This repository implements a Chinese-localized version of the Stanford Generative Agents simulation. It creates a virtual town with 25 AI agents powered by LLMs (primarily via Ollama) that autonomously organize events, form relationships, and exhibit human-like behaviors. The system consists of a simulation engine that runs agent interactions over time, a compression tool to generate replay data, and a Flask-based web frontend for visualizing the simulation. Agents maintain memory, generate plans, reflect on experiences, and interact through LLM-driven conversations and actions. The application supports local deployment with configurable LLMs and includes features like breakpoint resumption and conversation logging.