All notable changes to OpenGame will be documented in this file.
OpenGame is a fork and rebrand of qwen-code, specialized for end-to-end web game creation from natural-language prompts.
Highlights of this initial release:
- OpenGame agent framework: end-to-end pipeline that turns a high-level game design prompt into a fully playable web game.
- Game Skill: a reusable agent capability composed of Template Skill (stable project scaffolding for canvas / Phaser / three.js / etc.) and Debug Skill (systematic resolution of integration and runtime errors).
- GameCoder-27B integration: configurable as the underlying Code LLM, optimized via Supervised Fine-Tuning and Reinforcement Learning for game-engine mastery.
- OpenGame-Bench hooks: integration points for dynamically evaluating agent-built games against playability criteria.
- Rebranded CLI: the agent is invoked via the
opengamecommand. - Inherited from upstream: rich tool suite (Skills, SubAgents, Plan Mode), interactive and headless run modes, OpenAI-compatible API, IDE integrations, and the TypeScript SDK from qwen-code / Gemini CLI.
Note: settings still live under
~/.qwen/settings.jsonand.qwen/settings.jsonfor backward compatibility with the upstream agent runtime. A migration to~/.opengame/is planned for a future release.
For the upstream change history that this project inherits, see the qwen-code releases.