WorldX

WorldX is an open-source project (MIT license) that generates a complete virtual world from a single sentence. AI characters then live in it autonomously, making decisions, interacting with scenes, forming relationships, holding conversations, and remembering past events. The README marks the project as Alpha, with the core working and still being refined.

What It Is

WorldX is an LLM-driven world generation and multi-agent simulation tool. The user describes a scenario, such as a night market street in Song-dynasty Bianjing with a fortune teller, pawnbroker, thief, constable, and a time traveler. An orchestration pipeline designs the world structure, characters, and rules, and image-generation steps produce the map art and character sprites. The result is played in a Phaser 3 and React 19 client, backed by an Express and SQLite simulation server.

How the Workflow Runs

WorldX uses four separately configured model roles:

• Orchestrator: designs the world structure, characters, and rules.
• Image generation: produces map art and character portraits.
• Vision review: checks map quality and locates regions and elements.
• Simulation: drives runtime character behavior.

The roles use OpenAI-compatible chat/completions endpoints. Image generation can also use Google AI Studio's native image interface. Worlds can be created from the web page at /create or with an npm command-line script. Two pre-generated worlds are included for a quick trial that only needs the simulation model configured.

God Mode and Timelines

Users can intervene as a god: broadcast events, edit character personas or memories, and hold hypothetical conversations with any character. Worlds can run across multiple days with day-night cycles, and one world can branch into several timelines, each with its own SQLite persistence. The interface supports Chinese and English.

Setup Path

It requires Node.js 22.13 or newer, and the built-in node:sqlite module means no native modules need compiling. Users clone the repository, copy .env.example to .env, add model API keys, then run npm install and npm run dev. The client runs on port 3200 and the server on port 3100. The README also covers proxy settings for Node.js when a model provider is not reachable.