joonspk-research/generative_agents ? reverse-engineered prompt

Hey, can you build me a simulation environment for generative agents, based on that popular Stanford research paper? I want to create a small, 2D pixel-art town where multiple AI characters can live and interact with each other and their environment. Think of it like a tiny, self-running version of The Sims where the characters have their own goals.

These agents should be powered by a large language model, so I'll need an easy way to plug in my OpenAI API key. I want to be able to run everything locally. The main view should be a simple web page I can open on my browser that shows the town map. On it, I can watch the agents move around and go about their simulated lives. It should also have basic controls, maybe through a command line, to start the simulation, run it for a certain number of steps, and then save the results. I'd also love to be able to replay or demo a saved simulation.