dsh-speak 1.7.1 → 1.7.3

package/README.md

@@ -71,9 +71,14 @@ Windows:
 
 - Windows 10 or 11, PowerShell (any recent version).
 - Natural voices:
-  - **Windows 11**: natural voice packs are built into the system — no extra
-    installation. Enable/switch them in *Settings → Accessibility → Narrator* or
-    *Settings → Time & Language → Speech*.
+  - **Windows 11 (21H2–23H2)**: natural voice packs are built into the system —
+    no extra installation. Enable/switch them in *Settings → Accessibility →
+    Narrator* or *Settings → Time & Language → Speech*.
+  - **Windows 11 24H2/25H2**: natural voices moved to MSIX app packages, which
+    `System.Speech` may not enumerate (falls back to a robotic stock voice). As
+    on Windows 10, install
+    [NaturalVoiceSAPIAdapter](https://github.com/gexgd0419/NaturalVoiceSAPIAdapter)
+    to bridge them.
   - **Windows 10**: install
     [NaturalVoiceSAPIAdapter](https://github.com/gexgd0419/NaturalVoiceSAPIAdapter)
     and use its VoiceDownloader to download the natural voice pack(s) you want
@@ -315,7 +320,7 @@ document):
 | `longTextMessage` | `本次播报内容较长，请自行阅读。` | the fixed prompt spoken for over-long text in `message` mode (editable in the UI) |
 | `maxChars` | platform | per-utterance ceiling. **macOS default 0 (`say` has no ceiling); Windows default 300** (SAPI fails silently beyond ~375-470) |
 | `volume` | `50` | Windows only (0-100); macOS volume follows the system |
-| `rate` | `0` | `0` = engine default (Windows SAPI scale, e.g. 1; macOS words-per-minute, e.g. 175) |
+| `rate` | `0` | speech rate: Windows SAPI scale (-10 to 10, 0 = normal; try 1-3 for faster); macOS words-per-minute (default 175, 200 is a bit faster) |
 | `announceTurnEnd` | `false` | announce "第 N 轮对话完成/中断/异常结束" on turn end (`turn/end`) |
 | `announceCommandDone` | `false` | announce when a command finishes or fails (`command/done`) |
 | `announceGoalChange` | `false` | announce goal created/updated/completed/paused/resumed (`goal/change`, objective head) |

package/README.zh-CN.md

@@ -64,8 +64,12 @@ harness 事件（DSH 会话事件 / Claude Code Stop hook / 任意方式）
 
 - Windows 10 或 11，任意较新的 PowerShell。
 - 自然语音：
-  - **Windows 11**：系统已内置自然语音包，无需额外安装——在
+  - **Windows 11（21H2–23H2）**：系统已内置自然语音包，无需额外安装——在
     *设置 → 辅助功能 → 讲述人* 或 *设置 → 时间和语言 → 语音* 中启用/切换即可。
+  - **Windows 11 24H2/25H2**：自然语音已改为 MSIX 应用包，`System.Speech` 可能
+    枚举不到（`SpeechSynthesizer` 找不到自然语音、回退到机械音）——与 Windows 10
+    相同，需安装 [NaturalVoiceSAPIAdapter](https://github.com/gexgd0419/NaturalVoiceSAPIAdapter)
+    桥接。
   - **Windows 10**：需要安装
     [NaturalVoiceSAPIAdapter](https://github.com/gexgd0419/NaturalVoiceSAPIAdapter)，
     并用它的 VoiceDownloader 手动下载你需要的中文或其他语言的自然语音包。
@@ -295,7 +299,7 @@ speak.ps1 -Text "…" -Volume 50 -Rate 1 -MaxChars 300 -LongTextMessage "本次
 | `longTextMessage` | `本次播报内容较长，请自行阅读。` | `message` 模式下超长文本改念的固定提示语（UI 可编辑） |
 | `maxChars` | 平台相关 | 引擎单次朗读上限。**macOS 默认 0（`say` 无上限）；Windows 默认 300**（SAPI 超过约 375-470 字会静默失败） |
 | `volume` | `50` | 仅 Windows（0-100）；macOS 音量跟随系统 |
-| `rate` | `0` | `0` = 引擎默认（Windows SAPI 刻度如 1；macOS words-per-minute 如 175） |
+| `rate` | `0` | 语速：Windows SAPI 刻度（-10 到 10，0 = 正常，推荐 0 / 稍快 1-3）；macOS words-per-minute（默认 175，稍快 200） |
 | `announceTurnEnd` | `false` | 回合结束时播报"第 N 轮对话完成/中断/异常结束"（`turn/end`） |
 | `announceCommandDone` | `false` | 命令执行完成/失败时播报（`command/done`） |
 | `announceGoalChange` | `false` | 目标创建/更新/完成/暂停/恢复时播报（`goal/change`，含目标标题前 40 字） |

