Skip to content

YuGalk v2.0.21-安装包发布

Choose a tag to compare

@yuxin-lin52 yuxin-lin52 released this 29 Jul 05:37

YuGalk 2.0.21

这是 YuGalk 的第一个官方安装包版本。

YuGalk 是一个使用 Agent 开发的开源 Agent 项目,基于 Tauri + React + Rust 构建,将大模型对话、工具调用、本地推理与可扩展的技能 / 插件体系整合进一个轻量、可离线运行的桌面应用。

安装包说明

本 Release 提供 Windows x64 一键安装包 YuGalk_2.0.21_x64-setup.exe(约 327 MB)。安装包已内置运行所需的全部依赖,下载后直接安装即可使用,无需手动配置开发环境

  • 内置 WebView2 运行环境(离线安装,无需联网下载)
  • 内置 Node.js / Python / Git / Playwright 运行时
  • 内置 llama.cpp 推理引擎

主要功能

  • 大模型对话与工具调用:连接 OpenAI 兼容接口,支持多轮对话与工具编排。
  • 本地推理:内置 llama.cpp 引擎,可加载 GGUF 模型在本地离线运行(模型文件需自备)。
  • 可扩展体系:支持三类扩展——纯提示词技能(Skills)、原生扩展包(.galk)、以及 MCP 协议接入。
  • 开箱即用:运行依赖随安装包分发,离线环境下也能正常启动与使用。
  • 轻量桌面体验:基于 Tauri,体积小、资源占用低。

安装步骤

  1. 在下方 Assets 中下载 YuGalk_2.0.21_x64-setup.exe
  2. 双击运行,按向导选择安装目录并阅读许可协议。
  3. 完成安装后启动 YuGalk,在「设置」中配置模型:
    • 使用本地推理:将自备的 GGUF 模型放入「模型」目录并选择;
    • 或使用任意 OpenAI 兼容的云端 / 本地接口地址与密钥。

注意事项

  • 安装包体积较大,是因为打包了运行时与 WebView2,属正常现象。
  • 本地推理需要用户自行准备 GGUF 模型文件,本安装包不包含任何模型权重。
  • 应用窗口任务栏 / Alt+Tab 显示名称为 YuGalk
  • 当前仅提供 Windows x64 版本。

许可证

本项目以 MIT 许可证发布,并附带附加条款:衍生或再分发本软件时须显式标注「基于 YuGalk」及本项目仓库地址,且不得将本软件用于背书宣传。详见仓库内 LICENSE 文件。


YuGalk 2.0.21 (English)

This is the first official installer release of YuGalk.

YuGalk is an open-source Agent project built with Agents, powered by Tauri + React + Rust. It integrates LLM conversations, tool invocation, local inference, and an extensible skills / plugin system into a lightweight, offline-capable desktop application.

About the Installer

This release ships a Windows x64 one-click installer, YuGalk_2.0.21_x64-setup.exe (~327 MB). All runtime dependencies are bundled, so you can install and run it directly after download — no manual environment setup required.

  • Bundled WebView2 runtime (offline install, no network needed)
  • Bundled Node.js / Python / Git / Playwright runtimes
  • Bundled llama.cpp inference engine

Key Features

  • LLM chat & tool use: Connect to any OpenAI-compatible endpoint, with multi-turn dialogue and tool orchestration.
  • Local inference: Built-in llama.cpp engine can load GGUF models and run fully offline (model weights not included).
  • Extensible: Three extension types — prompt-only Skills, native .galk packages, and MCP protocol integration.
  • Zero-config runtime: Dependencies ship with the installer, so it works even in offline environments.
  • Lightweight desktop: Built on Tauri for a small footprint and low resource usage.

Installation

  1. Download YuGalk_2.0.21_x64-setup.exe from the Assets below.
  2. Run it and follow the wizard to choose an install directory and accept the license.
  3. Launch YuGalk and configure a model in Settings:
    • Local inference: place your GGUF model in the Models folder and select it;
    • Or use any OpenAI-compatible cloud / local endpoint with its key.

Notes

  • The large installer size is expected — it bundles runtimes and WebView2.
  • Local inference requires a user-supplied GGUF model; no model weights are included.
  • The app shows as YuGalk in the taskbar / Alt+Tab.
  • Windows x64 only at this time.

License

Released under the MIT License with additional terms: redistributions must clearly state "Based on YuGalk" with a link to this repository, and the software may not be used for endorsement. See the LICENSE file.