@goodandready/dsh-voice 0.9.32 → 0.9.34

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (5) hide show
  1. package/lib/client.js +601 -29
  2. package/lib/host-routes.js +66 -3
  3. package/lib/stats.js +51 -1
  4. package/lib/updater.js +75 -8
  5. package/package.json +1 -1
package/lib/client.js CHANGED
@@ -71,16 +71,10 @@ window.__ModuleLoader__.load({
71
71
  'webgpuStatus': 'Browser WebGPU Status',
72
72
  'webgpuAvailable': 'WebGPU Available (hardware acceleration ready)',
73
73
  'webgpuNotAvailable': 'WebGPU Not Detected (will fallback to server)',
74
- 'openaiHint': 'whisper-1 · key OPENAI_API_KEY',
75
- 'siliconflowHint': 'FunAudioLLM/SenseVoiceSmall · key SILICONFLOW_API_KEY',
76
- 'deepinfraHint': 'openai/whisper-large-v3-turbo · key DEEPINFRA_API_KEY',
77
- 'fireworksHint': 'whisper-v3-turbo · key FIREWORKS_API_KEY',
78
- 'mistralHint': 'voxtral-mini-latest · key MISTRAL_API_KEY',
79
- 'openrouterHint': 'google/gemini-2.5-flash · key OPENROUTER_API_KEY',
80
- 'localHint': 'set when the server is started',
81
- 'up': 'Up',
82
- 'down': 'Down',
83
- 'remove': 'Remove',
74
+ 'openaiHint': 'whisper-1 · key OPENAI_API_KEY', 'siliconflowHint': 'FunAudioLLM/SenseVoiceSmall · key SILICONFLOW_API_KEY',
75
+ 'deepinfraHint': 'openai/whisper-large-v3-turbo · key DEEPINFRA_API_KEY', 'fireworksHint': 'whisper-v3-turbo · key FIREWORKS_API_KEY',
76
+ 'mistralHint': 'voxtral-mini-latest · key MISTRAL_API_KEY', 'openrouterHint': 'google/gemini-2.5-flash · key OPENROUTER_API_KEY',
77
+ 'localHint': 'set when the server is started', 'up': 'Up', 'down': 'Down', 'remove': 'Remove',
84
78
  'addProvider': 'Add provider',
85
79
  'chainHint': 'Top to bottom is the order they are tried in',
86
80
  'customName': 'name used in the chain',
@@ -279,6 +273,55 @@ window.__ModuleLoader__.load({
279
273
  'assistantSpeakingGated': 'Assistant speaking (gated mode)',
280
274
  'jargonFormatHint': ' (spoken -> replacement)',
281
275
  'errorPrefix': 'Error',
276
+ 'wizardTitle': 'First-Run Setup & Diagnostics Wizard',
277
+ 'wizardDesc': 'Step-by-step diagnostic test: check secure context, microphone permissions, audio level, and STT round-trip.',
278
+ 'wizardBtnLaunch': 'Launch Mic & STT Wizard',
279
+ 'wizardClose': 'Close Wizard',
280
+ 'wizardStep1': '1. Audio & Environment',
281
+ 'wizardStep2': '2. Live Signal Test',
282
+ 'wizardStep3': '3. STT Transcription Test',
283
+ 'wizardSecureContextOk': 'Secure context / HTTPS verified',
284
+ 'wizardSecureContextBad': 'Insecure context (HTTPS or localhost required)',
285
+ 'wizardMediaDevicesOk': 'Browser mediaDevices API supported',
286
+ 'wizardMediaDevicesBad': 'Browser mediaDevices API missing',
287
+ 'wizardMediaRecorderOk': 'MediaRecorder API supported',
288
+ 'wizardMediaRecorderBad': 'MediaRecorder API unsupported in this browser',
289
+ 'wizardMicPermissionOk': 'Microphone permission granted',
290
+ 'wizardMicPermissionPrompt': 'Microphone permission not yet granted',
291
+ 'wizardMicPermissionBad': 'Microphone permission denied',
292
+ 'wizardRequestMic': 'Request Microphone Access',
293
+ 'wizardMicLabel': 'Microphone input device:',
294
+ 'wizardStartSignalCheck': 'Start Audio Level Test',
295
+ 'wizardStopSignalCheck': 'Stop Audio Test',
