used byt the chatbot
• $reset_prompt: Reset the prefix prompt to the default specified in bot.py
• $show_temperature: Show the current temperature configured for the chatbot
• $update_temperature <temperature>: Set the temperature used by the chatbot
• $reset_temperature: Reset the temperature to the default specified in bot.py

Limitations

Right now it doesn't seem that the ConversationalBufferMemory features in langchain aren't playing nicely with the way chat completion works in gpt-3.5-turbo. This means the bot currently has no memory; will fix once langchain is updated.