QT-Chen/dsh-mic-input

DSH Web 语音输入插件:利用浏览器 Web Speech API,为 DeepSeek Harness Web UI 提供麦克风语音输入功能。

Project Overview项目介绍

This is a native client-side plugin built exclusively for the DeepSeek Harness web UI that adds microphone-powered voice dictation functionality. It adds a mic button to the right side of the DSH composer input toolbar, right next to the send button. All recognition runs directly in the user’s browser using the built-in Web Speech API, so there are no external server components, no API keys required, and no models to download. Audio never leaves the browser except for processing by the browser’s own speech service, which is provided by Microsoft Azure for Edge and Google for Chrome.

To install the plugin, you can run the dsh plugin --profile web add github:QT-Chen/dsh-mic-input command via the DSH CLI, or install the prebuilt version from npm with dsh plugin --profile web add dsh-mic-input. After installation, you will need to restart dsh web or refresh the browser page to activate the plugin. On first run, your browser will ask for microphone permission, which you can safely grant because 127.0.0.1 and localhost are considered secure contexts for this permission. If you install from GitHub source, you may need to approve a one-time build for pnpm ≥ 10, but npm installs do not require this step.

During usage, click the mic button once to start recording, and click it again to stop and keep the transcribed text in your input box. The plugin includes multiple quality-of-life features, like live interim transcription, deduplication of repeated text, auto-reconnection after silent pauses to avoid cutting long sentences, and automatic smart punctuation that matches your input language. You can adjust settings like recognition language and punctuation mode from the DSH settings panel under General > Voice Input. The plugin stores no audio or text locally, so your privacy is fully protected, and it is released under the open source MIT license.

这是专为DeepSeek Harness(DSH)网页版开发的原生插件,为DSH的Web界面输入框工具栏添加了麦克风语音输入功能。整个插件完全在浏览器客户端运行,使用浏览器内置的Web Speech API进行语音识别,无需额外的服务器组件,也不需要配置API密钥。

点击输入框旁的麦克风按钮即可开始录音,语音会实时转写为文本填入输入框。插件解决了Web Speech API常见的重复文本问题,支持自动断句补标点、静默后自动重连、识别结束自动发送等功能,用户可以在DSH设置面板中调整识别语言、标点模式等参数。

安装可通过DSH命令行工具从GitHub仓库或npm源添加,安装完成后重启DSH网页端或刷新页面即可使用。首次使用时,浏览器会请求麦克风权限,本地localhost环境是安全上下文,可正常授权使用。识别效果受浏览器自带引擎限制,插件以MIT许可协议开源。

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 dsh-mic-input

把 QT-Chen/dsh-mic-input 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

dsh-mic-input

Microphone voice input for the DeepSeek Harness (DSH) Web UI — in the browser, no server, no keys.

中文说明见 README.zh.md

A pure client plugin that adds a microphone button to the composer tool row of the DSH Web GUI. Click to start dictating; recognized speech is transcribed live into the input box. Everything runs in your browser through the built-in Web Speech API:

  • Microsoft Edge → Microsoft Azure speech service
  • Google Chrome → Google/Chrome speech service

Zero server components, zero API keys, zero model downloads. Audio never leaves the browser (except to the browser's own speech service).

Features

  • Mic button in the composer tool row (right side, next to send) — click to start/stop; red pulsing ring while listening
  • Live transcription — interim results fill the input while you speak
  • No duplicated text — the draft is always rebuilt from the text that was in the input when recording started, and interim echo prefixes are deduplicated (fixes the classic "重复文字" bug of Web Speech plugins)
  • Auto-continue on silence — if the browser stops listening after a pause, the plugin silently reconnects and keeps the session (up to 6 silent ends), so long sentences are not cut off
  • Smart punctuation (default) — auto-completes sentence-ending punctuation: ? after 吗/呢 questions, ! after 吧/啊/呀/哦/嘛/啦/哇/哈, 。 otherwise (English: .); modes: 自动补全 / 保留原样 / 去除
  • Language selection — 跟随系统 (auto, follows browser language) / 中文 (zh-CN) / English (en-US); wrong language matching is the #1 cause of garbled recognition
  • Auto-send — optional: submit the message automatically when dictation ends
  • Settings row under 设置 → 通用 → 语音输入
  • Clear errors — permission-denied / network errors show bilingual hints on the button tooltip

Install

dsh plugin --profile web add github:QT-Chen/dsh-mic-input

or from npm (prebuilt, no build approval needed):

dsh plugin --profile web add dsh-mic-input

Restart dsh web (or refresh the page), then click the mic button to the right of the input box. The first time you use it, the browser asks for microphone permission — allow it (127.0.0.1 / localhost is a secure context, so it works).

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 deepseek-harness-tui 下一个 Next dsh-incognito →