dsh-voice
Full-duplex DeepSeek Harness voice plugin with local SenseVoice ASR, streaming Edge TTS, captions, and voice barge-in.
Install
$ dsh plugin --profile web add <repo-url-or-path>Plugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Records microphone audio and sends raw PCM to host-side SenseVoice ASR.
- Provides simplified-Chinese-oriented ASR with punctuation, inverse text normalization, and configurable automatic language selection.
- Streams model responses into sentence-segmented Edge TTS audio with live captions.
- Supports barge-in by clearing local playback, dropping queued or in-flight synthesis, and cancelling a running turn.
- Exposes SSE, ASR, cancellation, configuration, and Hugging Face model-proxy routes.
- Caches downloaded ASR model files on disk and supports cache warming with npm run prefetch.
- Offers toggle and hold microphone modes plus optional automatic submission.
Useful For
- Hands-free voice conversations with DeepSeek Harness.
- Chinese speech input with local host-side transcription.
- Interrupting a spoken response to ask a new question.
- Using a Hugging Face-compatible mirror for ASR model downloads.
Who It Fits
- DeepSeek Harness web-profile users who want microphone input and spoken responses.
- Chinese-speaking users seeking SenseVoice transcription features.
- Developers configuring host-side speech models and local model caches.
Documented Limitations
- Requires Node.js 22.19+ or 24+ because it uses node:zlib zstd APIs.
- ASR model files are fetched on first use unless the cache is prewarmed.
- Barge-in relies on browser echo cancellation; loud TTS may leak into the microphone on some platforms because there is no JavaScript-level AEC.
- The documented installation targets the DeepSeek Harness web profile.
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 2 runtime, 8 development, 3 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as MIT.
A root package.json was captured and can be inspected by the registry.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- haoku123/dsh-voice
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- e1ed20a · 2026-09-13
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 4GitHub stars
- Forks
- 6
- Open issues
- 0
- Last commit
- 2026-08-23
- Last release
- 2026-08-17
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Codex-style @file mentions and workspace path picker for the DeepSeek Harness web interface.
View plugin →VSCode-like modular sidebar and bottom panel workbench for DeepSeek Harness with terminal, editor, git, and plugin extension API.
View plugin →Full context lifecycle management, visualization dashboard, and /context command for DeepSeek Harness.
View plugin →Wallpaper Engine background integration with liquid glass UI styling for DSH web.
View plugin →