dsh-web-speech-input

Typing out a long prompt character by character in the Harness Web composer is not always the most natural workflow. dsh-web-speech-input adds a microphone button to the left of the input field that leverages the browser native Web Speech API, turning live speech into editable draft text with interim results visible as you talk. You review, edit, and submit on your own terms; the plugin never clicks Send for you.

There is no backend, no audio storage, and no network request from the plugin itself; all recognition is delegated to the browser speech engine. Language detection follows the page language, then the browser language, with zh-CN as the fallback. Common edge cases such as denied permissions, unavailable microphones, or network errors are surfaced with clear status messages, and the plugin includes keyboard focus management, ARIA labels, and reduced-motion support.

Ideal for Harness Web users on Chromium-based browsers (Chrome, Edge, etc.) who frequently compose long prompts or prefer thinking out loud. Firefox does not yet expose a compatible API in stable releases, and corporate or restricted environments may further limit availability; when it cannot work, the plugin tells you clearly rather than failing silently.