dsh-speech-input

When you are writing long prompts, jotting down ideas, or tidying code comments in the Harness Web composer, pure keyboard input can feel like the last mile never quite closes. dsh-speech-input drops a native microphone button into the input slot: click once to start dictating, watch recognized text land in the current draft in real time, click again or press Enter to stop and have end-of-sentence punctuation added. The plugin only writes to the draft and never sends the message for you.

It rides on the browser-native Web Speech API, works out of the box in Chrome and Edge, and follows the system language for both Chinese and English. While listening you see an oval pill with a cancel button, a live volume meter driven by locally computed RMS, and a stop button. Manual edits made mid-dictation are preserved; five seconds of detected silence triggers an auto-stop that keeps everything already recognized; canceling removes only the words produced by that round of speech. No API key, no server, no model download is required, and permission or network errors surface with a clear notice.

If you use Chrome or Edge in Harness Web to craft prompts, capture thoughts, or annotate code and want to speak while still staying in control, this plugin is built for you: zero-config, entirely in-browser, and the plugin itself neither stores nor uploads audio. Open it and start talking.