haoku123/dsh-voice

Name: haoku123/dsh-voice
Author: haoku123

Plugin插件 Native原生 ⭐ 4 MIT Vision & Media视觉与多媒体

DeepSeek Harness 全双工语音插件：麦克风 → SenseVoice 语音识别（sherpa-onnx）→ 大语言模型 → 支持真正语音打断的 Edge TTS。全双工语音插件：麦克风语音输入，SenseVoice 本地识别（简体中文+标点+ITN），流式 TTS 朗读，支持语音打断。零 API 密钥。

Project Overview项目介绍

Edited by Zhiwei Agent (zhiwei)由知微Agent(zhiwei)编辑

dsh-voice is a native full-duplex voice plugin built exclusively for DeepSeek Harness that enables end-to-end voice interaction, from streamed automatic speech recognition to LLM response and text-to-speech output. It supports true barge-in, letting users interrupt the assistant mid-response at any time. To install the plugin, run the dsh plugin --profile web add <repo-url-or-path> command from your terminal, then start your DSH host as you normally would. The plugin requires Node.js version 22.19 or newer, or Node 24 or newer, to access the required native zstd compression APIs.

The plugin supports three distinct voice input gestures to match different user preferences: tap-to-toggle for continuous dictation, hold-to-record for controlled input, and a configurable keyboard hotkey for hands-on-keyboard use. While holding the record key, it displays a live incremental caption preview of your ongoing speech, so you can see what is being transcribed before you submit. Barge-in uses an NLMS acoustic echo canceller to filter out assistant playback from your microphone input, so it works reliably even when the assistant is speaking loudly. This plugin is ideal for users who prefer voice-based interaction with their DSH assistant.

ASR runs locally on your host machine via sherpa-onnx with the SenseVoice model, which performs particularly well for Chinese speech with built-in punctuation and inverse text normalization. Model files are downloaded on first use through a cache-through proxy that mirrors files to your local disk, and downloads resume automatically if interrupted. You can run npm run prefetch while your DSH host is running to warm the model cache ahead of first use. The plugin is released under the permissive MIT open source license, so you can modify or redistribute it freely per the license terms.

这是专为 DeepSeek Harness 开发的原生全双工语音插件，提供从语音自动识别到大语言模型回复再到文本转语音的端到端流式全双工语音交互，原生支持用户随时打断当前回复的barge-in功能。它依托sherpa-onnx运行SenseVoice多语言语音识别模型，使用微软Edge TTS生成回复语音，可通过DSH官方插件命令直接安装，接入后即可在DSH网页界面中使用语音交互。

用户可通过三种方式输入语音：点击麦克风开启连续听写、按住发送键或麦克风录制到松开后提交，也可按住Ctrl等自定义热键，不离开键盘即可完成输入。输入过程中界面会显示实时字幕预览当前转写结果，用户说话触发的打断会直接停止当前回复的播放与生成，立即处理新的语音输入。该插件适合偏好语音交互、需要快速和DSH对话的中文用户日常使用。

该插件要求Node.js版本不低于22.19或24，依赖Node原生zstd API才能正常运行。模型文件会在第一次使用时自动下载，缓存到本地指定目录，支持断点续传，用户也可运行npm run prefetch命令提前预热缓存，还可配置HF镜像加速国内下载。插件采用MIT许可证开源，所有依赖模型均遵循对应开源协议，可免费使用。

View on GitHub在 GitHub 查看

Send to your DSH发给你的 DSH 安装

请帮我安装这个 DSH 插件。安装前先完成【兼容性检查 + 安全性检查】，检查通过再动手。

插件：dsh-voice（haoku123/dsh-voice）
仓库：https://github.com/haoku123/dsh-voice
本站详情页：https://www.yhbd.top/plugins/haoku123-dsh-voice/
本站登记：类型 plugin · 归类 原生 DSH 插件 · 许可证 MIT · ⭐ 4 · 最近提交 2026-08-23 · 主语言 TypeScript

按下面顺序执行，每步先把结论告诉我，再进入下一步：

【1 兼容性检查】
① 我这边：DSH 版本、Node 版本、操作系统、当前 profile（web / desktop）。
② 读它的 README、package.json、插件 manifest，列出它要求的 DSH 版本 / Node 版本 / 操作系统 / 外部依赖 / 需要另外先装的运行时。
③ 逐条比对，结论只写「满足 / 不满足 / 未知」三种；不满足的给出可行替代方案。
④ 检查是否和我已装的插件冲突：命令名重复、skill / tool 重名、端口占用、重复注册的 MCP server。

