dsh-dafeiyu-emotion 0.2.8 → 0.3.0

package/CHANGELOG.md

@@ -1,5 +1,33 @@
 # Changelog
 
+## 0.3.0
+
+- Added `empathyStyle` with four reply strategies: gentle, direct, humorous and concise. The choice is persisted and injected into the emotion-aware prompt.
+- Added an in-composer emotion timeline for analyses with multiple segments; hovering a segment shows its text, emotion and confidence.
+- Added privacy/runtime status to the settings panel and `/health`: last emotion, last audio size, model load time, worker PID and last temp cleanup time.
+- Added local-only `GET /plugins/dsh-dafeiyu-emotion/context` so other plugins can read the latest emotion context.
+- Settings now resolve the real `entry.ns` from `settings.describe()` instead of assuming the npm package name.
+- Settings writes now carry the settings revision and return 409 on a concurrent-window conflict.
+- Removed the hard `settings` inject; settings are resolved dynamically after the host is up.
+- Moved Python discovery to the first worker start, removing the synchronous filesystem stat from `apply`.
+- An explicit empty `backendRoot` now clears the environment fallback, so the UI can disable the enhanced backend.
+- Shortcut resolution now prefers the most recently used fish-ear instead of the last mounted one.
+- Narrow sidebar fields wrap instead of squeezing controls.
+- Removed test scripts from the published npm tarball.
+
+## 0.2.10
+
+- Removed `settings` from the hard inject list. A hard `settings` dependency can deadlock host startup while the settings service is still building plugin `Config` projections.
+- Settings are now resolved with `ctx.inject(['settings'], ...)` after the host is up; `/config` is read-only until then and becomes writable when the service arrives.
+- Added a regression assertion that `settings` is not a hard dependency.
+
+## 0.2.9
+
+- Fixed the host startup timeout: `apply` synchronously called `settings.describe()` while the host was still mounting plugins, which could deadlock startup for 60 seconds.
+- `apply` now uses the patch config only; settings are read/written through `/config` after the host is up.
+- Added cycle/depth guards to `plainConfig()` so an unexpected volatile-ref cycle cannot hang startup or a settings request.
+- Added a regression test where `settings.describe()` throws if called during `apply`.
+
 ## 0.2.8
 
 - Fixed `backendRoot` showing as `[object Object]` on the settings page. DSH projects volatile `Config` fields as `{ get() }` refs; `readSettings()` and `resolveOptions()` now unwrap them before path/string/number handling.

package/README.md

@@ -72,6 +72,10 @@
 - ⌨️ DSH 原生语音快捷键，可在快捷键设置中自行改键；
 - 🎛️ DSH 设置页里的完整设置卡片：启用、语言、增强后端、超时和录音上限；
 - 🧭 安装 Better Sidebar 后，可把「情绪耳朵」固定到右侧边栏；
+- 🎭 共情策略四档：温柔、直接、幽默、简洁；
+- 📈 多段语音的情绪时间线，可悬停查看每段文本、情绪和置信度；
+- 🔐 设置页内的隐私与运行状态：最近音频大小、模型加载、worker PID、临时文件清理；
+- 🔌 本机 `/context` 接口，其他插件可以读取最近一次情绪上下文；
 - 🐋 大肥鱼主题的“鱼耳”入口，和普通对话共用同一个 composer；
 - 🧠 情绪标签：喜悦 / 低落 / 生气 / 紧张 / 惊讶 / 厌烦 / 平静；
 - 📈 额外给出 confidence、valence、arousal 和语气摘要；
@@ -111,9 +115,9 @@ dsh plugin --profile desktop add github:LiDu-china/dsh-dafeiyu-emotion
 ### 通过 Release tarball
 
 ```bash
-curl -L -o dsh-dafeiyu-emotion-0.2.8.tgz \
-  https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/download/v0.2.8/dsh-dafeiyu-emotion-0.2.8.tgz
-dsh plugin --profile desktop add "D:/path/to/dsh-dafeiyu-emotion-0.2.8.tgz"
+curl -L -o dsh-dafeiyu-emotion-0.3.0.tgz \
+  https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/download/v0.3.0/dsh-dafeiyu-emotion-0.3.0.tgz
+dsh plugin --profile desktop add "D:/path/to/dsh-dafeiyu-emotion-0.3.0.tgz"
 ```
 
 > DSH Desktop 的 profile 由 Electron 应用管理；命令行安装请使用桌面应用自带的 `dsh` 命令。
@@ -130,6 +134,7 @@ dsh plugin --profile desktop add "D:/path/to/dsh-dafeiyu-emotion-0.2.8.tgz"
 
