Speech-to-text voice input for the DeepSeek Harness (DSH) web GUI.
- mic button in the composer tool row: click to record, click again to transcribe into the input box
- engines: browser Web Speech API (zero-config, Chrome/Edge) or OpenAI-compatible Whisper API (OpenAI / Groq)
- selectable model (whisper-1, whisper-large-v3, whisper-large-v3-turbo, distil-whisper-large-v3-en, custom), language and insert mode in Settings -> 语音输入
- API key kept in page memory only; non-secret config persisted via localStorage
Install: dsh plugin --profile web add dsh-stt-input (from this tarball)