296
+ 'wizardSignalOk': 'Audio signal detected (RMS: {level}%)',
297
+ 'wizardSignalZero': 'Zero audio level detected - speak into microphone or select correct device',
298
+ 'wizardSignalTesting': 'Listening for audio signal… speak into your microphone',
299
+ 'wizardProviderLabel': 'Active STT engine & model:',
300
+ 'wizardProcessingArea': 'Processing area:',
301
+ 'wizardAreaBrowser': 'In-browser (WebGPU / WebSpeech)',
302
+ 'wizardAreaServer': 'Local MiniPC server (SenseVoice / Whisper)',
303
+ 'wizardAreaCloud': 'Remote cloud provider',
304
+ 'wizardStartTestRecord': 'Record 3-Second Test Phrase',
305
+ 'wizardRecordingTest': 'Recording test phrase… speak now ({left}s)',
306
+ 'wizardTranscribingTest': 'Transcribing test phrase…',
307
+ 'wizardTestResult': 'STT Test Result:',
308
+ 'wizardTestSuccess': 'STT round-trip verified successfully! Latency: {latency}ms',
309
+ 'wizardTestEmpty': 'No speech detected in test recording. Please speak clearly and try again.',
310
+ 'wizardTestFailed': 'STT test failed: {error}',
311
+ 'wizardNextStepHint': 'Next step: {hint}',
312
+ 'wizardHintPermission': 'Click "Request Microphone Access" and allow browser permission prompt.',
313
+ 'wizardHintDevice': 'Check microphone connection or select a different input device.',
314
+ 'wizardHintHttps': 'Open DeepSeek Harness via HTTPS or localhost address.',
315
+ 'wizardHintProvider': 'Configure STT engine credentials or install SenseVoice model below.',
316
+ 'wizardFinishSuccess': 'Setup & Verification Complete',
317
+ 'wizardSkip': 'Skip Wizard',
318
+ 'wizardStatusCompleted': 'Microphone and STT verified',
319
+ 'exportDiagnosticsBtn': 'Export Diagnostics Report', 'exportingDiagnostics': 'Exporting report…',
320
+ 'diagnosticsExportSuccess': 'Report downloaded', 'recentRecognitionsTitle': 'Recent Recognition Telemetry',
321
+ 'noRecognitionsYet': 'No recent recognition records yet.', 'diagLatencyBreakdown': 'Latency breakdown:',
322
+ 'diagVad': 'VAD: {ms}ms', 'diagStt': 'STT: {ms}ms', 'diagPolish': 'Polish: {ms}ms',
323
+ 'diagTotal': 'Total: {ms}ms', 'diagBytes': 'Audio: {bytes} B', 'diagWords': 'Words: {count}',
324
+ 'diagExportNotice': 'Exported JSON contains anonymized timings, provider chain, MIME format, and error codes without raw audio or speech transcripts.',
282
325
  }
283
326
 
284
327
  const zh = {
@@ -318,16 +361,10 @@ window.__ModuleLoader__.load({
318
361
  'webgpuStatus': '浏览器 WebGPU 状态',
319
362
  'webgpuAvailable': 'WebGPU 可用 (硬件加速就绪)',
320
363
  'webgpuNotAvailable': '未检测到 WebGPU (将自动回退到服务端)',
321
- 'openaiHint': 'whisper-1 · 密钥 OPENAI_API_KEY',
322
- 'siliconflowHint': 'FunAudioLLM/SenseVoiceSmall · 密钥 SILICONFLOW_API_KEY',
323
- 'deepinfraHint': 'openai/whisper-large-v3-turbo · 密钥 DEEPINFRA_API_KEY',
324
- 'fireworksHint': 'whisper-v3-turbo · 密钥 FIREWORKS_API_KEY',
325
- 'mistralHint': 'voxtral-mini-latest · 密钥 MISTRAL_API_KEY',
326
- 'openrouterHint': 'google/gemini-2.5-flash · 密钥 OPENROUTER_API_KEY',
327
- 'localHint': '服务器启动时配置',
328
- 'up': '上移',
329
- 'down': '下移',
330
- 'remove': '移除',
364
+ 'openaiHint': 'whisper-1 · 密钥 OPENAI_API_KEY', 'siliconflowHint': 'FunAudioLLM/SenseVoiceSmall · 密钥 SILICONFLOW_API_KEY',
365
