3274375092/dsh-voice
30Last commit Aug 15, 2026
dsh-voice DSH plugin
dsh-voice adds a microphone button to the DSH web UI composer and a configurable global hotkey (default Ctrl+Space). When you speak, live streaming partial transcripts are echoed, and the final text is committed as a normal chat message without touching the agent's persona. It uses an adaptive dual engine: host-native ASR (sherpa-onnx + silero VAD, offline/private) with automatic fallback to browser Web Speech, so zero-config users get recognition immediately.
How to install the dsh-voice DSH plugin
dsh plugin --profile web add @nn12138/dsh-voiceCopying does not run this command. Review the repository and version before installing the dsh-voice DSH plugin.
dsh-voice DSH plugin data source
dsh-voice DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-voice DSH plugin can do
- Voice input via microphone button or configurable hotkey
- Live streaming recognition with partial transcripts echoed while speaking
- Adaptive dual engine: host-native ASR (sherpa-onnx + silero VAD) with automatic fallback to browser Web Speech
- Preset-agnostic: does not modify the agent persona or system prompt
- Optional offline native recognition models (download via dsh-voice-models)
Where the dsh-voice DSH plugin fits
- Hands-free chat input when typing is inconvenient
- Quickly compose messages using voice while keeping keyboard focus
- Use with any preset (code, standard, minimal, custom) without affecting behavior
- Offline voice recognition for privacy-sensitive environments (with native model)
Who the dsh-voice DSH plugin is for
- DSH users who prefer voice input over typing
- Users with accessibility needs or temporary mobility issues
- Power users who want a hotkey-driven voice workflow
dsh-voice DSH plugin limitations
- Input only: does not handle agent responses or persona modifications
- Native ASR requires additional installation of sherpa-onnx-node and model download (~100 MB)
- Browser Web Speech engine depends on browser support and may be unavailable in some environments
- Hotkey Ctrl+Space may conflict with other system shortcuts; customizable in config
- Microphone permission required; denied permission will prevent voice input
dsh-voice DSH plugin: from the repository README
Quoted from the 3274375092/dsh-voice README, the upstream source of the dsh-voice DSH plugin. Copyright remains with the original authors.
**English** | [中文](README.zh.md) A voice input plugin for DeepSeek Harness: click 🎤 in the web UI (or press a hotkey), speak, and the recognized text is submitted as a normal chat message. **Input only** — it never touches the agent preset/persona, so it behaves like "another input method" in every mode. [](https://www.npmjs.com/package/@nn12138/dsh-voice) [](LICENSE) ## Features - 🎤 **Voice input**: microphone button (platform design-system UI) + configurable global hotkey (Ctrl+Space by default) - ⚡ **Live recognition**: streaming partial transcripts are echoed while you speak; VAD finalization commits on stop (0.6s tail padding keeps sentence endings) - 🧠 **Adaptive dual engine**: host-native ASR (sherpa-onnx-node zipformer2 + silero VAD, offline/private) with automatic fallback to browser Web Speech (zero extra dependencies) - 🔌 **Preset-agnostic**: does not touch the persona/system prompt; works with code/standard/minimal/custom presets - 📦 **Optional models**: zero-config out of the box; run `dsh-voice-models` when you want native offline recognition
Read the full READMERepository license: MIT
dsh-voice DSH plugin questions
How do I install dsh-voice?
Run `dsh plugin --profile web add @nn12138/dsh-voice` in your terminal. For offline native recognition, also install the runtime with `dsh plugin --profile web add sherpa-onnx-node` and run `dsh-voice-models` to download the model (~100 MB). After installation, restart `dsh web` to see the microphone button.
Can I change the hotkey from Ctrl+Space?
Yes. Edit the plugin config in `~/.dsh/profiles/web/cordis.patch.yml` under the `voice` section, change `hotkey` to any key combination (e.g., `ctrl+alt+v`). The new hotkey takes effect after restarting `dsh web`.
Does dsh-voice work offline?
Yes, if you install the native ASR engine (sherpa-onnx-node) and download the model. With the model present, the plugin uses local offline recognition. Without it, it falls back to browser Web Speech which requires an internet connection for the speech service.
How do I switch between native ASR and browser Web Speech?
Set the `engine` option in the plugin config to `native`, `browser`, or `auto` (default). `auto` will use native if the model is available, otherwise fall back to browser. After changing the config, restart `dsh web`.
Will dsh-voice change the AI's personality or system prompt?
No. The plugin only sends the recognized text as a normal chat message. It never touches the agent preset, persona, or system prompt, so it works identically in all modes (code, standard, minimal, custom).