4thfever/cultivation-world-simulator
基于 AI Agent 工作流的修仙世界模拟器，旨在还原智能、开放的仙侠世界。| An open-source Cultivation World Simulator using Agentic Workflow to create a dynamic, emerging Xianxia world.
What it solves
This project creates a living, autonomous world simulation where every character is an independent AI agent. It solves the problem of static, scripted NPCs in simulation games by replacing pre-set narratives with emergent storytelling driven by Large Language Models (LLMs) and a rigorous rule-based system.
How it works
The simulator combines a strict framework of "cultivation" rules (spiritual roots, realms, techniques, and sect dynamics) with LLM-driven decision-making. Each NPC agent possesses its own personality, memory, and interpersonal relationships, allowing them to observe their environment and make autonomous decisions. The user acts as the "Heavenly Dao" (the world's overseer), observing the evolution of the world or intervening through events like heavenly tribulations.
Who it’s for
- Players who enjoy immersive, emergent simulation games and the "cultivation" genre.
- Developers interested in AI agent frameworks and the intersection of LLM-driven NPCs and rule-based game engines.
- Modders who looking to create their own autonomous AI worlds.
Highlights
- LLM-Driven NPCs: Every character has independent logic, memories, and goals, enabling complex social dynamics and emergent plots.
- Rule-Based Foundation: Prevents AI hallucinations by constraining imagination within a detailed system of cultivation levels, sects, and items.
- Emergent Narratives: No pre-written scripts; stories like sect wars and the rise of geniuses emerge naturally from agent interactions.
- Extensible Architecture: Supports various LLM backends (DeepSeek, MiniMax, Ollama) and provides an external control API for automation and custom agent integration.