Перейти к основному содержимому
V

dsh-plugin-voice-alert

venompool888/dsh-plugin-voice-alert

Spoken alerts on macOS that name the source and project: announces approval requests automatically and adds a voice-alert skill the agent uses for verification codes, finished long tasks or when it is stuck. DeepSeek TTS via an API key (DeepSeek official, a third-party base URL, or the key already set in DSH Models), tested before saving and falling back to the macOS system voice; speaks only after 30 seconds without keyboard or mouse input and repeats until you return, with AI-written lines in a chosen tone (professional, tsundere, cute, clingy, cold or custom); a Settings → Voice page with instant previews, speech speed, 10 spoken languages and an away mode.

Установка

dsh plugin --profile web add github:venompool888/dsh-plugin-voice-alert

README

dsh-plugin-voice-alert

中文 · English

中文

给 DeepSeek Harness(DSH)用的语音提醒插件。你不在屏幕前时,会话需要你批准、要验证码,或者长任务跑完了,它会把话念出来,开头先报来源和项目:

来自 DSH,3200 项目需要你批准。 来自 DSH,官网项目需要验证码:四,八,二,一。

安装

  1. 打开 DSH → 设置 → 插件 → 添加插件。
  2. 在输入框里填这个地址,点「安装」: https://github.com/Venompool888/dsh-plugin-voice-alert (需要本机能访问 GitHub;访问不了的话,到 Releases 下载 .tgz,解压后把文件夹路径填进去。)
  3. 按 ⌘Q 完全退出 DSH,再重新打开。设置侧栏里会出现「语音」。

也可以用命令行安装(桌面版用 desktop profile,网页版用 web):

dsh plugin --profile desktop add github:Venompool888/dsh-plugin-voice-alert

装好就能出声:没有可用的云端凭据时自动用 macOS 本地语音。打开「设置 → 语音」,在「试一试」里输入一句话点「朗读」,能听到声音就说明装好了。

功能

  • 等你批准时自动出声:念「来自 DSH,某某项目需要你批准」。插件只负责念,不会替你批准。
  • AI 自己判断要不要喊:插件注册了一个 voice-alert skill,遇到验证码、长任务结束、卡住需要你拍板时,AI 会调用它。
  • 两种声音:
    • DeepSeek 云端语音(默认),在「语音 → DeepSeek 语音」里选服务商:
      • DeepSeek 官方:只需要一个 API key(在 platform.deepseek.com 创建)。如果你已经在 DSH「设置 → 模型」里配置了 DeepSeek key,会自动用它,什么都不用填;页面上会显示正在用哪个 key。
      • 第三方:兼容 DeepSeek 语音接口的服务,填它的 base URL 和 key。请求发到 <base URL>/v1/tools/tts(base URL 以 /v1 结尾时是 <base URL>/tools/tts)。
      • 注意:DSH 的账号登录不能用于语音合成,DeepSeek 的语音接口只认 API key。
    • macOS 本地语音:系统自带,不联网、不花钱。云端失败时自动改用本地语音。
  • 10 种播报语言:中文、English、日本語、한국어、Français、Deutsch、Español、Português、Italiano、Русский。
  • 离开检测:键盘鼠标 30 秒没动才开口,你在电脑前时只响提示音、弹通知;之后每 30 秒喊一次(5 次后改为每 2 分钟),动一下键盘或鼠标就停,最多 20 分钟。DSH 里的批准一处理掉也会马上停。读的是 macOS 自带的空闲时间,不需要任何权限。
  • AI 现写、可选语气:每次喊的话由 AI 用你的 DeepSeek key 现写,可选专业、傲娇、可爱、撒娇、冷漠或自定义语气;验证码数字照常逐位念,生成失败时用固定句子。
  • 待命模式:跟 AI 说「我去吃饭了」「帮我盯着」,长任务结束也会喊;说「我回来了」就只在需要你动手时喊。
  • 可调:提示音、通知横幅、播报时自动调大音量、同一句话多少秒内不重复。

要求

  • macOS(播放用系统自带的 afplay / say)。Windows、Linux 上设置页能打开,但不会出声。
  • DSH 桌面版。

隐私

  • 所有设置都在「设置 → 语音」页里完成,不需要改任何文件。
  • API key 只保存在本机、只有你自己能读,只发给你选的那个服务商;页面上只显示末 4 位。
  • 插件只读取 DSH「模型」里的 DeepSeek key,不会修改它。

升级

DSH 规定第三方插件要先卸载再安装新版。设置和 key 不受影响。

命令行

插件会把播报脚本装到 ~/.dsh/bin/voice-alert,其他工具也能调用:

~/.dsh/bin/voice-alert --project 官网 --code 4821
~/.dsh/bin/voice-alert --project 官网 --style done "测试全部通过"
~/.dsh/bin/voice-alert doctor

English

A voice-alert plugin for DeepSeek Harness (DSH). When you're away from the screen and a session needs your approval, needs a verification code, or finishes a long task, it says so out loud, starting with the source and the project:

From DSH, project 3200 needs your approval.

Install

  1. Open DSH → Settings → Plugins → Add plugin.
  2. Enter this address and click Install: https://github.com/Venompool888/dsh-plugin-voice-alert (If GitHub isn't reachable, download the .tgz from Releases, unpack it, and enter the folder path instead.)
  3. Quit DSH completely (⌘Q) and reopen it. A Voice page appears in the Settings sidebar.

Or install from the command line (desktop profile for the desktop app, web for the web UI):

dsh plugin --profile desktop add github:Venompool888/dsh-plugin-voice-alert

It speaks right away: with no cloud credential it uses the macOS local voice. In Settings → Voice, type a sentence under Try it and click Read aloud to check.

Features

  • Speaks when approval is needed. It only announces; it never approves anything for you.
  • A voice-alert skill, so the AI can speak up for verification codes, finished long tasks, or when it's stuck.
  • Two voices: DeepSeek cloud voice (default — pick a provider: DeepSeek official needs only an API key from platform.deepseek.com, and reuses the DeepSeek key already set in DSH Settings → Models if there is one; Third-party takes the base URL and key of a DeepSeek-compatible service, called at <base URL>/v1/tools/tts. Signing in to DSH does not cover speech: DeepSeek's speech endpoint only accepts API keys) and the macOS local voice (offline, free, used automatically if the cloud fails).
  • Idle detection: speaks only after 30 seconds without keyboard or mouse input (otherwise just a chime and a notification), repeats every 30 seconds (every 2 minutes after 5 calls), and stops the moment you touch the keyboard or mouse, after 20 minutes at most, or when the DSH approval is handled. Uses macOS's own idle time; no permissions needed.
  • AI-written lines in a chosen tone: professional, tsundere, cute, clingy, cold or a custom persona, written with your DeepSeek key; verification code digits are still read one by one, and a fixed sentence is used if generation fails.
  • 10 spoken languages. Away mode for announcing finished tasks while you're gone.

Requirements

macOS for playback. On Windows and Linux the settings page opens, but nothing is spoken. DSH desktop.

Privacy

Everything is configured on the Settings → Voice page; no files to edit. API keys stay on your Mac, readable only by you, and are sent only to the provider you chose; the page shows just the last 4 characters. The DeepSeek key from DSH Models is read, never changed.

License

MIT

Похожие плагины