package/adapters/dsh/install.ps1

@@ -1,4 +1,4 @@
-# install.ps1 — one-command installer for the DSH adapter
+# install.ps1 — one-command installer for the DSH adapter
 # ========================================================
 # 1. copies engine/*.ps1 to ~/.dsh/hooks/
 # 2. copies speech-hook.js to ~/.dsh/profiles/web/plugins/

package/adapters/dsh/speech-hook.js

@@ -264,7 +264,9 @@ module.exports = {
         if (fullRead) args.push('-F')
         child = spawn('/bin/bash', [cfg.engine].concat(args), { detached: true, stdio: 'ignore' })
       } else {
-        child = spawn('powershell.exe', ['-NoProfile', '-ExecutionPolicy', 'Bypass', '-File', cfg.engine, '-File', tmp, '-Volume', String(cfg.volume), '-Rate', String(cfg.rate > 0 ? cfg.rate : 1), '-MaxChars', String(cfg.maxChars), '-LongTextMode', cfg.longTextMode, '-LongTextMessage', cfg.longTextMessage, '-CleanMarkdownFormatting', cfg.cleanMarkdownFormatting ? '1' : '0', '-ReadInlineCode', cfg.readInlineCode ? '1' : '0', '-CodeBlocks', cfg.codeBlocks, '-CodeBlockMaxChars', String(cfg.codeBlockMaxChars), '-CodeBlockReplacementText', cfg.codeBlockReplacementText, '-FullRead', fullRead ? '1' : '0'], { windowsHide: true, stdio: 'ignore' })
+        // Windows 语速 = SAPI 刻度（-10 到 10，0 = 正常；负值变慢），直接透传，
+        // 不要把 0/负值替换成 1（否则无法回到正常语速/减慢）
+        child = spawn('powershell.exe', ['-NoProfile', '-ExecutionPolicy', 'Bypass', '-File', cfg.engine, '-File', tmp, '-Volume', String(cfg.volume), '-Rate', String(cfg.rate), '-MaxChars', String(cfg.maxChars), '-LongTextMode', cfg.longTextMode, '-LongTextMessage', cfg.longTextMessage, '-CleanMarkdownFormatting', cfg.cleanMarkdownFormatting ? '1' : '0', '-ReadInlineCode', cfg.readInlineCode ? '1' : '0', '-CodeBlocks', cfg.codeBlocks, '-CodeBlockMaxChars', String(cfg.codeBlockMaxChars), '-CodeBlockReplacementText', cfg.codeBlockReplacementText, '-FullRead', fullRead ? '1' : '0'], { windowsHide: true, stdio: 'ignore' })
       }
       const token = ++speechToken
       activeSpeech = { process: child, tmp, token, item, gapMs: item.gapMs || 0 }

