Introduction¶
The DeepSeek Harness Web UI supports extending functionality through plugins. When interacting via the Web interface, Chinese input often requires frequent typing. This plugin addresses input efficiency by leveraging the browser’s native speech recognition capabilities, making voice-to-text operations lighter.
What Is This¶
This is a Web UI Chinese speech input plugin maintained by Dee3526.
It does only one thing: click the microphone button to the right of the chat input box, recognize Chinese speech into text, and append it to the input box. After recognition is complete, it does not send automatically; the user can still edit the content manually.
Core Features¶
- Speech-to-text: After clicking the microphone button, speech is recognized as text.
- Append input: Recognition results are appended directly to the current input box content instead of replacing it.
- Manual send: Messages are not sent automatically after recognition ends; the user can review or modify the text.
- Language restriction: Recognition language is fixed to
zh-CN. - Platform restriction: Only the Web UI is supported; desktop and other clients are not supported.
Installation and Enablement¶
Run the following command in the terminal to install the plugin:
dsh plugin --profile web add https://github.com/Dee3526/dsh-plugin-voice-input-demo.git
If the DeepSeek Harness Web service is already running, restart the corresponding process after installation.
Typical Usage¶
- Open the DeepSeek Harness Web UI.
- Click the microphone button to the right of the input box.
- On first use, the browser will request microphone access permission; choose Allow.
- After you finish speaking, click the stop button again, or wait for the single utterance recognition to end automatically.
- Review the text in the input box, edit as needed, then manually click Send.
Applicable Scenarios and Notes¶
- Applicable scenarios: Suitable for users who need to enter Chinese quickly, create speech-to-text records, or assist typing in the Web UI environment.
- Browser compatibility: The plugin relies on the browser’s
SpeechRecognitionorwebkitSpeechRecognitionAPI. Some browsers do not support this API; Chrome is recommended. - Privacy and offline use: The plugin does not save audio recordings. Some browsers may send audio to their speech recognition service, so offline processing cannot be guaranteed.
- Source review: It is recommended to review the source code before installation; the plugin license is MIT.
