ProxyNodeHub README

The project's README from github.com/wanvfx/ProxyNodeHub. As one markdown file: README.md.

ProxyNodeHub

GitHub 免费节点仓库监控与聚合工具

English | 中文

  

软件主界面


⭐ 如果 ProxyNodeHub 对你有用,请务必点一个 Star,这是我持续维护这个项目最大的动力。


项目简介

ProxyNodeHub 用于发现 GitHub 上活跃的免费节点仓库:分析活跃度、去重分散的节点、直接给出可用的订阅链接。点一次「搜索并分析」,剩下的梳理工作全部由它完成。

不必再逐个仓库翻找、比对、手动拼接订阅。它会挑出更新最勤、维护最稳的仓库,合并去重后一次导出;也能翻出那些藏在角落、不易察觉的冷门仓库——在免费节点这个领域,冷门往往意味着负载低、活得久,那才是真正好用的宝藏节点。

功能特性

六层探测引擎 特征库学习 → 硬编码映射 → 文件树筛选 → 常见路径探测 → README 解析 → 兜底候选
并发搜索 8 个关键词并发请求 GitHub REST API
活跃度评分 提交频率 + 均匀度 + 自动化 + 隐蔽度,100 分制
节点去重 以 server:port 为键去重,合并导出订阅
四种处理模式 🚀 极速 / ⚖️ 标准 / 🔍 深度 / 🐢 兼容,各自独立并发度
自动代理 13 个加速镜像,自动测速、失效回退、5 分钟结果缓存
搜索历史过滤 已搜索仓库自动跳过(1–30 天可调)
收藏管理 独立视图,重启后保留
特征库学习 自动记住探测成功过的路径,下次更快
日志持久化 日志滚动保存,保留天数可调,一键清理

运行环境

快速开始

  1. 从 Releases 获取构建包——无任何前置依赖请选 ProxyNodeHub_v0.0.1_self-contained.zip。
  2. 解压到任意目录,运行 ProxyNodeHub.exe。
  3. 按 F5(或点击「搜索并分析」)开始搜索与分析。

GitHub 秘钥为可选项,但建议填写:匿名搜索 API 每小时限额 60 次,填入后可提升至 5000 次。秘钥经 Windows 凭据管理器按用户加密存储。

编译

# 自包含版(单文件,约 68 MB,内含运行时)
dotnet publish src/ProxyNodeHub.csproj -c Release -r win-x64 -p:SelfContained=true -o dist/self-contained

# 框架依赖版(约 3 MB,需目标机安装 .NET 8 运行时)
dotnet publish src/ProxyNodeHub.csproj -c Release -r win-x64 -p:SelfContained=false -o dist/framework-dependent

需要 .NET 8 SDK 及 Windows Desktop 工作负载。

关于体积:自包含版无法裁剪。WinForms 不支持 PublishTrimmed(NETSDK1175),因此约 65 MB 是 .NET 8 WinForms 单文件发布的下限。

快捷键

快捷键 功能
F5 开始 / 取消搜索
Ctrl+C 复制最佳订阅链接
Ctrl+L 切换详情面板
Ctrl+D 切换收藏夹
Ctrl+F 聚焦筛选框
Esc 关闭弹窗

配置说明

工具学到的所有状态都放在可执行文件之外,位于 %AppData%\ProxyNodeHub\:

文件 / 路径 用途
known_repos.json 随 EXE 一起分发——已知仓库到订阅路径的映射
searched_repos.json 搜索历史,按条目设置过期时间
log.txt 日志滚动保存,启动时按保留天数清理
特征库 JSON 探测成功过的路径,供 L1 层复用

项目结构

src/
├── Program.cs               # 入口点
├── MainForm.cs              # 主窗体、工具栏、表格、日志面板
├── SubscriptionFinder.cs    # 六层探测引擎
├── GitHubService.cs         # GitHub HTTP + 代理镜像 + 测速
├── FeatureLibrary.cs        # 特征库 + 仓库分类器
├── CustomFeatureLibrary.cs  # 自定义特征码
├── NodeParser.cs            # 节点解析 / 去重 / 合并
├── KnownRepoLoader.cs       # known_repos.json 加载器
├── SearchHistory.cs         # 搜索历史持久化
├── LogStore.cs              # 日志持久化 + 保留天数
├── Settings.cs              # 设置、缓存、收藏
├── Models.cs                # 数据模型
├── Theme.cs                 # 颜色令牌
├── Fonts.cs                 # 字体系统
├── ModernButton.cs          # 按钮控件
├── BufferedGridView.cs      # 双缓冲表格
├── Win32Interop.cs          # Win32 互操作
├── LanguageFilter.cs        # 语言过滤
├── ExportDialog.cs          # 导出对话框
├── FeatureLibraryDialog.cs  # 特征库管理
├── AboutDialog.cs           # 关于对话框
├── ProxyNodeHub.csproj
├── known_repos.json         # 已知仓库映射
└── logo.ico / logo.png      # 应用图标

技术栈

作者

Zoyaya — github.com/wanvfx

许可协议

基于 MIT License 发布。

ProxyNodeHub (wanvfx/ProxyNodeHub)

github-api
Searching Database…

What is this?ProxyNodeHub is a Windows desktop tool that scans GitHub for active free proxy node repositories, scores their activity, deduplicates the nodes it finds, and outputs a ready-to-use subscription link with one click.
Is this a command-line tool or a graphical desktop app?It's a Windows desktop GUI application built with C# WinForms on .NET 8, not a CLI tool or a web service.
How do I get it running?Download a build from the Releases page — pick the self-contained .zip if you have no .NET installed, or the framework-dependent one otherwise — unzip it anywhere, run ProxyNodeHub.exe, then press F5 (or click 'Search and Analyze') to start.
Do I need .NET installed, or a GitHub API key?The self-contained build needs nothing extra, while the framework-dependent build requires the .NET 8 Desktop Runtime on Windows 10+ x64. A GitHub token is optional but recommended, since it raises the search API limit from 60 to 5000 requests per hour.
Any install pitfalls I should know about?The self-contained build can't be trimmed because WinForms doesn't support PublishTrimmed, so it stays around 65MB; the app is Windows-only (x64) with no macOS or Linux builds.
Where does my GitHub token get stored?If you enter a GitHub token, it is encrypted and stored per-user through the Windows Credential Manager rather than in plain text.
How does the activity scoring work?Each candidate repository is scored on a 100-point scale combining commit frequency, commit evenness, signs of automation, and how 'hidden' (low-traffic) the repo is, favoring quietly maintained repos that tend to have less overloaded nodes.
How does node deduplication work?Nodes pulled from different repositories are deduplicated using 'server:port' as the key before being merged into the final exported subscription link.
What are the four processing modes?You can choose between Fast, Standard, Deep, and Compatible modes, each running with its own level of search concurrency.
How does the six-layer detection engine find subscription links?It cascades through signature-library learning, hardcoded path mappings, file-tree filtering, common-path probing, README parsing, and a fallback candidate list to locate usable subscription files inside a repo.
Creator
wanvfx
Size
3 MB
Stars
467
Age
19d
Commits · 90 Days