/Whale-In-The-Shell

A crypto trading simulator with LLM agents.

Primary LanguagePythonMIT LicenseMIT

Overview

Whale in the Shell (WITS) is a crypto trading simulation game leveraging Large Language Models (LLMs) as agents and game masters. The game aims to simulate a dynamic trading environment, where agents, influenced by market news, opinions, and their unique traits, interact and make decisions.

Features

  • Agent-Based Simulation: Agents with unique personalities and decision-making processes, simulating diverse market participants.
  • Dynamic Event Generation: A game master module that creates realistic market scenarios, affecting agent actions and market conditions.
  • Simple World Model: The world model centers around a social media platform, forming the core of the simulated environment. Interactions and information flow on the platform influence market dynamics and agent decisions.
  • Simulated Crypto Market: The game features a persistent market environment with a variety of tradable assets created by the game master. Agents can interact with this market, making buying or selling decisions based on their analysis and the evolving market conditions.

Technical Specifications

  • Frontend: .NET Blazor server.
  • Backend: gRPC service written in Python.
  • Database: Supabase.
  • API Integration: OpenAI API for agent behavior and decision-making processes.