Back to directory

dsh-duet

In a curated listMaintenance: Active

yums/dsh-duet

Adds full-duplex voice control and speech playback for DSH sessions, including composer edits, send, switch, fork, rename, cancel, and question/approval replies.

View on GitHub
$ npm install -g @deepseek-ai/dsh

16

stars

2

forks

JavaScript

Language

Apache-2.0

License

2026-09-18

Created

2026-11-20

Last push

Apache-2.0 DSH plugin that adds full-duplex Chinese voice control and speech playback to DSH Web and Desktop sessions.

DSH integration

Native runtime

Author-claimed

Safety audit

Unaudited

Last verified

2026-11-15

License

Apache-2.0

01What can it help you accomplish?

  • Operate a DSH session by voice instead of the keyboard

    Spoken composer input, edits and clearing, send and stop commands, and new / switch / rename / fork / archive session actions, plus spoken answers to DSH questions and free chat

    DSH Web and Desktop users who want to run sessions while away from the keyboard or while interrupting an in-flight reply

  • Hear task results without watching the DSH screen

    A short spoken announcement when a task finishes (playback mode), or optionally a 3D companion that reacts to your voice and shows pending items and task notifications

    Users running long DSH tasks who leave the screen and only need the result or a prompt to step in

02How to install into DeepSeek Harness

Prerequisites

  • DSH Web or DSH Desktop — the README supports both
  • Web profile only: Node.js 22 LTS from 22.19.0, or Node.js 24 and above
  • A workspace created or selected in DSH before first use

Installation steps

  1. 01

    Install and open DeepSeek Harness: `npm install -g @deepseek-ai/dsh`, then `dsh web`

    $ npm install -g @deepseek-ai/dsh

  2. 02

    In another terminal, add the plugin to the Web profile: `dsh plugin --profile web add dsh-duet`, then restart with `dsh web`

    $ dsh plugin --profile web add dsh-duet

  3. 03

    Desktop: open the plugin manager in DSH Desktop, choose 添加插件 (add plugin), enter `dsh-duet` and install

  4. 04

    Update: rerun `dsh plugin --profile web add dsh-duet@latest` and restart DSH on Web; update from the plugin manager on Desktop

    $ dsh plugin --profile web add dsh-duet@latest

Verify the integration

  • After installing on Desktop, if the duet button does not appear, quit and reopen DSH Desktop
  • On first voice use, follow the prompt back on the DSH page and click 启用声音 (enable sound)

03DSH integration and capability boundaries

DSH integrationNative runtime

DSH Web / Desktop plugin: installed with `dsh plugin --profile web add dsh-duet` on the Web profile or through the DSH Desktop plugin manager, and it adds duet voice buttons to the DSH session UI

  • Playback mode

    a finished DSH task→a short spoken announcement of the result

  • Full-duplex interaction mode

    microphone audio while interaction mode is on→composer edits, DSH commands and spoken replies, with barge-in while the model is replying

    microphone audio captured during interaction mode is uploaded to the duet voice server
  • Voice session and task commands

    spoken commands such as “发送”“提交” for sending or “停止当前任务” for stopping→composer input, edits and clearing, task send / stop, session create / switch / rename / fork / archive, answers to DSH questions, and assistant renaming

  • 3D companion avatar (experimental)

    voice and session events→a desktop or in-page character with subtitles that follow the voice, nodding and lip movement, pending-item list and task notifications

    the avatar is toggled from the avatar button and right-clicking the desktop character resizes it, switches bubble style, shows pending items or returns to DSH

04Who is it for? When not to use it?

Good for

  • DSH Web and Desktop users who want to run sessions while away from the keyboard or while interrupting an in-flight reply
  • Users running long DSH tasks who leave the screen and only need the result or a prompt to step in

Not for

  • The voice service only supports Chinese for now — the README states Chinese-only support, so English or other-language speech is not covered by the spoken commands or playback.
  • duet's voice service runs on a server. While interaction mode is on, the interaction audio captured by the microphone, session info (session id, name, running state, current session name), composer content and task results are uploaded to it; the README warns not to say passwords or keys.

05Compatibility, maintenance and safety notes

  • The voice service only supports Chinese for now — the README states Chinese-only support, so English or other-language speech is not covered by the spoken commands or playback.
  • duet's voice service runs on a server. While interaction mode is on, the interaction audio captured by the microphone, session info (session id, name, running state, current session name), composer content and task results are uploaded to it; the README warns not to say passwords or keys.
  • The 3D desktop avatar ships for macOS Apple Silicon and Windows x64 only; other platforms have to use the in-page avatar instead.
2026-09-182026-11-11v0.2.1

Apache-2.0 · latest release v0.2.1 (2026-11-11)

06Frequently asked questions

How do I install duet in DSH?

Web: run `dsh plugin --profile web add dsh-duet` in another terminal, then restart with `dsh web`. Desktop: open the plugin manager, choose 添加插件, enter `dsh-duet` and install — if the duet button does not appear, quit and reopen DSH Desktop.

Which Node.js version does the Web profile need?

Node.js 22 LTS from 22.19.0, or Node.js 24 and above. The README states this requirement only for the Web profile.

What can I actually do by voice?

Dictate, edit or clear the composer, send a task with “发送”“提交”, stop the current task, create / switch / rename / fork / archive sessions, answer DSH questions, rename the assistant, and chat freely. Playback mode only announces task results.

What data leaves my machine?

duet's voice service is served from a server. While interaction mode is on, the microphone's interaction audio, session info (id, name, running state, current session name), composer content and task results are uploaded to understand speech, run session actions, generate replies and speech. The README says not to hand over passwords or keys.

Does duet work in English?

No — the README states Chinese-only support for now. Both the spoken commands and the playback are built around Chinese speech.

08Data and sources

  • Author-claimedgithub.comcd2b0842ff97…

    在另一个终端为 Web profile 安装插件,再重启 DSH

  • Author-claimedgithub.comcd2b0842ff97…

    在 DSH Desktop 的插件管理页面选择“添加插件”,输入 `dsh-duet` 并安装。

  • Author-claimedgithub.comcd2b0842ff97…

    支持 DSH Web 和 Desktop。

This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-11-15. Found an error? Submit a correction.

🏆

Best DeepSeek Harness Plugins

Twelve plugins worth installing first — picked from the whole catalog, across every category.

DSH Plugins is an independent community directory of DeepSeek Harness plugins. Not affiliated with or endorsed by DeepSeek. Third-party plugins are not security-audited — review the source before installing.

New DeepSeek Harness plugins, weekly. No spam.