The plugin architecture of DeepSeek Harness (DSH) allows extensions on the web side. For developers, keyboard typing speed on the web can become a productivity bottleneck, especially when quickly capturing ideas or switching input methods is required. This plugin provides a direct voice input entry point in the input box.

What It Is

This is a client-side plugin maintained by NewDaNew. It adds a microphone button to the right side of the DSH Web input box to provide speech-to-text functionality. Recognition is fully based on the browser-native Web Speech API, runs locally, does not upload audio data, and requires no external API key or server configuration.

Core Features

  • Input box integration: Adds a microphone button to the input box toolbar.
  • Local recognition: Converts speech to text using browser capabilities.
  • Automatic sending: Supports automatic message sending after recognition.
  • Configuration options: Supports setting the recognition language (defaults to the browser language) and toggling automatic sending.

Installation and Enabling

The official installation script is recommended.

Run the following command in PowerShell:

irm https://raw.githubusercontent.com/NewDaNew/dsh-voice-input/main/install.ps1 | iex

The script automatically locates the DSH directory, downloads the plugin, and writes the configuration. After installation is complete, restart the dsh web process and refresh the browser page.

If you need to install manually:

  1. Download and extract the repository ZIP file.
  2. In PowerShell, enter the extracted directory and run:
    powershell -ExecutionPolicy Bypass -File install.ps1
  1. Restart dsh web and refresh the browser.

Typical Usage

After installation is enabled, a microphone button will appear on the right side of the input box. The workflow is as follows:

  1. Click the microphone button to start listening.
  2. After you stop speaking, the transcribed content is displayed in real time above the button, the recognition result is automatically filled into the input box, and listening stops automatically.
  3. Click the microphone button again to immediately stop the current recognition and insert the content into the input box.
  4. Click the small arrow icon to the right of the microphone button (Settings) to configure whether to enable “automatic sending after recognition” or adjust the recognition language (supports Chinese, English, or follow the browser).

Notes

  • Environment requirements: Requires DSH version 0.1.0-rc.6+ (web profile).
  • Browser compatibility: Use the Chrome or Edge browser (supports the Web Speech API). The button will be automatically disabled in unsupported browsers such as Firefox.
  • State limitations: The microphone button will be temporarily disabled when no session has been entered (hero state) or during message submission or decision stages.
  • Language settings: The recognition language follows the browser language by default (default to zh-CN on Chinese systems).

Conclusion

This plugin is open-sourced under the MIT License. By performing speech recognition locally in the browser, it offers convenience while avoiding the need to upload data. For more details and source code, please refer to the project homepage.