 - 启用 / 停用情绪耳朵；
 - 识别语言：中文、English 或自动；
+- 共情策略：温柔、直接、幽默、简洁；
 - VoiceSense 增强后端目录与 Python 路径；
 - 单次录音上限、后端启动超时、单次分析超时。
 
@@ -184,6 +189,7 @@ dsh plugin --profile desktop add "D:/path/to/dsh-dafeiyu-emotion-0.2.8.tgz"
 | `backendRoot` | 空 | VoiceSense 项目根目录；空则使用 DSH Voice Input 轻量模式 |
 | `pythonPath` | 空 | 自定义 Python；空则自动寻找 `.venv-ser` |
 | `language` | `zh` | 识别语言：`zh` / `en` / `auto` |
+| `empathyStyle` | `gentle` | 共情策略：`gentle` / `direct` / `humorous` / `concise` |
 | `maxAudioBytes` | 12 MB | 单次录音大小上限 |
 | `startupTimeoutMs` | 180000 | 首次加载增强后端的最长等待 |
 | `requestTimeoutMs` | 180000 | 单次分析请求超时 |

package/cordis.patch.yml

@@ -28,6 +28,7 @@
         backendRoot: ""
         pythonPath: ""
         language: zh
+        empathyStyle: gentle
         maxAudioBytes: 12582912
         startupTimeoutMs: 180000
         requestTimeoutMs: 180000

