Preface

In the DSH plugin ecosystem, client capabilities can be extended to the Web UI through plugins. dsh-voice-input-web is one such plugin: it adds a microphone button to the chat input toolbar in the DeepSeek Harness Web UI, allowing the browser’s Web Speech API to transcribe speech to text in real time and append it to the input draft.

It requires no API key and no server-side changes. Below, we introduce its features, installation methods, configuration options, and usage limitations.

What This Is

dsh-voice-input-web is a DeepSeek Harness client plugin for the Web platform. Its goal is to integrate voice input into the chat input box:

  • Adds a microphone button to the chat input toolbar
  • Click the button to start or stop listening
  • Uses the browser’s Web Speech API for real-time speech transcription
  • Appends the final recognized text to the input draft

The npm package name, plugin ID, and localStorage key referenced in verified sources all consistently use dsh-voice-input-web. The license is MIT. The repository address is:

https://github.com/CrazyGummies/dsh-voice-input

Verified sources do not provide star counts, so this article does not include specific star data.

Core Features

Microphone Button

The plugin adds a microphone button to the chat input toolbar in the DeepSeek Harness Web GUI. Clicking the button starts or stops listening.

Real-Time Transcription Overlay

During listening, an overlay displays real-time interim transcription results.

Final Text Goes to Draft

The final recognized text is automatically appended to the input draft. You can then edit and send it as you normally would with the input box.

Recognition Language

The plugin supports one-click language switching:

  • Chinese: zh-CN
  • English: en-US
  • Auto: auto

Optional Auto-Send

If dsh-voice-input-web.autoSend is set to "1", a completed speech segment will be sent automatically.

Implementation

The plugin is described as zero-dependency, zero-key, with no server-side changes. Recognition is performed via the browser’s Web Speech API; speech recognition requires an internet connection, as the browser calls system speech services.

Installation and Enablement

Online Installation

First, install the plugin:

dsh plugin --profile web add dsh-voice-input-web

After installation, restart the web profile and refresh the browser page:

dsh --profile web

After these steps, the microphone button will appear in the input toolbar.

Offline Installation

For offline installation, follow these three steps:

  1. Place the package into the web profile’s node_modules directory:
   ~/.dsh/profiles/web/node_modules/dsh-voice-input-web/
  1. Append to ~/.dsh/profiles/web/cordis.patch.yml:
   - insert:
       - id: dsh-voice-input-web
         name: 'dsh-voice-input-web'
  1. Restart the web profile and refresh the browser page:
   dsh --profile web

For offline installation, the plugin ID, package directory name, and localStorage key all use dsh-voice-input-web. Verified sources note that dsh-voice-input is already taken on npm by a SenseVoice solution with the same name, so be careful not to confuse the package names.

Typical Usage

Switching Recognition Language

Write the language configuration to the browser’s localStorage:

localStorage.setItem('dsh-voice-input-web.lang', 'zh-CN');

You can also change it to:

localStorage.setItem('dsh-voice-input-web.lang', 'en-US');

Or:

localStorage.setItem('dsh-voice-input-web.lang', 'auto');

Auto-Send

Enable auto-send:

localStorage.setItem('dsh-voice-input-web.autoSend', '1');

Disable auto-send:

localStorage.setItem('dsh-voice-input-web.autoSend', '0');

Use Cases and Considerations

This plugin is suitable for browsers that support the Web Speech API, such as Chrome / Edge, or other browsers that support webkitSpeechRecognition. Firefox does not support the Web Speech API and will show a prompt.

Before use, please ensure:

  • The page is accessed via localhost / 127.0.0.1 or HTTPS
  • The browser can access system speech services
  • Speech recognition requires an internet connection
  • The plugin is a client-side plugin running with the current dsh process permissions; review the source code and license before installation

The verified source indicates the license is MIT.

Conclusion

The value of dsh-voice-input-web lies in adding a low-configuration voice input option to the DeepSeek Harness Web UI input box, merging the browser’s Web Speech API transcription results directly into the chat draft. It is suitable for scenarios requiring fast input where the browser environment supports the Web Speech API.

Verified sources do not provide a directory page URL, so this article does not include a directory page link. The GitHub repository is:

https://github.com/CrazyGummies/dsh-voice-input