ET10 The AI Native Framework
ET10 is a next-generation game development framework designed for the AI coding era.
Unlike traditional game frameworks that only focus on runtime performance, ET10 is built to make AI-assisted game development practical for large-scale online games.
ET10 combines:
Distributed MMO architecture Fiber-based actor runtime Cloud-native server infrastructure Hot reload workflows Automated gameplay testing AI harness skills Unity runtime bridge Analyzer-enforced architecture
so AI can safely develop, test, and maintain complex online game projects without breaking project structure.
Built for AI-Assisted Development
Most AI coding tools can generate code.
ET10 ensures the generated code can:
compile successfully pass gameplay tests obey architecture rules survive hot reload scale across distributed servers integrate safely with existing systems
before being accepted into production.
ET10 includes a powerful analyzer ecosystem to enforce large-scale project architecture, including:
async safety rules entity lifecycle protection package dependency constraints actor usage validation runtime safety checks
This allows AI to work within strict architectural boundaries instead of generating uncontrolled code.
Designed for Large-Scale Online Games
ET10 provides a modern MMO runtime with:
distributed actor model dynamic room scaling service discovery cloud-native deployment location services hotfix workflows runtime observability multi-process server architecture
The framework is designed for long-running, scalable online games rather than small standalone projects.
AI Harness + Automated Verification
ET10 goes beyond code generation.
It provides a complete AI development workflow including:
automated compilation gameplay verification multi-bot testing Unity Editor integration runtime log analysis test harness execution automated validation pipelines
AI is not only able to write code, but also verify whether the game actually works.
Unity Runtime Integration
ET10 includes a Unity runtime bridge that allows AI to interact directly with the engine runtime.
Examples include:
creating GameObjects modifying components controlling Play Mode refreshing assets reading logs running gameplay tests inspecting runtime entities
This enables a true AI-driven game development workflow instead of simple file-based code generation.
Why ET10?
Traditional frameworks are designed for human developers.
ET10 is designed for a future where:
AI writes code analyzers enforce architecture automated systems verify gameplay cloud infrastructure scales dynamically developers focus on game design instead of repetitive engineering tasks ET10 AI writes code. ET10 makes it production-ready.
For running instructions, please refer to Book/1.1Running Guide.md.
开源版本使用ET License,只可以用于学习,不可二次修改开源,商业使用需要购买许可,价格4999,具体请看LICENSE文件
| 包名 | 详细内容 |
|------|------|
| cn.etetet.harness | ai各种skills跟分析器,包括et-code, et-async, et-excel, et-luban, et-build, et-test, et-tdd等skills,还有包分析器, await分析器等等分析器, 主要用于严格限制ai行为 |
| cn.etetet.unitybridge | 该包是Unity Skills, 实现了各种Unity操作,并且极其方便扩展,提出需求让ai实现即可。目前注册的68个指令全部包括:状态与生命周期 Ping、HostState、Compile、Refresh、RegenProject、EnterPlay、ExitPlay、Reload、EditorGetStateRequest、EditorPauseRequest、EditorUndoRequest、EditorRedoRequest;资源 AssetSearchRequest、AssetFindRequest、AssetGetPathRequest、AssetLoadRequest、AssetImportRequest、AssetRefreshRequest、AssetReadTextRequest;场景 SceneGetHierarchyRequest、SceneGetActiveRequest、SceneLoadRequest、SceneSaveRequest、SceneNewRequest;选择集 SelectionGetRequest、SelectionSetRequest、SelectionAddRequest、SelectionRemoveRequest、SelectionClearRequest;GameObject GameObjectCreateRequest、GameObjectDestroyRequest、GameObjectDuplicateRequest、GameObjectFindRequest、GameObjectGetInfoRequest、GameObjectRenameRequest、GameObjectSetActiveRequest;Transform TransformGetRequest、TransformSetPositionRequest、TransformSetRotationRequest、TransformSetScaleRequest、TransformSetParentRequest、TransformLookAtRequest、TransformResetRequest、TransformSetSiblingIndexRequest;Inspector InspectorGetComponentsRequest、InspectorGetPropertiesRequest、InspectorGetPropertyRequest、InspectorFindPropertyRequest、InspectorSetPropertyRequest、InspectorSetPropertiesRequest、InspectorAddComponentRequest、InspectorRemoveComponentRequest;Prefab PrefabInstantiateRequest、PrefabSaveRequest、PrefabUnpackRequest、PrefabGetInfoRequest、PrefabGetHierarchyRequest、PrefabApplyRequest;GameView GameViewGetResolutionRequest、GameViewListResolutionsRequest、GameViewSetResolutionRequest;其它 ConsoleGetLogsRequest、EditorLogRequest、ScreenshotCaptureRequest、UnityTestRunRequest、MenuItemExecuteRequest、BatchExecuteRequest、TestEcho |
| cn.etetet.test | 测试框架,可以模拟整个游戏环境,实现整条客户端跟服务器交互。比如组队功能,可以写一个test,里面创建5个客户端机器人,然后让机器人发送消息实现组队,从而可以测试整个客户端跟服务端的组队逻辑。有了test框架,ai的工作检查变得极其轻松,只需要让ai写test,然后 review自己的test,确保test执行通过即可。不仅成实现逻辑层Test,UI Test也完全可以实现|
| cn.etetet.behaviortree | 行为树包,提供BTNode、Sequence、Selector、Not、Condition、Action等行为树节点定义与运行时支持,并带有Unity编辑器可视化编辑、节点参数配置、运行路径调试和配置代码导出能力,适合实现怪物AI、任务流程、技能释放、复杂条件判断等可配置决策逻辑。 |
| cn.etetet.servicediscovery | 服务发现包,提供服务注册、注销、查询、订阅、心跳、租约和变更通知能力,进程级ServiceDiscoveryAgent会把请求转发到当前服务发现主节点,并支持主节点切换、故障恢复和多客户端并发访问,适合多进程多服务器架构下动态发现服务 |
| cn.etetet.conditionexpr | ConditionExpr包,可以把Excel/Luban中填写的条件表达式编译成行为树运行,支持数值比较、&&、||、!、括号、错误码绑定、多owner key和专用条件节点,适合副本进入条件、奖励领取条件、任务接取条件、功能开启条件等策划可配置条件判断 |
| cn.etetet.spell | Spell包,提供技能与Buff系统,包含Spell/Buff配置模型、运行时施法逻辑、客户端表现组件和相关Unity编辑器工具,并复用行为树配置导出链路,适合实现技能释放、Buff结算、技能表现和复杂战斗流程配置。Unity菜单 ET->Spell->Spell Editor提供集成化的编辑器。甚至直接让ai自己配置即可 |
| cn.etetet.aspire | Aspire包,整合ET框架与.NET Aspire,会根据StartConfig为每个ET进程和副本创建Aspire服务,自动传递SceneName、StartConfig、内外网端口等启动参数,并接入OTLP/OpenTelemetry,适合分布式应用编排、运行观测和监控 |
game develop ai harness支持,ET10遥遥领先! 绝大部分框架要么客户端要么服务端,要想实现完善的vibe coding test框架是非常不舒服的,只有ET10这种前后端一起,才能极其方便开发双端test,直接打通前后端逻辑,联调都不用,test直接完成双端逻辑开发,完成联调工作。逻辑开发的工作量只剩下UI了。
8.1跟8区别不大,代码结构完全没有变化,主要是编译方式的修改,可以在Unity中按F6进行编译,也可以在IDE中进行编译,运行中reload可以先按F6编译,再按F7进行热重载
| 日期 | 榜单 | 排名 | 周期新增 | 总 STAR |
|---|