package/docs/MARKETPLACE.md

@@ -72,7 +72,7 @@ https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/download/v0.2.3/dsh-d
 dsh plugin --profile desktop add dsh-dafeiyu-emotion
 ```
 
-Published as `dsh-dafeiyu-emotion@0.2.8` on npm. The 1024Store entry refreshes its one-click install automatically. GitHub-source and Release-tarball installs remain available as fallbacks.
+Published as `dsh-dafeiyu-emotion@0.3.0` on npm. The 1024Store entry refreshes its one-click install automatically. GitHub-source and Release-tarball installs remain available as fallbacks.
 
 ## Privacy
 

package/docs/RELEASE-0.2.10.md

@@ -0,0 +1,20 @@
+## 0.2.10
+
+修复 Host 启动阶段的依赖死锁，并修正 volatile 配置解包。
+
+- 移除 `settings` 硬依赖。改为 `ctx.inject(['settings'], ...)` 动态获取，避免 settings 服务尚未就绪时插件 fiber 一直 pending。
+- `apply` 不再调用 `settings.describe()`，只在 `/config` 请求时读取运行时配置。
+- `plainConfig()` 改为递归栈 + `finally` 清理，修复共享对象被误判为循环引用而丢失的问题。
+- 新增共享引用、真实循环、`settings.describe()` 不应在启动阶段调用的回归测试。
+
+## 注意
+
+当前环境如果仍无法启动，请先确认 `magpie-mcp-git` / `magpie-mcp-anysearch` 没有阻塞 Host 启动。这两个 MCP 客户端在连接不可用时会等待重试，可能把 Host 启动拖过 60 秒。
+
+## 安装
+
+```bash
+dsh plugin --profile desktop add dsh-dafeiyu-emotion
+```
+
+升级后请完全退出并重新启动 the agent harness。

package/docs/RELEASE-0.2.9.md

@@ -0,0 +1,17 @@
+## 0.2.9
+
+修复“the agent harness 打不开 / Host startup exceeded 60 seconds”。
+
+- 根因：0.2.8 在 `apply` 期间同步调用 `settings.describe()`。此时 Host 还在挂载插件，遍历所有插件配置可能造成启动死锁，最终 60 秒超时。
+- 0.2.9 的 `apply` 只读取 `cordis.patch.yml` 的 patch 配置，不再在启动事务里调用 `settings.describe()`。
+- 设置页的读取/写入仍通过 `/plugins/dsh-dafeiyu-emotion/config`；`PATCH` 会更新共享运行时配置并调用 `settings.update()` 持久化。
+- `plainConfig()` 增加循环引用和深度保护，避免异常 volatile ref 卡死启动或设置请求。
+- 新增回归测试：`settings.describe()` 在 `apply` 期间如果被调用会直接抛错。
+
+## 安装
+
+```bash
+dsh plugin --profile desktop add dsh-dafeiyu-emotion
+```
+
+升级后请完全退出并重新启动 the agent harness。若之前因 0.2.8 无法启动，可先进入安全模式或临时移除插件，再安装 0.2.9。

package/docs/RELEASE-0.3.0.md

@@ -0,0 +1,29 @@
+## 0.3.0
+
+这是情绪耳朵的功能版本，新增共情策略、情绪时间线、隐私状态和可订阅的情绪上下文。
+
+### 新增
+
+- **共情策略**：温柔 / 直接 / 幽默 / 简洁，保存后会影响情绪化 prompt。
+- **情绪时间线**：多段分析时在输入框内显示一条可悬停的情绪条，标出每段文本、情绪和置信度。
+- **隐私与运行状态**：设置页显示最近情绪、最近音频大小、模型加载耗时、worker PID 和最近临时文件清理时间。
+- **`/context` 接口**：本机 `GET /plugins/dsh-dafeiyu-emotion/context` 返回最近一次情绪上下文，供其他插件读取。
+
+### 修复
+
+- settings 从 `describe()` 解析真实 `entry.ns`，不再硬编码包名。
+- 保存设置时携带 revision，两个窗口同时修改会返回 409。
+- `settings` 改为动态注入，避免 Host 启动阶段硬依赖。
+- Python 查找延迟到首次启动 worker，`apply` 不再同步 stat 文件系统。
+- 显式清空 `backendRoot` 可以覆盖环境变量。
+- 快捷键优先使用最近操作过的鱼耳，而不是最后挂载的实例。
+- 窄侧边栏字段自动换行。
+- 发布包不再包含测试脚本。
+
+### 安装
+
+```bash
+dsh plugin --profile desktop add dsh-dafeiyu-emotion
+```
+
+升级后请完全退出并重新启动 the agent harness。

package/lib/client.js

@@ -37,19 +37,35 @@ window.__ModuleLoader__.load({
     }
 
     function registerVoiceShortcut(handler) {
-      voiceShortcutHandlers.push(handler)
-      return () => {
-        const index = voiceShortcutHandlers.lastIndexOf(handler)
+      const entry = { ...handler, lastActiveAt: Date.now() }
+      voiceShortcutHandlers.push(entry)
+      const dispose = () => {
+        const index = voiceShortcutHandlers.lastIndexOf(entry)
         if (index >= 0) voiceShortcutHandlers.splice(index, 1)
       }
+      dispose.touch = () => {
+        entry.lastActiveAt = Date.now()
+      }
+      return dispose
     }
 
     function canUseVoiceShortcut(phase, locked) {
       return !locked && (phase === 'idle' || phase === 'recording' || phase === 'error')
     }
 
+    function activeVoiceShortcutHandler() {
+      const candidates = [...voiceShortcutHandlers].sort((left, right) => right.lastActiveAt - left.lastActiveAt)
+      return candidates.find((handler) => {
+        try {
+          return handler.canHandle?.() === true
+        } catch {
+          return false
+        }
+      }) || candidates[0]
+    }
+
     function canInvokeVoiceShortcut() {
-      const handler = voiceShortcutHandlers.at(-1)
+      const handler = activeVoiceShortcutHandler()
       if (!handler) return false
       try {
         return handler.canHandle?.() === true
@@ -59,7 +75,7 @@ window.__ModuleLoader__.load({
     }
 
     function invokeVoiceShortcut() {
-      const handler = voiceShortcutHandlers.at(-1)
+      const handler = activeVoiceShortcutHandler()
       if (!handler) return false
       try {
         return handler.run?.() === true
@@ -175,6 +191,23 @@ window.__ModuleLoader__.load({
         fontSize: 11,
         opacity: 0.6,
       },
+      timeline: {
+        display: 'inline-flex',
+        alignItems: 'center',
+        gap: 1,
+        width: 58,
+        height: 6,
+        flex: '0 0 auto',
+        borderRadius: 99,
+        overflow: 'hidden',
+        background: 'color-mix(in srgb, currentColor 10%, transparent)',
+      },
+      timelinePart: {
+        height: '100%',
+        minWidth: 2,
