Skip to main content

Prerequisites

...so, if your curiosity has been piqued and you’ve been hunting for a "Try it out" button... sorry, not here. Running an LLM for everyone is just too expensive for us normal folks. But don’t worry, I’ll show you how to get it running locally. There’s just one small hurdle... but hey, consider it the first step in your text adventure—obtain the key! Yes, just like in a classic adventure game, but instead of a rusty old key, you'll need an API key from OpenAI to summon GPT-4. 😉

There are two versions: a console version and a UI version. The UI version comes with Chrome’s great speech-to-text feature, but for getting started, the console version is just perfect.

  • Console Version: The console version can run with user input or voice input by using different Speech-to-Text (STT) engines like Whisper, Faster Whisper, or Google STT.
  • Web Version: The web version uses the highly advanced Speech-to-Text engine of the Google Chrome browser, which is powered by a Google backend.

The prerequisites are required for either the bash version and for the Web version

Clone the Repository​

git clone https://github.com/freegroup/ArcaneAI.git
cd ArcaneAI

Install Python 3.12.7 and Dependencies​

Other versions may work as well, but this is tested and works perfectly.

# Go into the ./game directory to prepare the Python environment
cd game 

# Create a virtual environment to keep the global environment clean
python3.12 -m venv venv 

# Activate the virtual environment
source ./venv/bin/activate  

# Install required libraries
pip install -r requirements.txt

An API Key for Your Preferred LLM​

Due to the use of function calling, the only fully working LLM for now is OpenAI GPT-4o. Other models may work with adjustments to the prompt, but as of now, they require more tweaking, and I haven't gotten them running yet.

I use a .env file or environment variables to expose API keys for the different LLM and TTS providers. Let’s create one and configure it:

Create a file ./game/.env with the following content:

OPENAI_API_KEY=#############your-key##########

# Uncomment and set up additional API keys if needed.
# By default, this project uses OpenAI only.
#
#GEMINI_API_KEY=#####gemini-key##################
#GOOGLE_APPLICATION_CREDENTIALS=./secrets/gen-lang-client-#############.json

# info, debug, error, fatal
ARCANEAI_LOG_LEVEL=info

# Configuration for the web UI
#
MAP_FILE=TheTipsyQuest
USERNAME=<username>
PASSWORD=<password>

now you are ready to run a demo setup...