Introduction

When using DeepSeek Harness (DSH), developers usually rely on interface state changes to determine whether an agent has started responding, completed normally, or encountered an error. When tasks run for a long time or users need to operate multiple interface elements at the same time, relying solely on visual feedback may be delayed or may be missed. dsh-answer-sound aims to fill this gap with audio feedback, providing sound support for the answer lifecycle of agents in the Web GUI.

What Is This?

dsh-answer-sound is a plugin that provides sound support for the DeepSeek Harness Web GUI. It listens to the agent’s answer lifecycle and plays the corresponding sound when an answer starts, when it finishes, or when an error occurs. The plugin is maintained by the community and is released under the MIT license.

Core Features

  • Answer lifecycle sounds: Monitors dialogue runtime state in the background and triggers the corresponding sound when the agent starts answering, when answering is complete, or when an error occurs.
  • Type-specific settings: Adds four configuration options to the settings interface (master switch + three sound types), supporting independent volume adjustment, selection of local audio files, and preview.
  • No external resources: Default sounds are synthesized in real time by the Web Audio API; custom audio files are read locally only and are not uploaded to any server.
  • Local persistence: User settings are stored in the browser’s localStorage (with the key dsh.answer-sound-effects.config).

Installation and Activation

Run the following command to install the plugin:

dsh plugin --profile web add github:<owner>/dsh-answer-sound

After installation, restart the dsh web service and open the Settings → General tab to see the newly added sound settings.

Typical Usage

After restarting the service and opening the settings page, locate the General section. Enable the “Master” switch and, as needed, set the volume for Start, Done, and Error, or replace them with local audio files.

Use Cases and Notes

This plugin is suitable for Web interface users who need audio cues to judge agent status. It runs only in the browser; the Node side is a no-op and does not make any network requests. Audio files do not leave your local machine. Please review the source code and license before use.

Conclusion

With this mechanism, developers can confirm an agent’s running state by sound without interrupting their workflow, improving the debugging experience.