+ 'deepinfraHint': 'openai/whisper-large-v3-turbo · 密钥 DEEPINFRA_API_KEY', 'fireworksHint': 'whisper-v3-turbo · 密钥 FIREWORKS_API_KEY',
366
+ 'mistralHint': 'voxtral-mini-latest · 密钥 MISTRAL_API_KEY', 'openrouterHint': 'google/gemini-2.5-flash · 密钥 OPENROUTER_API_KEY',
367
+ 'localHint': '服务器启动时配置', 'up': '上移', 'down': '下移', 'remove': '移除',
331
368
  'addProvider': '添加服务商',
332
369
  'chainHint': '自上而下为故障转移尝试顺序',
333
370
  'customName': '备用链中使用的名称',
@@ -526,6 +563,55 @@ window.__ModuleLoader__.load({
526
563
  'assistantSpeakingGated': '助手正在发言（门控模式）',
527
564
  'jargonFormatHint': ' (口语 -> 替换文本)',
528
565
  'errorPrefix': '错误',
566
+ 'wizardTitle': '首次启动与麦克风诊断向导',
567
+ 'wizardDesc': '分步诊断检查：安全上下文、麦克风权限、音频信号电平及实时 STT 回环测试。',
568
+ 'wizardBtnLaunch': '启动麦克风与 STT 诊断向导',
569
+ 'wizardClose': '关闭向导',
570
+ 'wizardStep1': '1. 音频与运行环境',
571
+ 'wizardStep2': '2. 实时信号测试',
572
+ 'wizardStep3': '3. STT 识别回环测试',
573
+ 'wizardSecureContextOk': '安全上下文 / HTTPS 已验证',
574
+ 'wizardSecureContextBad': '非安全上下文（需 HTTPS 或 localhost）',
575
+ 'wizardMediaDevicesOk': '浏览器 mediaDevices API 支持正常',
576
+ 'wizardMediaDevicesBad': '浏览器缺少 mediaDevices API 支持',
577
+ 'wizardMediaRecorderOk': 'MediaRecorder API 支持正常',
578
+ 'wizardMediaRecorderBad': '当前浏览器不支持 MediaRecorder API',
579
+ 'wizardMicPermissionOk': '麦克风权限已授予',
580
+ 'wizardMicPermissionPrompt': '尚未获得麦克风授权',
581
+ 'wizardMicPermissionBad': '麦克风权限被拒绝',
582
+ 'wizardRequestMic': '请求麦克风权限',
583
+ 'wizardMicLabel': '麦克风输入设备：',
584
+ 'wizardStartSignalCheck': '开始音频电平测试',
585
+ 'wizardStopSignalCheck': '停止音频测试',
586
+ 'wizardSignalOk': '检测到音频信号 (RMS: {level}%)',
587
+ 'wizardSignalZero': '未检测到声音信号 — 请对着麦克风说话或选择正确设备',
588
+ 'wizardSignalTesting': '正在监听声音信号… 请对着麦克风发声',
589
+ 'wizardProviderLabel': '当前语音引擎与模型：',
590
+ 'wizardProcessingArea': '处理区域：',
591
+ 'wizardAreaBrowser': '浏览器端处理 (WebGPU / WebSpeech)',
592
+ 'wizardAreaServer': '本地服务器处理 (SenseVoice / Whisper)',
593
+ 'wizardAreaCloud': '远程云端服务',
594
+ 'wizardStartTestRecord': '录制 3 秒测试语句',
595
+ 'wizardRecordingTest': '正在录制测试音频… 请说话 ({left}秒)',
596
+ 'wizardTranscribingTest': '正在请求语音识别…',
597
+ 'wizardTestResult': 'STT 测试结果：',
598
+ 'wizardTestSuccess': 'STT 链路验证成功！耗时：{latency}毫秒',
599
+ 'wizardTestEmpty': '录音中未检测到有效语音，请清晰发音后重试。',
600
+ 'wizardTestFailed': 'STT 测试失败：{error}',
601
+ 'wizardNextStepHint': '建议下一步：{hint}',
602
+ 'wizardHintPermission': '点击“请求麦克风权限”并在浏览器弹窗中允许访问。',
603
+ 'wizardHintDevice': '检查麦克风硬件连接或切换选择其他输入设备。',
604
+ 'wizardHintHttps': '请通过 HTTPS 或 localhost 访问 DeepSeek Harness。',
605
+ 'wizardHintProvider': '请在下方配置 STT 密钥凭据或一键安装 SenseVoice 本地模型。',
606
+ 'wizardFinishSuccess': '设置与验证已完成',
607
+ 'wizardSkip': '跳过向导',
608
+ 'wizardStatusCompleted': '麦克风与 STT 诊断已就绪',
609
+ 'exportDiagnosticsBtn': '导出诊断报告', 'exportingDiagnostics': '正在导出报告…',
610
+ 'diagnosticsExportSuccess': '诊断报告已下载', 'recentRecognitionsTitle': '近期识别遥测详情',
611
+ 'noRecognitionsYet': '暂无近期识别记录。', 'diagLatencyBreakdown': '耗时分布：',
612
+ 'diagVad': 'VAD：{ms}毫秒', 'diagStt': 'STT：{ms}毫秒', 'diagPolish': '后处理：{ms}毫秒',
613
+ 'diagTotal': '总耗时：{ms}毫秒', 'diagBytes': '音频大小：{bytes} 字节', 'diagWords': '词数：{count}',
614
+ 'diagExportNotice': '导出的 JSON 报告包含脱敏的耗时统计、备用链尝试、MIME 格式与错误码，不含任何原始音频或语音文本。',
529
615
  }
530
616
 
531
617
  // Strings are also needed outside components — in recording handlers and
