Ollamarama is an AI chatbot for IRC which uses offline LLMs with LiteLLM and Ollama. It can roleplay as almost anything you can think of. You can set any default personality you would like. It can be changed at any time, and each user has their own separate chat history with their chosen personality setting. Users can interact with each others chat histories for collaboration if they would like, but otherwise, conversations are separated.
Also available for the Matrix chat protocol at ollamarama-matrix. Terminal-based version at ollamarama
Based on my earlier project, InfiniGPT, which uses OpenAI, available at infinigpt-irc
Install and familiarize yourself with Ollama, make sure you can run offline LLMs, etc.
You can install it with this command:
curl https://ollama.ai/install.sh | sh
Once it's all set up, you'll need to download the models you want to use. You can play with the available ones and see what works best for you. Add those to the config.json file. If you want to use the ones I've included, just run ollama pull modelname for each. You can skip this part, and they should download when the model is switched, but the response will be delayed until it finishes downloading.
You also need to install the irc and litellm modules
pip3 install litellm irc
Fill in the variables for channel, nickname, password and server in config.json Password is optional, but registration is required for some channels.
python3 ollamarama.py
.ai message or botname: message Basic usage. Personality is preset by bot operator.
.x user message This allows you to talk to another user's chat history. user is the display name of the user whose history you want to use
.persona personality Changes the personality. It can be a character, personality type, object, idea. Don't use a custom system prompt here.
.custom prompt Set a custom system prompt istead of the roleplay prompt
.reset Reset to preset personality
.stock Remove personality and set to standard model settings
.help botname Display the help menu
.models Show current model and available models (admin only)
.model name Set a model (admin only)
.model reset Reset to default model (admin only)
.temperature Set temperature value between 0 and 1. To reset to default, type reset instead of a number. (bot owner only)
.top_p Set top_p value between 0 and 1. To reset to default, type reset instead of a number. (bot owner only)
.repeat_penalty Set repeat_penalty between 0 and 2. To reset to default, type reset instead of a number. (bot owner only)
.clear Resets all bot history and sets default model and settings. (bot owner only)
.auth user Add user to admins (bot owner only)
.deauth user Remove user from admins (bot owner only)
.gpersona persona Change global personality (bot owner only)
.gpersona reset Reset global personality (bot owner only)
To get a longer response, you can tell the bot to "ignore the brackets after this message".
When using a coding LLM, remove the personality with the stock command, or set an appropriate personality, such as a python expert.
I have not extensively tested the models included in the json, add and remove models as you see fit. They each have their strenghts and weaknesses. I am using the default 4-bit quant versions for simplicity.