dsh-plugin-voice-input

Seamlessly integrate voice input into the DeepSeek Harness Web UI without leaving your familiar writing environment. This plugin addresses the pain points of web-based speech interaction by allowing you to record audio directly via a microphone button next to the input field, significantly boosting typing efficiency. Its core strength lies in completely "local offline" processing. Utilizing the sherpa-onnx and SenseVoice models, the plugin performs speech-to-text (ASR) entirely on your device. This eliminates the need for cloud uploads, protecting your privacy while avoiding network latency. It supports mixed Chinese/English and multiple other languages; the transcribed text is automatically appended to the draft and never auto-sends, giving you full control. Designed for users who value efficiency and privacy, particularly developers and power users of DeepSeek Harness. The plugin adheres to official design standards, placing a microphone button to the left of the send key without disrupting the layout. With a simple installation, you can enjoy a smooth offline voice input experience directly in your browser.