@@ -3074,18 +3160,61 @@ function cycleDictationMode() {
3074
3160
  ),
3075
3161
  )
3076
3162
  }
3077
- // Provider Telemetry & Latency Dashboard
3163
+ // Provider Telemetry & Latency Dashboard (#266)
3078
3164
  function ProviderDashboard(props) {
3079
3165
  const t = props.t || moduleT
3080
3166
  const statsData = props.statsData || {}
3167
+ const recentDiagnostics = props.recentDiagnostics || []
3168
+ const [exporting, setExporting] = React.useState(false)
3169
+ const [exportMsg, setExportMsg] = React.useState('')
3170
+
3171
+ const exportReport = async () => {
3172
+ setExporting(true)
3173
+ setExportMsg('')
3174
+ try {
3175
+ const res = await fetch('/dsh-voice/diagnostics/export')
3176
+ const data = await res.json()
3177
+ if (!res.ok || !data || !data.ok) {
3178
+ throw new Error((data && data.error && data.error.message) || `HTTP ${res.status}`)
3179
+ }
3180
+ const blob = new Blob([JSON.stringify(data.report || {}, null, 2)], { type: 'application/json' })
3181
+ const url = URL.createObjectURL(blob)
3182
+ const a = document.createElement('a')
3183
+ a.href = url
3184
+ a.download = `dsh-voice-diagnostics-${new Date().toISOString().replace(/[:.]/g, '-')}.json`
3185
+ document.body.appendChild(a)
3186
+ a.click()
3187
+ document.body.removeChild(a)
3188
+ URL.revokeObjectURL(url)
3189
+ setExportMsg('✓ ' + t('diagnosticsExportSuccess'))
3190
+ } catch (err) {
3191
+ setExportMsg(String(err && err.message ? err.message : err))
3192
+ } finally {
3193
+ setExporting(false)
3194
+ }
3195
+ }
3196
+
3081
3197
  return React.createElement('div', { className: 'cb-section-card' },
3082
3198
  React.createElement('div', { className: 'cb-section-title' },
3083
3199
  React.createElement('span', null, '📊 ' + t('providerDashboard')),
3084
- React.createElement('button', {
3085
- type: 'button', className: 'cb-btn', style: { fontSize: '12px', padding: '4px 8px' },
3086
- onClick: props.onRefresh,
3087
- }, '🔄 ' + t('refreshStats')),
3200
+ React.createElement('div', { style: { display: 'flex', gap: '8px', alignItems: 'center' } },
3201
+ React.createElement('button', {
3202
+ type: 'button',
3203
+ className: 'cb-btn',
3204
+ style: { fontSize: '12px', padding: '4px 10px' },
3205
+ disabled: exporting,
3206
+ onClick: exportReport,
3207
+ }, exporting ? t('exportingDiagnostics') : ('📥 ' + t('exportDiagnosticsBtn'))),
3208
+ React.createElement('button', {
3209
+ type: 'button', className: 'cb-btn', style: { fontSize: '12px', padding: '4px 8px' },
3210
+ onClick: props.onRefresh,
3211
+ }, '🔄 ' + t('refreshStats')),
3212
+ ),
3088
3213
  ),
3214
+ exportMsg ? React.createElement('div', {
3215
+ className: exportMsg.startsWith('✓') ? 'dvs-ok' : 'cb-alert-err',
3216
+ style: { fontSize: '12px' },
3217
+ }, exportMsg) : null,
3089
3218
  React.createElement('div', { className: 'dvo-dash' },
3090
3219
  React.createElement('div', { className: 'dvo-dash-grid' },
3091
3220
  Object.keys(statsData).length === 0
@@ -3125,6 +3254,41 @@ function cycleDictationMode() {
3125
3254
  }),
3126
3255
  ),
3127
3256
  ),
3257
+ // Per-recognition telemetry breakdown list (#266)
3258
+ React.createElement('div', { style: { marginTop: '12px', borderTop: '1px solid var(--dsw-alias-border-l2)', paddingTop: '10px' } },
