Texas Holdem Agents is a poker simulation designed to be played by some combination of human and ai players. The player interface is a Python class that inherits from model.player.
The simulation currently includes baseline rules based players, a human player cli interface, and two LLM agents using models from Anthropic and OpenAI. If you use an LLM model, you need to configure your API access keys as an environment error. If you miss the API keys, you should see a helpful message.
When you run a tournament, results of each hand are saved to a csv file in the results folder for later analysis.
There's not much for an interface yet. You can define a tournament or series of them as I have in src/main.py. The easiest way to get started is to install python requirements and then run main.py. You can modify it to change to the player mix to your liking. If you run it with a Human player, it will prompt you for input.
The source code is rooted in the src directory.
cd src
` Install dependencies
pip install -r requirements.txt
Run the simulation
python3 main.py
Currently, Texas Holdem Agents implements these agents.
Chad Gipiti is powered by OpenAI's API's. The state of the game is passed in a prompt to ChatGPT with instructions to return it's response in json format.
Claude is powered by Anthropic's API's. The state of the game is passed in a prompt to Claude with instructions to return it's response in json format.
Callie Calls is a rules based player that will attempt to always CALL.
Ricky Raise is a rules based player that will attempt to always RAISE
Robbie Random is a rules based player that will play a random action
Floyd Folds is a rules based player that will always FOLD
Paula Pairs is a rules based player that will play any pair
Johnny Jacks is a rules based player that will attempt to play hands with Jacks or Better