DeepSeek Harness (DSH) achieves a high degree of extensibility through its plugin system. When interacting via the Web client, input mainly relies on the keyboard. If you need to quickly capture ideas or enter natural-language instructions, this approach may not be intuitive. The following introduces the dsh-voice-input plugin, which uses the browser-native Web Speech API to provide Chinese and English voice input for DSH Web.

This is a plugin that adds Chinese and English voice input to the DeepSeek Harness Web input box. It is maintained by wepar1212 and categorized as an online tool. It is based on the browser Web Speech API, requires no server-side support, and does not depend on an API key or a local model.

Core Features

  • Adds Chinese and English voice input functionality to the Web input box.
  • Supports spoken Chinese, English, and mixed Chinese-English input.
  • Relies on the browser Web Speech API.
  • No server required; no API key or local model required.

Installation and Activation

  1. In the root directory of the deepseek-harness source code, run the installation command:
    pnpm dsh plugin --profile web add github:wepar1212/dsh-voice-input#v0.1.0
  1. Restart DSH Web.
  2. Click the microphone icon 🎤 on the left side of the input box to start continuous recognition.
  3. Click the icon again to stop recognition. The recognized result is automatically inserted into the input box.

Typical Usage

After restarting DSH Web, a microphone icon appears on the left side of the input box on the conversation page. Click the icon and start speaking; the plugin converts speech to text in real time. Click the icon again to stop; the text content is inserted into the input box. At this point, you can edit it and send it.

Use Scenarios and Notes

  • Use scenarios: Suitable for scenarios that require quick entry of long text or mixed-language content on the Web and where privacy requirements are relatively relaxed.
  • Notes:
    • The plugin’s recognition locale is set to zh-CN. The actual accuracy of Chinese, English, and mixed speech depends on the speech recognition service used by your browser.
    • Requires Chrome or Edge browsers that support the Web Speech API.
    • Speech recognition may be processed by the online service of the browser vendor. Before use, review your browser’s privacy policy to decide whether to use it.
    • Verified on DeepSeek Harness source edition 0.1.0-rc.5.
    • The plugin runs with the permissions of the current DSH process. Check the source code and license before installation.

With this plugin, you can move beyond keyboard limitations in DSH Web and interact directly through voice. For more plugin information, see the Skill Hub directory or GitHub repository.