3259
+ React.createElement('div', { className: 'dvs-h', style: { marginBottom: '8px' } }, t('recentRecognitionsTitle')),
3260
+ recentDiagnostics.length === 0
3261
+ ? React.createElement('div', { className: 'dvs-sub' }, t('noRecognitionsYet'))
3262
+ : React.createElement('div', { style: { display: 'flex', flexDirection: 'column', gap: '8px' } },
3263
+ recentDiagnostics.slice(-5).reverse().map((d) => React.createElement('div', {
3264
+ key: d.id,
3265
+ className: 'dvs-card',
3266
+ style: { padding: '8px 12px', fontSize: '12px' },
3267
+ },
3268
+ React.createElement('div', { className: 'cb-row', style: { justifyContent: 'space-between', alignItems: 'center' } },
3269
+ React.createElement('div', { style: { display: 'flex', gap: '6px', alignItems: 'center' } },
3270
+ React.createElement('span', { className: 'cb-badge ' + (d.success ? 'cb-badge-ok' : 'cb-badge-err') },
3271
+ d.success ? '✓ ' + (d.provider || 'STT') : '✗ ' + (d.error ? d.error.slice(0, 30) : 'fail')
3272
+ ),
3273
+ React.createElement('span', { className: 'dvs-sub' }, `[${d.mode}/${d.subMode}]`),
3274
+ d.audioBytes ? React.createElement('span', { className: 'dvs-sub' }, t('diagBytes').replace('{bytes}', d.audioBytes)) : null,
3275
+ ),
3276
+ React.createElement('span', { style: { fontWeight: '600' } },
3277
+ t('diagTotal').replace('{ms}', d.totalTookMs)
3278
+ ),
3279
+ ),
3280
+ React.createElement('div', { className: 'cb-row', style: { gap: '12px', marginTop: '4px', fontSize: '11px', color: 'var(--dsw-alias-label-secondary)' } },
3281
+ React.createElement('span', null, t('diagVad').replace('{ms}', (d.breakdown && d.breakdown.vadMs) || 0)),
3282
+ React.createElement('span', null, t('diagStt').replace('{ms}', (d.breakdown && d.breakdown.sttMs) || 0)),
3283
+ d.breakdown && d.breakdown.polishMs ? React.createElement('span', null, t('diagPolish').replace('{ms}', d.breakdown.polishMs)) : null,
3284
+ typeof d.wordCount === 'number' ? React.createElement('span', null, t('diagWords').replace('{count}', d.wordCount)) : null,
3285
+ ),
3286
+ ))
3287
+ ),
3288
+ React.createElement('div', { className: 'dvs-sub', style: { marginTop: '8px', fontStyle: 'italic' } },
3289
+ t('diagExportNotice')
3290
+ ),
3291
+ ),
3128
3292
  )
3129
3293
  }
3130
3294
  // ------------------------------------------------------------------ 72-dictation-modes
@@ -3823,14 +3987,21 @@ function MacrosSection(props) {
3823
3987
  const whisperRunning = props.whisperRunning
3824
3988
  const providerCount = props.providerCount
3825
3989
  const onRefresh = props.onRefresh
3990
+ const onOpenWizard = props.onOpenWizard
3826
3991
 
3827
3992
  return React.createElement('div', { className: 'cb-section-card' },
3828
3993
  React.createElement('div', { className: 'cb-section-title' },
3829
3994
  React.createElement('span', null, t('statusTitle')),
3830
- React.createElement('button', {
3831
- type: 'button', className: 'cb-btn', style: { fontSize: '12px', padding: '4px 8px' },
3832
- onClick: onRefresh,
3833
- }, '🔄 ' + t('refreshStats')),
3995
+ React.createElement('div', { style: { display: 'flex', gap: '8px', alignItems: 'center' } },
3996
+ onOpenWizard ? React.createElement('button', {
3997
