DeepSeek Harness Plugin Hub

发布与管理完整 Harness Profiles，发现适合你的插件。

探索

插件目录环境预设文档中心动态

社区

发布插件联系我们报告问题

相关链接

Plugin Hub GitHubDeepSeek Harness 官方项目系统状态隐私说明
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

独立、非官方社区项目，与 DeepSeek 官方无隶属、授权或背书关系。

Voice Input — DeepSeek Harness 插件（DSH Plugin）
DeepSeek Harness Plugin Hub
ProfilesPlugins分类动态文档登录管理 Profiles
ProfilesPlugins分类动态文档登录
← Plugins
V

@hackerfish/dsh-voice-input

Voice Input

语音输入：对话框麦克风按钮，浏览器内置 Web Speech API（zh-CN）识别，结果填入输入框草稿

插件会安装到这里；不确定时保持 web。

npx -y @deepseek-ai/dsh plugin --profile web add github:hackerFish/dsh-voice-input#c5855427c77a5b5a7e8f2936349a3c70de5478df
README兼容性版本

兼容性与来源证明

Voice Input 以 @hackerfish/dsh-voice-input 发布，当前版本为 0.1.0。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。

DSH 兼容范围
*
运行环境
web
发布来源
github
Registry 更新时间
2026/8/24

版本

0.1.0stable
2026/8/24

相关插件

正在加载相关插件…

最新版
0.1.0
DSH
*
HMR
重启进程
Tree shaking
未声明可安全裁剪
解包体积
未提供
文件数
未提供
Surface
web
许可证
MIT
发布源
github
GitHub
★ 1
周下载
0
最近提交
2026/8/24
查看源码 ↗
README Badge

点击下方 Badge 复制 Markdown，粘贴到 README 即可。

这是你的 Plugin？认领权益 · 优先安全扫描

验证 package.json 声明的 GitHub 仓库，即可管理这个公开页面。认领后，Hub 会优先安排当前版本的安全扫描，并在通过后公开展示结果。

认领这个 Plugin →
报告问题

相关插件

继续浏览 vision-media 分类下经过校验的插件。

Tool Describe Image@linxin666/dsh-tool-describe-image面向模型的 describe_image 工具，用于 dsh Web GUI：通过在兼容 OpenAI 的端点调用视觉语言模型，为文本模型提供图像理解能力，以描述一张图像（本地路径、http(s) URL 或附件引用）。可热插拔 —Modlens@liustack/modlens面向仅支持文本的 LLM 的插件视觉能力，由免费的 Antigravity CLI 提供支持Deepseek Ivideodeepseek-ivideoiPolloWork HyperFrames Video Studio，以及 27 个可编辑视频模板，以原生 DeepSeek Harness 对话视图呈现。Codexdsh-codexChatGPT OAuth、Codex 模型、搜索、read_image URL 支持，以及适用于 DeepSeek Harness 的 gpt-image-2 生成

README

dsh-voice-input · Voice Input Plugin

中文

A voice-input plugin for DSH (DeepSeek Harness): adds a 🎤 mic button to the chat composer. It uses the browser-native Web Speech API (zh-CN) to transcribe Chinese speech in real time; when you finish speaking, the transcript is filled into the input draft for review/editing before you press Enter to send. Zero config, zero keys, pure browser capability (Chrome / Edge).

Features

  • Mic button on the right side of the composer (conversation.input.right slot) — click to start listening, click again or pause to stop;
  • Live transcript with a pulsing indicator while listening; the final text is appended to the current draft (never auto-sends);
  • Clear error hints for denied permission / no microphone / restricted network / unsupported browser;
  • Settings tab "Voice Input" (settings.plugins.tab): browser support status + plugin load status;
  • UI fully localized (zh / en) — follows the DSH language preference automatically.

Structure

src/
  host/index.ts     host half: /dsh-voice-input/health route (recognition itself is pure browser)
  client/index.ts   client half: mic button + Web Speech API + settings page (i18n zh/en)
scripts/
  wrap-client.mjs       wraps the client bundle with the official __ModuleLoader__.load protocol
  self-test-client.mjs  structure / load / registration assertions
tsup.config.ts      dual build: host (esm, node) + client (cjs, browser)
cordis.patch.yml   profile stack insertion patch

Develop / Build

npm install
npm run build   # tsup build + client wrap + self-test

Outputs: lib/host/index.mjs (host), lib/client/index.js (client bundle).

Install into dsh

Local development (a file: reference; reinstall after code changes):

  1. npm run build to make sure lib/ is up to date;
  2. In the target profile's package.json ($DSH_HOME/profiles/<profile>/package.json):
    • add "@hackerfish/dsh-voice-input": "file:<absolute path to the plugin dir>" to dependencies;
    • add "@hackerfish/dsh-voice-input" to dsh.profile.bundles;
  3. Run pnpm install in that profile directory;
  4. Restart dsh and refresh the page — you will see the 🎤 button on the right of the composer.

Install from GitHub:

dsh plugin --profile web add https://github.com/hackerFish/dsh-voice-input

(or add github:hackerFish/dsh-voice-input to the profile's dependencies and run pnpm install.)

Notes & Limitations

  • Recognition depends on the online speech service built into Chrome/Edge; when offline or unreachable, a "network restricted" hint is shown.
  • The first click requests microphone permission (a getUserMedia pre-check); a denial shows clear guidance.
  • One utterance at a time: it stops automatically after a pause (continuous=false), or click again to stop early.
  • The transcript only fills the draft and never auto-sends, so misrecognitions can be corrected before sending.