secretxuan/dsh-computer-use-win

DeepSeek Harness 的 Windows 电脑操控插件

Project Overview项目介绍

This repository is a purpose-built plugin for DeepSeek Harness (DSH) that adds full computer-use capabilities to DSH running on Windows systems. It is the Windows counterpart to the macOS-first dsh-computer-use plugin, filling the gap for Windows-native automation and visual observation for DSH agents. It ships with over 15 distinct tools covering screen capture, window enumeration, UI element discovery, OCR processing, mouse and keyboard input injection, and direct Windows Settings launching. It supports both vision-capable DeepSeek routes and text-only routes, normalizing screenshots to fit model pixel and byte budgets automatically. It works with DSH version 0.1.1-rc.2 or newer, following DSH's official plugin conventions.

The plugin enables DSH agents to complete a wide range of desktop automation tasks on Windows, from launching system settings to interacting with custom-rendered third-party applications. It follows different documented recommended workflows based on whether you are using a vision-capable model or a text-only model. For text-only models, the workflow starts with listing windows and UI elements, falls back to OCR and structured visual observation when UI Automation does not expose enough control details, then executes actions and verifies results. For vision models, you can just capture a screenshot, let the model analyze the image, then execute the required input actions directly.

The plugin has no external native binary dependencies, all capabilities are implemented via Windows built-in APIs called from a sandboxed PowerShell child process. To install, you need a working DSH CLI installation on Windows, and you can install it in one click via Oh My DSH, or manually add it to your profile's cordis.patch.yml configuration. It is released under the permissive MIT open source license, and has several documented known limitations, including that elevated admin windows may ignore input from non-elevated processes, and some VDI environments filter certain key combinations, and UI coverage depends on app accessibility metadata.

这是一个专门为 DeepSeek Harness (DSH) 开发的 Windows 原生计算机控制插件,作为 macOS 优先的 dsh-computer-use 的 Windows 分支,填补了 DSH 在 Windows 平台上的端侧计算机操作能力空白。它提供屏幕截图、窗口枚举、UI 元素发现、OCR 文本识别、鼠标键盘控制和快速打开 Windows 设置等能力,同时适配 DSH 的视觉大模型路线和纯文本大模型路线。

它面向需要让大模型在 Windows 桌面完成自动化任务的 DSH 用户,支持从目标窗口定位到元素点击输入的全流程自动化操作。针对纯文本模型,推荐先枚举窗口、获取可访问 UI 元素,UIA 覆盖不足时用 OCR 和视觉观察生成结构化 UI 地图,再执行对应点击输入,最后验证操作结果。对于视觉模型,可以直接截图让模型分析后再执行操作。

插件没有额外原生二进制依赖,所有能力都通过受限 PowerShell 子进程调用 Windows 原生 API 实现,采用 MIT 许可证开源。已知限制包括管理员窗口可能忽略非提权进程的输入,部分 VDI 会过滤 Win 组合键,UI 自动化覆盖度依赖目标应用的可访问性元数据。安装通过 Oh My DSH 一键完成,也可手动配置到 cordis.patch.yml。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 3 stars - very few users, little community feedback星标只有 3,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add https://github.com/secretxuan/dsh-computer-use-win/releases/download/v0.7.0/dsh-computer-use-win-0.7.0.tgz

把 secretxuan/dsh-computer-use-win 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

dsh-computer-use-win

中文 | English

Windows computer-use bundle for DeepSeek Harness (dsh): screenshots, window and UI Automation element discovery, OCR/visual observation for text-only models, mouse/keyboard control, and direct Windows Settings launch.

With DSH 0.1.1-rc.2 or newer, the built-in deepseek-v4-flash-vision-exp route receives computer_screenshot captures as native image content. DSH normalizes each screenshot to the route's pixel and byte budget, then uses the DeepSeek Files API with an inline fallback. Text-only routes keep the OCR-compatible saved-file flow.

The macOS-first sibling is Anionex/dsh-computer-use; this plugin fills the Windows gap.

What It Adds

Tool Purpose
computer_screenshot Capture the full screen or a specific window. Vision-capable routes receive an image; text-only routes receive a saved PNG path.
computer_list_windows List visible top-level windows with process, pid, title, screen rectangle, and active state.
computer_list_elements Search controls exposed through Windows UI Automation and return names, types, physical-pixel rectangles, click=(x,y) centers, and diagnostics when UIA cannot see inside custom UI.
computer_ocr_image Run Windows OCR on an existing image file and return recognized text boxes with absolute coordinates.
computer_ocr_window Capture a window and OCR it, useful for CEF/Chromium/custom-rendered apps that UIA exposes only as a shell.
computer_observe_image Turn an existing image into a generic text-model UI map: OCR, coarse regions, clickable text centers, and heuristic visual candidates.
computer_observe_window Capture a window and return the same UI map, useful for icon-only controls and custom-rendered apps.
computer_observe_media_controls Observe the bottom/player area and report generic play/pause toggle state without custom pixel scripts.
computer_search_in_window Generic search flow: locate a likely search field, click it, type a query, and press Enter.
computer_locate_text Locate visible text in a window through OCR without moving the mouse.
computer_click_text OCR a window, find visible text, and click the matching text center with an explicit purpose.
computer_open_settings Open Windows Settings through the documented ms-settings: URI.
computer_click Move the mouse and click at absolute screen coordinates.
computer_type_text Type literal text through clipboard paste first, with SendInput fallback. This keeps CJK text reliable under VDI/RDP filtering.
computer_press_key Press a key or key combination. SendKeys is primary; SendInput is used for combinations SendKeys cannot spell.
computer_scroll Scroll the mouse wheel, optionally at given coordinates.
computer_drag Left-button drag between two screen points.
computer_move_mouse Move the cursor without clicking.

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-sparrow 下一个 Next dsh-kimi-formula →