package/client/client.js

@@ -66,7 +66,8 @@ window.__ModuleLoader__.load({
       volume: '音量',
       volumeHint: '仅 Windows（0-100）；macOS 音量跟随系统。',
       rate: '语速',
-      rateHint: '0 = 引擎默认（Windows SAPI 刻度 / macOS words-per-minute）。',
+      rateHintWin: 'Windows SAPI 语速（-10 到 10，0 = 正常；推荐 0，稍快 1-3）。',
+      rateHintMac: 'words-per-minute（wpm），数值越大语速越快：默认 175，稍快 200。',
       engine: '引擎路径',
       engineHint: '自定义引擎脚本路径；留空自动解析（包内引擎 → ~/.dsh/hooks）。',
       longTextBehavior: '超长文本行为',
@@ -133,7 +134,8 @@ window.__ModuleLoader__.load({
       volume: 'Volume',
       volumeHint: 'Windows only (0-100); macOS volume follows the system.',
       rate: 'Speech Rate',
-      rateHint: '0 = engine default (Windows SAPI scale / macOS words-per-minute).',
+      rateHintWin: 'Windows SAPI rate scale (-10 to 10, 0 = normal; try 1-3 for faster).',
+      rateHintMac: 'words-per-minute (wpm) — higher = faster: default 175, 200 is a bit faster.',
       engine: 'Engine Path',
       engineHint: 'Custom engine script path; empty = auto-resolve (package engine → ~/.dsh/hooks).',
       longTextBehavior: 'Long Text Behavior',
@@ -274,7 +276,15 @@ window.__ModuleLoader__.load({
         return e('div', { className: 'dsh-speak-field' }, inline ? e('div', { className: 'dsh-speak-inline-field' }, e('div', { className: 'dsh-speak-field-label' }, label), children) : e('div', { className: 'dsh-speak-field-label' }, label), inline ? null : children, e('p', { className: 'dsh-speak-field-hint' }, hint))
       }
       function Toggle({ label, value, disabled, onChange, hint }) { return e(Field, { label: `${label}:`, hint, inline: true }, e('div', { className: 'dsh-speak-option-row' }, e(Button, { variant: 'outline', size: 'sm', disabled, 'aria-pressed': value, onClick: () => onChange(!value) }, value ? t('toggleOn') : t('toggleOff')))) }
-      function SettingInput({ label, value, disabled, numeric, onChange, hint }) { const id = `dsh-speak-${label.toLowerCase().replace(/[^a-z0-9]+/g, '-')}`; return e(Field, { label: `${label}:`, hint }, e('div', { className: 'dsh-speak-input-row' }, e(Input, { id, value: String(value), disabled, inputMode: numeric ? 'numeric' : undefined, onChange: event => onChange(event.target.value) }))) }
+      function SettingInput({ label, value, disabled, numeric, onChange, hint }) {
+        const id = `dsh-speak-${label.toLowerCase().replace(/[^a-z0-9]+/g, '-')}`
+        // 缓冲输入：受控 value 来自配置，直接绑定会让"-"这类中间态被弹回
+        // （Number('-') = NaN 拒绝后 value 不变），负数字段无法输入。这里用
+        // 局部 state 缓冲，输入过程自由，onChange 里校验合法后才写配置。
+        const [text, setText] = React.useState(String(value))
+        React.useEffect(() => { setText(String(value)) }, [value])
+        return e(Field, { label: `${label}:`, hint }, e('div', { className: 'dsh-speak-input-row' }, e(Input, { id, value: text, disabled, inputMode: numeric ? 'numeric' : undefined, onChange: event => { setText(event.target.value); onChange(event.target.value) } })))
+      }
       function Options({ label, value, disabled, onChange, hint, options }) { return e(Field, { label, hint }, e('div', { className: 'dsh-speak-option-row' }, ...options.map(option => e(Button, { key: option.value, variant: value === option.value ? 'primary' : 'outline', size: 'sm', disabled, 'aria-pressed': value === option.value, onClick: () => onChange(option.value) }, option.label)))) }
       function MarkdownCleaning({ value, clean, disabled, set }) {
         const [open, setOpen] = React.useState(true); const controlsDisabled = disabled || !clean
@@ -290,6 +300,7 @@ window.__ModuleLoader__.load({
         const [eventsOpen, setEventsOpen] = React.useState(false)
         const snapshot = useSettings(); if (snapshot.status !== 'ready' || !snapshot.value) return null
         const value = snapshot.value; const disabled = !snapshot.writable; const clean = value.cleanMarkdownFormatting !== false; const set = (field, next) => { void settings.set(field, next).catch(console.error) }
+        const isMac = /Mac|iPhone|iPad|iPod/.test(navigator.userAgent || navigator.platform || '')
         return e('section', { 'aria-label': t('settingsAria') }, e('h3', null, t('settingsTitle')), e('p', null, t('settingsIntro')),
           e(Toggle, { label: t('masterSwitch'), value: value.enabled !== false, disabled, onChange: next => set('enabled', next), hint: t('masterSwitchHint') }),
           e(Toggle, { label: t('automaticSpeech'), value: value.automaticSpeech !== false, disabled, onChange: next => set('automaticSpeech', next), hint: t('automaticSpeechHint') }),
@@ -299,7 +310,7 @@ window.__ModuleLoader__.load({
           e(Toggle, { label: t('cleanMarkdown'), value: clean, disabled, onChange: next => set('cleanMarkdownFormatting', next), hint: t('cleanMarkdownHint') }), e(MarkdownCleaning, { value, clean, disabled, set }),
           e(SettingInput, { label: t('maxChars'), value: value.maxChars == null ? 0 : value.maxChars, numeric: true, disabled, onChange: next => { if (/^\d+$/.test(next)) set('maxChars', Number(next)) }, hint: t('maxCharsHint') }),
           e(SettingInput, { label: t('volume'), value: value.volume == null ? 50 : value.volume, numeric: true, disabled, onChange: next => { if (/^\d+$/.test(next)) set('volume', Number(next)) }, hint: t('volumeHint') }),
-          e(SettingInput, { label: t('rate'), value: value.rate == null ? 0 : value.rate, numeric: true, disabled, onChange: next => { if (/^-?\d*\.?\d*$/.test(next)) set('rate', Number(next)) }, hint: t('rateHint') }),
+          e(SettingInput, { label: t('rate'), value: value.rate == null ? 0 : value.rate, numeric: true, disabled, onChange: next => { if (/^-?\d*\.?\d*$/.test(next)) { const n = Number(next); if (Number.isFinite(n)) set('rate', n) } }, hint: isMac ? t('rateHintMac') : t('rateHintWin') }),
           e(SettingInput, { label: t('engine'), value: value.engine || '', disabled, onChange: next => set('engine', next), hint: t('engineHint') }),
           e(Options, { label: t('longTextBehavior'), value: value.longTextMode || 'message', disabled, onChange: next => set('longTextMode', next), hint: t('longTextBehaviorHint'), options: [{ value: 'message', label: t('longTextMessageOption') }, { value: 'heading', label: t('longTextHeadingOption') }] }),
           e(SettingInput, { label: t('fixedPrompt'), value: value.longTextMessage || '本次播报内容较长，请自行阅读。', disabled: disabled || value.longTextMode !== 'message', onChange: next => set('longTextMessage', next), hint: t('fixedPromptHint') }),