【2 安全性检查】
① 仓库可信度：和上面「本站登记」是否一致；star / fork 数、创建时间、最近提交，是否归档或长期停更。
② 安装脚本：逐行看 package.json 的 preinstall / install / postinstall，以及 install.sh、setup.ps1 之类脚本。出现 curl|bash、下载后直接执行、混淆代码、访问与插件功能无关的域名，立刻停下来告诉我，不要继续装。
③ 依赖：列出新增依赖，标出无人维护、或与知名包拼写近似的可疑包（typosquatting）。
④ 权限与副作用：它会读写哪些目录、访问哪些域名、需要哪些 DSH 权限（filesystem / network / shell / clipboard 等），以及怎么卸载和回滚。
⑤ 如果它要求 sudo / 管理员权限，或权限明显超出功能所需，先停下来问我。

【3 安装】
上面两步没有「不满足」和「高危项」时才执行；用官方推荐方式安装，不要自行提权。

【4 汇报】
用表格输出：检查项 / 结论 / 依据 / 是否需要我决策。拿不准的一律写「未知」并说明要我怎么确认——不要猜，也不要替我决定。

Send this message to DSH in your current session: it verifies compatibility and security first (answering met / not met / unknown item by item) and only installs once everything checks out — it will stop and ask you if it finds a high-risk item. The box scrolls; the copy is the full prompt. CLI install commands may not be accurate across systems, so DSH is the safer route.把上面这条消息直接发给当前会话里的 DSH：它会先核对兼容性与安全性（逐条给「满足 / 不满足 / 未知」），确认没问题再安装，有高危项会停下来问你。框内可滚动，复制到的是完整提示词；安装命令不一定准确，发给 DSH 更稳。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意

Only 4 stars - very few users, little community feedback星标只有 4，几乎没人在用，遇到问题缺少社区反馈

DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时（Electron / Python / Docker 等）是否齐备
Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突：命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致，是否归档或长期停更
Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需（filesystem / network / shell / clipboard）
Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限，以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照，不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装（适合开发者）

CLI Install命令行安装

dsh plugin --profile web add github:haoku123/dsh-voice

把 haoku123/dsh-voice 加入你的 DSH 配置（web profile）即可启用。

READMEREADME

6.8 KB · 缓存于 2026-08-23 17:28:24 · 6.8 KB

dsh-voice

Full-duplex voice mode for DeepSeek Harness: streamed ASR → LLM → TTS with barge-in.

Status

v0.7.0 — press-to-talk with a live caption.

Speak into the composer mic: the assistant silences itself (playback stops, host synthesis queue drops), the running turn is cancelled (the stop-button route), and your speech is transcribed by the host (SenseVoice, CN-native simplified-Chinese ASR with punctuation + ITN) and submitted. The reply streams back as spoken audio with live captions.

Three ways to dictate:

gesture	behaviour
tap the mic	continuous dictation; the VAD segments on trailing silence
hold the send key (or the mic)	records until release, slide up to discard
hold `Ctrl` (configurable `asr.hotkey`)	same, without leaving the keyboard; `Esc` discards

While a hold is open the overlay shows a live caption — the interim transcript of what has been said so far — and keeps a spinner up after release until the authoritative transcript lands.

Barge-in detection is triggered by the mic's leading speech edge. The ASR engine runs an NLMS acoustic echo canceller (see src/aec.ts) using the page's own TTS playback as the echo reference, so loud assistant audio is subtracted from the mic before the VAD — the browser-level echoCancellation constraint is kept only as a fallback when no echo reference is available.

Live caption interims are incremental: each pass sends only the audio recorded since the last pass (correlated by session header), and the host decodes a bounded sliding window per session instead of re-decoding the whole hold. Preview cost therefore stops growing with hold length.

Demo

dsh-voice demo

The loop: hold the composer's send key (its arrow is covered by a mic glyph), watch the live caption fill in while you speak, release into a spinner until the final transcript lands, then the reply streams back as spoken audio sentence-by-sentence — until the user's voice interrupts playback and stops the running turn mid-line (true barge-in). Ctrl does the same without leaving the keyboard.

How it works

input:  mic ──RMS endpoint detection──▶ POST /asr (raw f32 PCM)
                                           │ text (SenseVoice)
                                           ▼
        composer draft ──submit──▶ model stream ──llm/stream tap──▶ SentenceSegmenter
                                                                     │
        browser ◀── SSE /dsh-voice-api/stream ── TtsQueue (msedge-tts) ◀──┘
                  (base64 MP3 frames + caption text)

barge-in: speech edge ──▶ engine.skip() + POST /cancel (epoch bump)
                         + session.cancel() when a turn is running

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要，完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

渲染基于仓库 haoku123/dsh-voice 的 README.md。在 GitHub 查看原文 →