Full-duplex voice for the DSH Web UI: tap-to-toggle or hold-to-talk dictation (send key or Ctrl) with a live caption, host-side SenseVoice ASR via sherpa-onnx, sentence-by-sentence spoken replies, and true barge-in where speaking interrupts playback and the running turn. Includes NLMS acoustic echo cancellation using the page's own TTS playback as the echo reference. No API key.
dsh plugin --profile web add @haoku123/dsh-voice