+ type: 'button', className: 'cb-btn cb-btn-primary', style: { fontSize: '12px', padding: '4px 10px' },
3998
+ onClick: onOpenWizard,
3999
+ }, '✨ ' + t('wizardBtnLaunch')) : null,
4000
+ React.createElement('button', {
4001
+ type: 'button', className: 'cb-btn', style: { fontSize: '12px', padding: '4px 8px' },
4002
+ onClick: onRefresh,
4003
+ }, '🔄 ' + t('refreshStats')),
4004
+ ),
3834
4005
  ),
3835
4006
  React.createElement('div', { className: 'cb-row', style: { alignItems: 'center', gap: '12px', flexWrap: 'wrap' } },
3836
4007
  React.createElement('span', { className: 'cb-badge ' + (hostOnline ? 'cb-badge-ok' : 'cb-badge-err') },
@@ -3931,6 +4102,7 @@ function MacrosSection(props) {
3931
4102
  setServerConfig(data.config)
3932
4103
  if (typeof data.revision === 'number') setRevision(data.revision)
3933
4104
  setDraft((curr) => curr || deepClone(data.config))
4105
+ if (data.config.wizardCompleted !== true) setShowWizard(true)
3934
4106
  }
3935
4107
  })
3936
4108
  .catch(() => {})
@@ -3942,6 +4114,7 @@ function MacrosSection(props) {
3942
4114
  // Hotkey picker: not a name field — press the key you want.
3943
4115
  const [catching, setCatching] = React.useState(false)
3944
4116
  const [jargonInput, setJargonInput] = React.useState(null)
4117
+ const [showWizard, setShowWizard] = React.useState(false)
3945
4118
  React.useEffect(() => {
3946
4119
  if (!catching) return undefined
3947
4120
  const onKey = (event) => {
@@ -4208,6 +4381,7 @@ function MacrosSection(props) {
4208
4381
  React.createElement('span', { className: 'dvs-sub' }, hint),
4209
4382
  )
4210
4383
  const statsData = (statusData && statusData.providerStats) || {}
4384
+ const recentDiagnostics = (statusData && statusData.recentDiagnostics) || []
4211
4385
  const hostOnline = statusLatency !== null
4212
4386
  const whisperRunning = !!(statusData && statusData.whisperRunning)
4213
4387
  const providerCount = (statusData && Array.isArray(statusData.providers)) ? statusData.providers.length : chainOptions.length
@@ -4219,7 +4393,15 @@ function MacrosSection(props) {
4219
4393
  whisperRunning: whisperRunning,
4220
4394
  providerCount: providerCount,
4221
4395
  onRefresh: fetchStatus,
4396
+ onOpenWizard: () => setShowWizard((v) => !v),
4222
4397
  }),
4398
+ showWizard ? React.createElement(FirstRunWizard, {
4399
+ t: t,
4400
+ draft: draft,
4401
+ setTop: setTop,
4402
+ writable: writable,
4403
+ onClose: () => setShowWizard(false),
4404
+ }) : null,
4223
4405
  React.createElement('div', { className: 'cb-section-card' },
4224
4406
  React.createElement('div', { className: 'cb-section-title' },
4225
4407
  React.createElement('span', null, '🎙️ ' + t('dictationBtn')),
@@ -4432,6 +4614,7 @@ function MacrosSection(props) {
4432
4614
  React.createElement(ProviderDashboard, {
4433
4615
  t: t,
4434
4616
  statsData: statsData,
4617
+ recentDiagnostics: recentDiagnostics,
4435
4618
  onRefresh: fetchStatus,
4436
4619
  }),
4437
4620
  React.createElement(UpdaterCard, {
@@ -4451,6 +4634,395 @@ function MacrosSection(props) {
4451
4634
  ),
4452
4635
  )
4453
4636
  }
4637
+ // ------------------------------------------------------- first-run wizard (#265)
4638
+ function FirstRunWizard(props) {
4639
