Skip to content

RailGPT v3.0.0

Latest

Choose a tag to compare

@EasonWheng EasonWheng released this 17 Aug 08:30

RailGPT v3.0.0 / 正式发布

中文

RailGPT v3.0.0 是本地优先铁路智能助手的正式稳定版。它在 v2 可观测执行引擎之上,完成了前级语义路由、结构化上下文、Memory OS、能力合同和工具证据门禁的升级。

重点更新

  • 支持 DeepSeek、OpenAI(ChatGPT)与任意 OpenAI-compatible 服务商。
  • Settings 提供主 Key、Thinker Key、三种 Agent 模式的独立模型配置,以及自定义兼容端点。
  • API Key 继续由 Windows DPAPI 按当前用户加密保存;安装包与仓库均不携带用户密钥。
  • FAST-GO 保持低延迟,FAST-PLUSDEEP 保持推理优先;模型变更对下一次请求即时生效。
  • 改进 AgentContextPackage、活动主题帧、日期裁决、工具能力合同、RailGo 运营缓存与事实证据隔离。
  • 发布 Windows 便携版与支持自选安装目录、桌面快捷方式和卸载的正式安装包。

下载说明

  • RailGPT_v3.0.0_Setup_x64.exe:推荐普通 Windows 用户使用。
  • RailGPT_v3.0.0_win64.zip:免安装便携版。

首次启动后,请在 Settings > API 中配置自己的服务商、API Key 和模型。外部数据的覆盖、时效和可用性由相应服务提供方决定。

English

RailGPT v3.0.0 is the formal stable release of the local-first railway AI assistant. It retains the observable v2 execution engine and upgrades semantic routing, structured context, the Memory OS, capability contracts, and tool-evidence gates.

Highlights

  • DeepSeek, OpenAI (ChatGPT), and arbitrary OpenAI-compatible providers are supported.
  • Settings offers primary and thinker keys, independent model configuration for all three Agent modes, and a custom compatible endpoint.
  • API keys remain encrypted with current-user Windows DPAPI; neither the repository nor the installer contains user credentials.
  • FAST-GO remains latency-oriented, while FAST-PLUS and DEEP remain reasoning-oriented. Model changes apply to subsequent requests without a restart.
  • AgentContextPackage, active topic frames, date resolution, capability contracts, RailGo operational caching, and fact-evidence isolation are improved.
  • A Windows portable package and a professional installer with selectable installation directory, desktop shortcut, and uninstaller are provided.

Downloads

  • RailGPT_v3.0.0_Setup_x64.exe: recommended for standard Windows installation.
  • RailGPT_v3.0.0_win64.zip: portable no-install build.

After first launch, configure your own provider, API key, and model IDs in Settings > API. Coverage, freshness, and uptime of external data are controlled by their respective providers.