zemanzhang809/dsh-stt-plugin ↗★ 1
dsh-stt-plugin
Speech-to-text (voice input) plugin for DeepSeek Harness — a microphone toggle in the composer tool row and a Voice Input settings page. 适合需要通过麦克风进行连续听写或单次语音输入转文字的用户。
インストール
$
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:zemanzhang809/dsh-stt-pluginドキュメント
README 全文を読む ↗Usage
- Open a session so the composer is live.
- Click the mic button and speak. Recognized text is appended to the draft (a space joins the draft and each phrase; interim results are rewritten live).
- Click the mic again to stop. With auto-send enabled in single-shot mode, the message submits automatically when a phrase finishes; otherwise send as usual.
Settings reference
Open Settings → Voice Input:
| Setting | Default | Meaning |
|---|---|---|
| Recognition language | auto | auto follows the browser locale, or pick a BCP-47 code (zh-CN, en-US, ja-JP, de-DE, fr-FR, es-ES, ru-RU, …). |
| Continuous dictation | off | Keep the microphone open until you click the mic button again; phrases accumulate in the draft. |
| Auto-send after recognition | off | Single-shot mode only: submit the message as soon as a phrase finishes. |
Preferences persist in the DSH settings store and survive refresh and restart. When that store is not writable in your composition (settings service absent, or the page served from a non-loopback address), the page transparently falls back to this browser's local storage — the controls stay enabled either way, and a hint states where values are kept.