PiyotaHu/muxiva-dsh-voice

Muxiva Voice adds local-first, full-duplex voice to DeepSeek Harness: the microphone, VAD, ASR, sentence scheduling, TTS and barge-in path run on the Mac through the Muxiva graph, while DSH keeps its tools, sessions and Web transcript, connected by a versioned loopback bridge that patches neither upstream project.

Muxiva Voice adds local-first, full-duplex voice to DeepSeek Harness. Your microphone, VAD, ASR, sentence scheduling, TTS, playback and interruption path stay on the Mac, orchestrated by the Muxiva graph; DSH remains the stateful agent harness and keeps its tools, sessions and Web transcript, and the plugin owns only the bridge - additive DSH UI, a versioned loopback protocol and project Node Packs, patching neither upstream project. The pipeline runs PCM16 into a typed-frame graph (Silero VAD for barge-in, Zipformer preview, SenseVoice zh/en final) and streams assistant deltas back through a sentence buffer, speech formatter and normalizer into Qwen3-TTS/MLX playback. A large voice orb above the composer follows input energy and moves through listening, hearing, thinking and speaking; mute pauses the audio source without stopping the WebSocket, and a small End control performs a full shutdown. It was certified on an M1 Pro at 130/130 turns and 30/30 interruptions with zero TTS underruns.

Vision & Multimodal ★ 2
View on GitHub ↗

Install

dsh plugin --profile web add @muxiva/dsh-voice@alpha

Copied verbatim from the README's alpha install block; the README also documents a source install dsh plugin --profile web add . from a checkout.

Compatibility

Alpha: requires macOS arm64, Node.js 22.19+, Python 3.11-3.13, the Muxiva 0.1.1 CLI and the official DSH CLI. DSH rc.5 and rc.6 are supported; alpha.2 is certified against rc.6. The speech bridge binds only to loopback and accepts one active client; model artifacts are pinned by immutable revision/URL with SHA-256 verification. English README.

Details

Alternatives

54xkeee/dsh-vision · 54xkeee/dsh-youreyes · Cheng-cheng9669/dsh-deepseek-vision

More plugins in Vision & Multimodal

Browse more in Vision & Multimodal

Guides for Vision & Multimodal plugins