+ const t = props.t || moduleT
4640
+ const draft = props.draft
4641
+ const setTop = props.setTop
4642
+ const writable = props.writable
4643
+ const onClose = props.onClose
4644
+
4645
+ const isSecure = typeof window !== 'undefined' ? (window.isSecureContext || window.location.protocol === 'https:' || window.location.hostname === 'localhost' || window.location.hostname === '127.0.0.1') : true
4646
+ const hasMediaDevices = typeof navigator !== 'undefined' && !!navigator.mediaDevices && typeof navigator.mediaDevices.getUserMedia === 'function'
4647
+ const hasMediaRecorder = typeof MediaRecorder !== 'undefined'
4648
+
4649
+ const [micPermission, setMicPermission] = React.useState('prompt')
4650
+ const [devices, setDevices] = React.useState([])
4651
+ const [selectedDev, setSelectedDev] = React.useState((draft && draft.micDeviceId) || '')
4652
+ const [signalTesting, setSignalTesting] = React.useState(false)
4653
+ const [currentRms, setCurrentRms] = React.useState(0)
4654
+ const [peakRms, setPeakRms] = React.useState(0)
4655
+ const [testingStt, setTestingStt] = React.useState(false)
4656
+ const [testCountdown, setTestCountdown] = React.useState(0)
4657
+ const [testResult, setTestResult] = React.useState(null)
4658
+ const [testError, setTestError] = React.useState('')
4659
+ const [testLatency, setTestLatency] = React.useState(null)
4660
+
4661
+ const signalStreamRef = React.useRef(null)
4662
+ const audioCtxRef = React.useRef(null)
4663
+ const animFrameRef = React.useRef(null)
4664
+
4665
+ React.useEffect(() => {
4666
+ if (typeof navigator !== 'undefined' && navigator.permissions && navigator.permissions.query) {
4667
+ navigator.permissions.query({ name: 'microphone' })
4668
+ .then((p) => {
4669
+ if (p) {
4670
+ setMicPermission(p.state)
4671
+ p.onchange = () => setMicPermission(p.state)
4672
+ }
4673
+ })
4674
+ .catch(() => {})
4675
+ }
4676
+ }, [])
4677
+
4678
+ const refreshDevices = () => {
4679
+ if (!navigator.mediaDevices || !navigator.mediaDevices.enumerateDevices) return
4680
+ navigator.mediaDevices.enumerateDevices()
4681
+ .then((list) => {
4682
+ const audioInputs = list.filter((d) => d.kind === 'audioinput')
4683
+ setDevices(audioInputs)
4684
+ })
4685
+ .catch(() => {})
4686
+ }
4687
+
4688
+ React.useEffect(() => {
4689
+ refreshDevices()
4690
+ }, [])
4691
+
4692
+ const stopSignalTest = () => {
4693
+ if (animFrameRef.current) {
4694
+ cancelAnimationFrame(animFrameRef.current)
4695
+ animFrameRef.current = null
4696
+ }
4697
+ if (signalStreamRef.current) {
4698
+ try { signalStreamRef.current.getTracks().forEach((trk) => trk.stop()) } catch (_) {}
4699
+ signalStreamRef.current = null
4700
+ }
4701
+ if (audioCtxRef.current) {
4702
+ try { audioCtxRef.current.close() } catch (_) {}
4703
+ audioCtxRef.current = null
4704
+ }
4705
+ setSignalTesting(false)
4706
+ setCurrentRms(0)
4707
+ }
4708
+
4709
+ React.useEffect(() => {
4710
+ return () => {
4711
+ stopSignalTest()
4712
+ }
4713
+ }, [])
4714
+
4715
+ const requestPermissionAndTestSignal = async () => {
4716
+ stopSignalTest()
4717
+ setTestError('')
4718
+ try {
4719
+ const constraints = {
4720
+ audio: selectedDev ? { deviceId: { exact: selectedDev } } : true,
4721
+ }
4722
