aaaadrop/dsh-voice-kit

README 摘要

dsh-voice-kit 🎙️ Voice input (Web Speech API) and read-aloud (speechSynthesis) for the DeepSeek Harness web GUI. 中文说明 Status: v0.3.1, published on npm — pnpm typecheck passes, 35 unit tests pass, pnpm build emits the ecosystem-standard closure-factory bundle (host half + browser half). Voice input and read-aloud are wired against the real conversation API (session kit's inputActions / useInput / useSession ), and read-aloud supports Microsoft Edge neural voices (host-side synthesis via msedge-tts , cached, with automatic fallback to the system voice). Verified inside a real DSH Desktop profile. Install: dsh plugin add dsh-voice-kit . Features - 🎤 Voice input — mic button in the composer's left rail; speech is transcribed (Chrome/Edge webkitSpeechRecognition ) and appended to the draft. Auto-restarts across Chrome's 60s session limit; Esc cancels. - 🔊 Read aloud with neural voices — per-message button at each assistant message tail. Default engine is Microsoft Edge neural TTS (晓晓/云希/ 云健/云扬…), synthesized host-side and played back with an element; falls back to speechSynthesis when offline or on failure. Markdown and emoji are stripped before speaking; long replies are chunked at …

分类