dsh-speech-input is an independent implementation for DeepSeek Harness.

Prior-art review included the MIT-licensed dsh-mic-input project by QT-Chen
(https://github.com/QT-Chen/dsh-mic-input). This project was implemented from
new source against the public Harness slot and input-action contracts; the
acknowledgement is retained because the prior project informed product-level
behavior such as a composer microphone and continuous browser recognition.