DeepSeek Harness (DSH) is an extensible framework that supports feature enhancement through plugins. For web users, voice input is a high-frequency requirement, but solutions relying on the Google Web Speech API cannot be used under mainland China network conditions. The dsh-voice-input-cn plugin addresses this pain point by using a local Python bridge as a relay, enabling China-accessible voice input based on Alibaba Cloud DashScope ASR.

What Is This

This is a DeepSeek Harness Web voice input plugin for mainland China, maintained by Schumchanvi. It solves the issue where the original version cannot be accessed in China due to its dependency on the Google API, and implements speech-to-text using Alibaba Cloud DashScope ASR and a local Python bridge.

Core Features

  • Microphone button in the input box: Click to start speaking; recording stops automatically after 1.5 seconds of silence.
  • Real-time insertion: Recognized text is inserted into the input box in real time at the cursor position, without overwriting existing draft content.
  • Language switching: Supports Chinese / English.
  • Automatic sending: Optional “auto-send after recognition.”
  • Local relay: Relayed through a local bridge (Python), connecting directly to Alibaba Cloud within China for low latency.

Installation and Enablement

1. Install the Plugin

Use the DSH CLI to install the plugin:

dsh plugin --profile web add github:Schumchanvi/dsh-voice-input-cn

2. Install Bridge Dependencies

A Python 3.10+ environment is required. Install the necessary dependencies:

pip install dashscope websockets

3. Start the Local Bridge

Start the local Python service, listening on 127.0.0.1:8765:

python bridge/voice-bridge.py

4. Configure the API Key

  1. Refresh the DSH page.
  2. Click the microphone button next to the input box, then click the arrow on the right to open settings.
  3. Enter your Alibaba Cloud DashScope API Key (available from the Alibaba Cloud Bailian console).
  4. Select a language and save.

Typical Usage

  1. Click the microphone button next to the input box to start speaking.
  2. After you finish speaking, the recognized text is inserted into the input box in real time.
  3. If automatic sending is disabled, you can continue editing and send it manually; if enabled, the recognized text is submitted directly.
  4. Click the microphone button again or pause for 1.5 seconds to stop recording.

Use Cases and Notes

  • Use cases: DSH users who need to use voice input under mainland China network conditions.
  • Security notes:
    • The API Key is stored in the browser’s localStorage. This plugin does not upload the key to any third party; however, other scripts in the same browser can read the key. Do not use this on shared computers.
    • The bridge listens only on 127.0.0.1 and is not exposed to external networks.
    • Audio is sent only to Alibaba Cloud DashScope for recognition, and the server does not retain it.
  • Runtime permissions: The bridge runs as a local process. Ensure that your firewall allows local connections.

Summary

By introducing a local Python bridge, this plugin resolves the authentication issue with Alibaba Cloud DashScope ASR (direct browser requests would return 401) and provides a smooth voice input experience under mainland China network conditions. For developers or users who need to interact with DeepSeek Harness using voice, this is a practical supplementary tool.

GitHub repository | Plugin directory