package/docs/DESIGN.md

@@ -253,7 +253,7 @@ config:
   longTextMessage: '本次播报内容较长，请自行阅读。'
   maxChars: 300                # macOS default 0 = unlimited
   volume: 50                   # Windows only
-  rate: 0                      # 0 = engine default
+  rate: 0                      # rate: Win SAPI scale -10..10 (0=normal) / mac wpm (175)
   # —— optional event announcements (off by default) ——
   announceTurnEnd: false     # turn/end
   announceCommandDone: false # command/done

package/docs/DESIGN.zh-CN.md

@@ -232,7 +232,7 @@ config:
   longTextMessage: '本次播报内容较长，请自行阅读。'
   maxChars: 300                # macOS 默认 0 = 不限
   volume: 50                   # 仅 Windows
-  rate: 0                      # 0 = 引擎默认
+  rate: 0                      # 语速：Win SAPI 刻度 -10~10（0=正常）/ mac wpm（175）
   # —— 可选事件播报（默认全关）——
   announceTurnEnd: false     # turn/end
   announceCommandDone: false # command/done

package/engine/speak.ps1

@@ -68,7 +68,12 @@ if (-not $fullReadMode -and $text.Length -gt $MaxChars -and $LongTextMode -eq 'h
     $candidate = ''
     $bestLevel = 7
     $firstNonEmpty = ''
+    # 代码块 fence 内的行跳过：其中的 "# 注释" 不是 markdown 标题，
+    # 否则长回复里的代码注释会被误当成标题只念注释
+    $inCodeBlock = $false
     foreach ($line in ($text -split "`n")) {
+        if ($line -match '^\s*```') { $inCodeBlock = -not $inCodeBlock; continue }
+        if ($inCodeBlock) { continue }
         if ($line -match '^\s*#{1,6}\s+') {
             $level = ([regex]::Match($line, '^(\s*)(#+)')).Groups[2].Value.Length
             if ($level -lt $bestLevel) {

package/engine/speak.sh

@@ -56,8 +56,10 @@ fi
 # -F (full read, manual replay) skips this guard entirely.
 if [ "$FULL_READ" != "1" ] && [ "$MAX_CHARS" -gt 0 ] && [ "${#TEXT}" -gt "$MAX_CHARS" ] && [ "$LONG_MODE" = "heading" ]; then
   TEXT=$(printf '%s' "$TEXT" | /usr/bin/perl -CSD -e '
-    my $best = 7; my $cand = ""; my $first = "";
+    my $best = 7; my $cand = ""; my $first = ""; my $inCode = 0;
     while (<STDIN>) {
+      if (/^\s*```/) { $inCode = !$inCode; next; }   # 跳过代码块内的 "# 注释"
+      next if $inCode;
       if (/^[ \t]*(\#{1,6})[ \t]+(.*)$/) { my $n = length($1); if ($n < $best) { $best = $n; $cand = $2; } }
       elsif ($first eq "" && /\S/) { $first = $_; }
     }

package/engine/speech-prompt.ps1

@@ -1,4 +1,4 @@
-# speech-prompt.ps1 — Short prompt announcement (synchronous, blocking)
+# speech-prompt.ps1 — Short prompt announcement (synchronous, blocking)
 # Use when a harness/agent needs the user's attention (a question, an approval
 # request). Reads the text through engine/speak.ps1 and waits until it finishes,
 # so the caller knows the announcement was actually spoken.

package/engine/speech-summary.ps1

@@ -1,4 +1,4 @@
-# speech-summary.ps1 — Reply summary announcement (synchronous, blocking)
+# speech-summary.ps1 — Reply summary announcement (synchronous, blocking)
 # For harnesses with no "reply finished" event (e.g. DSH has no Stop hook): the
 # agent calls this at the end of its final reply.
 #

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-speak",
-  "version": "1.7.1",
+  "version": "1.7.3",
   "description": "Make your AI harness speak — voice announcements for DSH and other AI coding harnesses (Windows SAPI5 + macOS system voices)",
   "main": "adapters/dsh/speech-hook.js",
   "exports": {