Repolode
4thfever

4thfever/cultivation-world-simulator

An AI-driven Xianxia (cultivation) world simulator where every cultivator is an LLM-powered agent that acts within a rule-based world.

Stars
2,116
Forks
238
Open issues
1
Last commit
2 months ago

What is it?

Cultivation World Simulator is a text-and-map game in which you play as the "Heavenly Dao" observing a Xianxia world that evolves on its own. Each cultivator is an independent agent driven by an LLM, with its own personality, memory, relationships and goals, and the world runs on a rule system covering spiritual roots, realms, techniques, sects, pills, weapons, auctions and lifespan.

The rule layer is meant to keep AI behavior grounded, so events such as sect conflicts, sect-versus-demon rivalries and the fall of talented cultivators emerge from the world logic rather than a scripted storyline. Players can watch the world unfold or intervene with events like heavenly tribulations and changes to a cultivator's mind.

The project offers a desktop build on the Epic Games Store, Docker deployment, and a source-based development mode using Python and a Vue/TypeScript web frontend. It also exposes query and command HTTP APIs so external scripts or agents can observe the world and make interventions.

Key features

  • 01Each NPC is an independent LLM-driven agent with personality, memory and relationships.
  • 02Rule system covering spiritual roots, cultivation realms, techniques, sects and lifespan constrains AI behavior.
  • 03Emergent storylines from sect rivalries, large events such as auctions and martial tournaments.
  • 04Players can act as the Heavenly Dao and intervene with tribulations or altered minds.
  • 05Supports Docker deployment, a desktop build, and source development with configurable model presets like DeepSeek, MiniMax and Ollama.
  • 06Provides query and command HTTP APIs for external agents to observe and influence the world.

Use cases

  • Playing a sandbox cultivation game where the story is generated rather than scripted.
  • Studying multi-agent LLM behavior inside a constrained rule world.
  • Building external agents that observe and intervene in a simulated world through its API.
  • Prototyping LLM-driven NPC systems for narrative games.

Quick start

  1. 1

    Install backend dependencies

    pip install -r requirements.txt
  2. 2

    Install frontend dependencies (requires Node.js)

    cd web && npm install && cd ..
  3. 3

    Start the server with frontend in dev mode, then configure a model preset in the settings page

    python src/server/main.py --dev
  4. 4

    Open the frontend address shown in the startup log (usually http://localhost:5173)

Things to know

  • You need a working LLM model service, such as DeepSeek, MiniMax, or Ollama, to play.
  • The README marks the Docker deployment as untested.
  • Mobile UI is not fully adapted yet.

Topics

FAQ

What is cultivation-world-simulator?
An AI-driven Xianxia (cultivation) world simulator where every cultivator is an LLM-powered agent that acts within a rule-based world. Cultivation World Simulator is a text-and-map game in which you play as the "Heavenly Dao" observing a Xianxia world that evolves on its own. Each cultivator is an independent agent driven by an LLM, with its own personality, memory, relationships and goals, and the world runs on a rule system covering spiritual roots, realms, techniques, sects, pills, weapons, auctions and lifespan.
Is cultivation-world-simulator free to use? What license does it have?
cultivation-world-simulator is open source under the Other license. Check the license terms before using it in commercial products.
Is cultivation-world-simulator still maintained?
Updated within 6 months. The last commit to cultivation-world-simulator was on Aug 16, 2026.
What can I use cultivation-world-simulator for?
Playing a sandbox cultivation game where the story is generated rather than scripted. Studying multi-agent LLM behavior inside a constrained rule world. Building external agents that observe and intervene in a simulated world through its API. Prototyping LLM-driven NPC systems for narrative games.
How do I install cultivation-world-simulator?
Install backend dependencies: pip install -r requirements.txt Install frontend dependencies (requires Node.js): cd web && npm install && cd .. Start the server with frontend in dev mode, then configure a model preset in the settings page: python src/server/main.py --dev Open the frontend address shown in the startup log (usually http://localhost:5173)
Where can I get cultivation-world-simulator?
The source code and installation instructions are on GitHub: https://github.com/4thfever/cultivation-world-simulator

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects