v1.2.0

中文

dsh-sound-lab 1.2.0 —— DSH 声音工坊（Sound Lab）

DeepSeek Harness（桌面版 / web 版）的声音工坊：对话事件音效、AI 角色语音生成、音效库管理。

功能总览

• 四个触发条件，独立配置：会话结束 / 弹出选项 / 请求许可 / 停止，每个可单独【启用 + 选择音效】
• 三种音效选择：内置提示音（Web Audio 琶音，零依赖）/ 不播放 / 音效库具体音效
• 注意类事件始终响铃：弹出选项 / 请求许可出现即播，不受会话运行/查看状态限制
• AI 生成角色音频：音效库上方入口，用你自己的阿里云百炼 API Key 与复刻音色ID，输入文本一键生成专属音色语音（可自定义文件名 / 试听 / 删除，一键加入音效库自动登记），弹窗内附图文详细教程
• 可拖动悬浮球：拖到屏幕边缘自动缩成小半球；配置弹窗可拖动、置顶
• 三种外观：鲸鱼娘（默认）/ 纯白 / 纯黑，语音名字可自定义
• 音量滑杆（0–100%）+ 测试音效：可试听具体音效与内置提示音
• 音效库管理：插件 sounds/ 目录本地音频（sounds.json 控制文件管理，放进目录即自动登记），支持上传（拖拽/选择）、删除、重命名（✎）、恢复被隐藏的附带音效
• 配置双持久化：localStorage + 宿主 config.json，重启不丢；加载时逐字段清洗，脏数据回退默认
• 启动即预加载：首次播放秒出；无音效时内置提示音兜底

本次更新（1.2.0）

• 包名更名：npm 包名由 dsh-client-ui-event-sounds 改为 dsh-sound-lab（与仓库名一致），安装命令、部署目录（$DSH_HOME/plugins/dsh-sound-lab）同步更新
• 新增「AI生成角色音频」：宿主端移植百炼 CosyVoice 合成协议（长文本分段 + WAV 帧拼接）；生成时可自定义文件名（重名 409 拦截），音频可试听 / 删除，一键加入音效库自动登记（不复制文件，直接引用）；token 消耗由用户自负
• 凭据三档状态：未配置（缺任一）/ 未通过（全配置但调用失败）/ 配置成功；防抖合并保存 + 生成请求直接携带 + 关闭弹窗强制落盘，杜绝「填了却提示未配置」
• 详细教程内置：弹窗内「📖 详细教程」图文展示获取 API Key 与音色ID 的完整步骤（可展开，弹窗最高 4:6、不超过页面）
• 音效库重构为 manifest 控制：新增 sounds/sounds.json 控制文件驱动音效库，磁盘文件未登记时自动对账补录（放进目录即可用）；生成目录移到 sounds/generated/，直接链接引用；音效分 ai / local / bundled 三类
• 附带音效升级为 3 个：新增「啊哇哇！」，与「hirari do～」「呢？」同为内置附带音效，不可删除/重命名（删除=软隐藏可随时恢复）
• 音效重命名：非附带音效可在列表中 ✎ 行内重命名（自动补扩展名、重名 409、联动触发条件引用）
• 隐藏音效实时试听：隐藏列表可直接试听且秒出——启动时预加载全部音效（含隐藏）进缓存，试听复用缓存不再二次下载
• UI 微调：「自定义」按钮改名「上传」（上传成功自动关弹窗，恢复不关，隐藏列表实时同步）；音效库超 4 条上下滚动；音效名不显示文件后缀；「不播放」固定置顶
• 工程化：宿主端 TTS 接口 + manifest 对账/重命名/登记冒烟测试扩充（43 项）；教程图片随包发布（lib/tutorial/）

安装（桌面版 / web 版皆可用，任选其一）

• 方式一（推荐，零依赖）：如果 dsh 尚未运行，先启动一次（桌面应用或 npx @deepseek-ai/dsh web）完成初始化；然后双击 dsh-sound-lab-Setup-1.2.0-x64.exe，安装器会自动重启 dsh 即可生效
• 方式二（需 Node.js）：如果 dsh 尚未运行，先启动一次（npx @deepseek-ai/dsh web）完成初始化；然后二选一，根据你的实际运行方式执行其中一条，最后重启 dsh：
  • 桌面版：npx @deepseek-ai/dsh plugin --profile desktop add dsh-sound-lab --config.minimumReleaseAge=0
  • web 版：npx @deepseek-ai/dsh plugin --profile web add dsh-sound-lab --config.minimumReleaseAge=0
• 未安装/未初始化 dsh 时：Setup 安装器与 npm run setup 会自动完成首次初始化（自动下载 Node.js 并运行 npx @deepseek-ai/dsh web）；手动方式则先下载 DeepSeek Harness 桌面版，或安装 Node.js（https://nodejs.org）后运行 npx @deepseek-ai/dsh web。如需安装后自动重启，推荐方式一 Setup 安装器。
• 卸载：双击 dsh-sound-lab-UnSetup-1.2.0-x64.exe，完成后会自动重启 dsh

粉丝向非官方个人项目，音效素材版权归《明日方舟》官方及声优所有，仅供个人学习娱乐，不用于商业用途。

---

English

dsh-sound-lab 1.2.0 — DSH Sound Lab (Angelina「hirari do～」)

A DSH Sound Lab for both DeepSeek Harness Desktop and dsh web: event sounds, AI character voice lines and a sound library

Features

• 4 independently configured triggers: session end / options popup / permission request / stop — each with its own 【enable + sound】
• 3 sound choices per trigger: built-in chime (Web Audio arpeggio, zero dependencies) / no sound / a specific sound from the library
• Attention events always ring: options popup & permission request play as soon as they appear, regardless of the running/viewing state
• AI voice generation: an entry above the sound library — bring your own Alibaba Cloud Bailian API Key and cloned voice ID, type a line and generate speech in your custom voice (custom file name / preview / delete, one-click add to the library with automatic registration), with an illustrated step-by-step tutorial built in
• Draggable floating ball that collapses into a half-ball at screen edges; draggable always-on-top settings dialog
• 3 appearances: Whale Girl (default) / Pure White / Pure Black, plus a customizable voice name
• Volume slider (0–100%) + test sound with preview for both specific sounds and the built-in chime
• Sound library management: local audio in the plugin sounds/ folder (driven by a sounds.json control file — drop a file in and it's registered automatically), custom upload (pick/drag-drop), delete, rename (✎), restore hidden bundled sounds
• Dual-persisted config: localStorage + host config.json; every field is sanitized on load, bad values fall back to defaults
• Preload on startup for instant first play; built-in chime as fallback

What's new in 1.2.0

• Renamed package: npm package renamed from dsh-client-ui-event-sounds to dsh-sound-lab (now matching the repo name); install commands and the deploy dir ($DSH_HOME/plugins/dsh-sound-lab) updated accordingly
• New "AI Voice" dialog: the host side ports the Bailian CosyVoice synthesis protocol (long-text chunking + WAV frame concat); custom file names (409 on duplicates), preview / delete, one-click add to the library with automatic registration (no file copy — direct reference); token consumption is on the user
• 3-state credential status: unconfigured (any missing) / failed (all set but the call errors) / configured; debounced merged saving + credentials carried by the generate request + forced flush on dialog close — no more "configured but prompts unconfigured"
• Built-in illustrated tutorial covering how to get your API Key and cloned voice ID (expandable, dialog up to 4:6, never taller than the page)
• Sound library now manifest-driven: a sounds/sounds.json control file drives the library; unregistered files on disk are reconciled automatically (drop a file in and it works); generated audio lives in sounds/generated/ referenced by path; kinds: ai / local / bundled
• 3 bundled sounds now: "AWAWA!" joins "hirari do～" and "Huh?" — all bundled sounds cannot be deleted or renamed (delete = soft-hide, restorable anytime)
• Sound renaming: non-bundled sounds can be renamed inline via ✎ (auto extension, 409 on duplicates, trigger references update together)
• Hidden sounds preview instantly: all sounds (incl. hidden) are preloaded on startup; preview reuses the cache — no more waiting
• UI polish: the "Custom" button is now "Upload" (closes on upload, stays open on restore, hidden list syncs live); the library scrolls beyond 4 entries; file extensions hidden from names; "No sound" pinned on top
• Engineering: host-side smoke tests expanded to 43 cases (TTS endpoints + manifest reconcile/rename/register); tutorial images ship with the package (lib/tutorial/)

Install (works for both Desktop & web users — pick either)

• Option A (recommended, zero deps): if dsh isn't running, start it once (Desktop app or npx @deepseek-ai/dsh web) to initialize profiles; then double-click dsh-sound-lab-Setup-1.2.0-x64.exe — it restarts dsh automatically
• Option B (needs Node.js) — if dsh isn't running, start it once (npx @deepseek-ai/dsh web) to initialize profiles; then pick one based on how you run dsh and restart dsh:
  • Desktop: npx @deepseek-ai/dsh plugin --profile desktop add dsh-sound-lab --config.minimumReleaseAge=0
  • web: npx @deepseek-ai/dsh plugin --profile web add dsh-sound-lab --config.minimumReleaseAge=0
• No dsh installed/initialized yet? The Setup installer and npm run setup handle first-time initialization automatically (they download Node.js if needed and run npx @deepseek-ai/dsh web); manually, download the DeepSeek Harness Desktop app, or install Node.js (https://nodejs.org) and run npx @deepseek-ai/dsh web. For auto-restart after install, prefer Option A (the Setup installer).
• Uninstall: double-click dsh-sound-lab-UnSetup-1.2.0-x64.exe, then dsh restarts automatically

Fan-made unofficial project; sound assets belong to Arknights and its voice actors. For personal learning & entertainment only — not for commercial use.