+        borderRadius: 99,
+        opacity: 0.9,
+      },
     }
 
     const T = {
@@ -599,15 +632,15 @@ window.__ModuleLoader__.load({
     function SettingsField({ icon, label, hint, children, wide = false }) {
       return h('div', {
         style: {
-          display: 'grid',
-          gridTemplateColumns: wide ? 'minmax(0, 1fr)' : 'minmax(0, 1fr) auto',
+          display: 'flex',
+          flexWrap: 'wrap',
           gap: wide ? 7 : 12,
           alignItems: 'center',
           padding: '11px 0',
           borderBottom: `1px solid ${T.border}`,
         },
       },
-        h('div', { style: { minWidth: 0 } },
+        h('div', { style: { minWidth: 0, flex: wide ? '1 1 100%' : '1 1 200px' } },
           h('div', {
             style: {
               display: 'flex',
@@ -789,6 +822,7 @@ window.__ModuleLoader__.load({
         || draft.backendRoot !== config.backendRoot
         || draft.pythonPath !== config.pythonPath
         || draft.language !== config.language
+        || draft.empathyStyle !== config.empathyStyle
         || draft.maxAudioBytes !== config.maxAudioBytes
         || draft.startupTimeoutMs !== config.startupTimeoutMs
         || draft.requestTimeoutMs !== config.requestTimeoutMs
@@ -797,19 +831,25 @@ window.__ModuleLoader__.load({
       const save = async () => {
         if (!config || !draft || saving) return
         const patch = {}
-        for (const key of ['enabled', 'backendRoot', 'pythonPath', 'language', 'maxAudioBytes', 'startupTimeoutMs', 'requestTimeoutMs']) {
+        for (const key of ['enabled', 'backendRoot', 'pythonPath', 'language', 'empathyStyle', 'maxAudioBytes', 'startupTimeoutMs', 'requestTimeoutMs']) {
           if (draft[key] !== config[key]) patch[key] = draft[key]
         }
         if (Object.keys(patch).length === 0) return
         setSaving(true)
         setNotice(null)
         try {
+          const body = { ...patch }
+          if (Number.isInteger(config.revision)) body.expectedRevision = config.revision
           const response = await fetch(CONFIG_ENDPOINT, {
             method: 'PATCH',
             headers: { 'content-type': 'application/json' },
-            body: JSON.stringify(patch),
+            body: JSON.stringify(body),
           })
           const payload = await response.json().catch(() => ({}))
+          if (response.status === 409) {
+            setNotice({ kind: 'error', text: '设置已被其他窗口修改，请重新读取后再保存' })
+            return
+          }
           if (!response.ok) throw new Error(payload.error || `保存失败（${response.status}）`)
           setConfig(payload)
           setDraft(payload)
@@ -942,6 +982,18 @@ window.__ModuleLoader__.load({
         cursor: dirty && !saving ? 'pointer' : 'not-allowed',
         opacity: dirty && !saving ? 1 : 0.45,
       }
+      const stat = (label, value) => h('div', { style: { minWidth: 0 } },
+        h('div', { style: { color: T.faint, fontSize: 10.5, lineHeight: 1.35 } }, label),
+        h('div', {
+          style: {
+            marginTop: 2,
+            color: T.text,
+            fontSize: 12,
+            lineHeight: 1.4,
+            wordBreak: 'break-word',
+          },
+        }, value),
+      )
 
       return h(Root, {
         'data-testid': 'dsh-dafeiyu-emotion-settings',
@@ -1166,6 +1218,22 @@ window.__ModuleLoader__.load({
               ],
               onChange: (value) => update('language', value),
             })),
+            h(SettingsField, {
+              icon: 'spark',
+              label: '共情策略',
+              hint: '决定大肥鱼先怎么接住情绪，再回答问题。',
+            }, h(SettingsSegmented, {
+              value: draft.empathyStyle || 'gentle',
+              label: '共情策略',
+              disabled,
+              options: [
+                { value: 'gentle', label: '温柔' },
+                { value: 'direct', label: '直接' },
+                { value: 'humorous', label: '幽默' },
+                { value: 'concise', label: '简洁' },
+              ],
+              onChange: (value) => update('empathyStyle', value),
+            })),
           ),
           h(SettingsSection, { title: '增强后端', hint: '可选' },
             h(SettingsField, {
@@ -1203,6 +1271,36 @@ window.__ModuleLoader__.load({
               style: { ...inputStyle, width: '100%' },
             })),
           ),
+          h(SettingsSection, { title: '隐私与运行状态' },
+            h('div', {
+              style: {
+                display: 'grid',
+                gridTemplateColumns: 'repeat(2, minmax(0, 1fr))',
+                gap: '10px 14px',
+                padding: '6px 0 2px',
+              },
+            },
+              stat('最近情绪', health?.lastAnalysis
+                ? `${health.lastAnalysis.emotionZh || '—'} · ${Math.round((Number(health.lastAnalysis.confidence) || 0) * 100)}%`
