litanlitudan/skyagi
A Python package implementing Generative Agents for simulating believable human-like NPC behaviors in role-playing games using LLMs.

SkyAGI uses large language models to create autonomous agents that exhibit believable human-like behavior patterns. The project implements the Generative Agents architecture, where AI NPCs generate contextually appropriate responses in interactive game scenarios. Users can use pre-built character configurations from popular media franchises or define custom characters via JSON files. The system is built on LangChain for LLM orchestration and provides a role-playing game interface demonstrating LLM emergent capabilities in human behavior simulation.