+                : '还没有分析记录'),
+              stat('最近音频', health?.privacy?.lastAudioBytes
+                ? `${Math.round(health.privacy.lastAudioBytes / 1024)} KB`
+                : '—'),
+              stat('模型加载', health?.privacy?.modelLoadMs != null
+                ? `${health.privacy.modelLoadMs} ms`
+                : '—'),
+              stat('worker PID', health?.privacy?.workerPid ?? '—'),
+              stat('最近清理', health?.privacy?.lastCleanupAt
+                ? new Date(health.privacy.lastCleanupAt).toLocaleTimeString()
+                : '—'),
+              stat('共情策略', ({
+                gentle: '温柔',
+                direct: '直接',
+                humorous: '幽默',
+                concise: '简洁',
+              })[draft.empathyStyle] || '温柔'),
+            ),
+          ),
           h('details', {
             onToggle: (event) => {
               const marker = event.currentTarget.querySelector('[data-chevron]')
@@ -1359,6 +1457,7 @@ window.__ModuleLoader__.load({
       const generation = useRef(0)
       const levelsRef = useRef([])
       const shortcutAction = useRef(null)
+      const shortcutRegistration = useRef(null)
       const deliveryGate = useRef(null)
       if (!deliveryGate.current) deliveryGate.current = createDeliveryGate()
       const useInputSnapshot = typeof useInput === 'function' ? useInput : useNoInput
@@ -1619,10 +1718,21 @@ window.__ModuleLoader__.load({
         }
       }, [locked, phase, sessionId, pluginEnabled])
 
-      useEffect(() => registerVoiceShortcut({
-        canHandle: () => shortcutAction.current?.canHandle?.() === true,
-        run: () => shortcutAction.current?.run?.() === true,
-      }), [])
+      useEffect(() => {
+        const registration = registerVoiceShortcut({
+          canHandle: () => shortcutAction.current?.canHandle?.() === true,
+          run: () => shortcutAction.current?.run?.() === true,
+        })
+        shortcutRegistration.current = registration
+        return () => {
+          registration()
+          if (shortcutRegistration.current === registration) shortcutRegistration.current = null
+        }
+      }, [])
+
+      const touchShortcut = () => {
+        shortcutRegistration.current?.touch?.()
+      }
 
       useEffect(() => {
         let active = true
@@ -1682,6 +1792,7 @@ window.__ModuleLoader__.load({
             event.preventDefault()
             event.currentTarget.style.transform = 'translateY(0)'
           },
+          onFocus: touchShortcut,
           onClick: () => void start(),
         },
           h('span', { style: { display: 'inline-flex', color: '#3e7cb1' } }, h(WhaleMark, { size: 16 })),
@@ -1692,7 +1803,12 @@ window.__ModuleLoader__.load({
       const emotion = analysis?.emotion || 'neutral'
       const color = EMOTION_COLORS[emotion] || EMOTION_COLORS.neutral
       const emotionText = analysis?.emotionZh || EMOTION_LABELS[emotion] || '分析中'
-      return h('div', { style: S.row, 'data-voice-emotion': phase },
+      return h('div', {
+        style: S.row,
+        'data-voice-emotion': phase,
+        onPointerDown: touchShortcut,
+        onFocusCapture: touchShortcut,
+      },
         phase === 'recording'
           ? bars(level, levels)
           : h('span', { style: { ...S.pill, color: phase === 'result' ? color : 'inherit' } },
@@ -1707,6 +1823,21 @@ window.__ModuleLoader__.load({
                       ? emotionText
                       : '出错',
             ),
+        phase === 'result' && analysis?.segments?.length > 1
+          ? h('span', {
+              style: S.timeline,
+              role: 'img',
+              'aria-label': '情绪时间线',
+            }, ...analysis.segments.map((segment, index) => h('i', {
+              key: index,
+              title: `${segment.text} · ${EMOTION_LABELS[segment.emotion] || segment.emotion} ${Math.round((Number(segment.confidence) || 0) * 100)}%`,
+              style: {
+                ...S.timelinePart,
+                width: `${Math.max(2, (Number(segment.end) - Number(segment.start)) / Math.max(0.1, Number(analysis.duration) || 1) * 100)}%`,
+                background: EMOTION_COLORS[segment.emotion] || EMOTION_COLORS.neutral,
+              },
+            })))
+          : null,
         h('span', { style: S.message, title: message || '正在录音' },
           phase === 'recording'
             ? h(React.Fragment, null,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-dafeiyu-emotion",
-  "version": "0.2.8",
+  "version": "0.3.0",
   "description": "大肥鱼语音情绪耳朵：听语气、识别情绪，并让 DSH 按对方当下的状态自然回复。 · Voice + emotion-aware companion for DeepSeek Harness.",
   "type": "module",
   "main": "src/index.js",
@@ -17,7 +17,6 @@
     "docs/*.md",
     "docs/*.svg",
     "docs/*.png",
-    "scripts/*.mjs",
     "assets/*.png",
     "screenshots.json",
     "icon.svg",