+ const stream = await navigator.mediaDevices.getUserMedia(constraints)
4723
+ signalStreamRef.current = stream
4724
+ setMicPermission('granted')
4725
+ refreshDevices()
4726
+
4727
+ const AC = typeof AudioContext !== 'undefined' ? AudioContext
4728
+ : (typeof webkitAudioContext !== 'undefined' ? webkitAudioContext : null)
4729
+ if (!AC) return
4730
+ const actx = new AC()
4731
+ audioCtxRef.current = actx
4732
+ if (actx.state === 'suspended') { actx.resume().catch(() => {}) }
4733
+ const src = actx.createMediaStreamSource(stream)
4734
+ const analyser = actx.createAnalyser()
4735
+ analyser.fftSize = 128
4736
+ src.connect(analyser)
4737
+
4738
+ setSignalTesting(true)
4739
+ const buf = new Uint8Array(analyser.frequencyBinCount)
4740
+
4741
+ const checkLevel = () => {
4742
+ if (!signalStreamRef.current) return
4743
+ analyser.getByteFrequencyData(buf)
4744
+ let sum = 0
4745
+ for (let i = 0; i < buf.length; i++) sum += buf[i]
4746
+ const rms = Math.min(100, Math.round((sum / buf.length / 255) * 100 * 2.2))
4747
+ setCurrentRms(rms)
4748
+ setPeakRms((prev) => Math.max(prev, rms))
4749
+ animFrameRef.current = requestAnimationFrame(checkLevel)
4750
+ }
4751
+ animFrameRef.current = requestAnimationFrame(checkLevel)
4752
+ } catch (err) {
4753
+ setMicPermission('denied')
4754
+ setTestError(String(err && err.message ? err.message : err))
4755
+ stopSignalTest()
4756
+ }
4757
+ }
4758
+
4759
+ const runSttTest = async () => {
4760
+ stopSignalTest()
4761
+ setTestError('')
4762
+ setTestResult(null)
4763
+ setTestLatency(null)
4764
+ setTestingStt(true)
4765
+ setTestCountdown(3)
4766
+
4767
+ let testStream = null
4768
+ let recorder = null
4769
+ const chunks = []
4770
+
4771
+ try {
4772
+ const constraints = {
4773
+ audio: selectedDev ? { deviceId: { exact: selectedDev } } : true,
4774
+ }
4775
+ testStream = await navigator.mediaDevices.getUserMedia(constraints)
4776
+ let mime = 'audio/webm;codecs=opus'
4777
+ if (!MediaRecorder.isTypeSupported(mime)) mime = ''
4778
+ recorder = mime ? new MediaRecorder(testStream, { mimeType: mime }) : new MediaRecorder(testStream)
4779
+ recorder.ondataavailable = (e) => {
4780
+ if (e.data && e.data.size > 0) chunks.push(e.data)
4781
+ }
4782
+ recorder.start()
4783
+
4784
+ let left = 3
4785
+ const countdownTimer = setInterval(() => {
4786
+ left -= 1
4787
+ setTestCountdown(left)
4788
+ if (left <= 0) {
4789
+ clearInterval(countdownTimer)
4790
+ if (recorder.state !== 'inactive') recorder.stop()
4791
+ }
4792
+ }, 1000)
4793
+
4794
+ await new Promise((resolve) => recorder.addEventListener('stop', resolve, { once: true }))
4795
+
4796
+ try { testStream.getTracks().forEach((trk) => trk.stop()) } catch (_) {}
4797
+
4798
+ const blob = new Blob(chunks, { type: mime || 'audio/webm' })
4799
+ if (blob.size < 400) {
4800
+ setTestError(t('wizardTestEmpty'))
4801
+ setTestingStt(false)
4802
+ return
4803
+ }
4804
+
4805
+ setTestCountdown(0)
4806
+ const startMs = Date.now()
4807
+ const dataBase64 = await blobToBase64(blob)
4808
+ const payload = {
4809
+ dataBase64,
4810
+ mimeType: mime || recorder.mimeType || 'audio/webm',
4811
+ mode: 'dictation',
4812
+ subMode: 'clean',
4813
+ }
4814
+
4815
+ const res = await fetch('/dsh-voice/transcribe', {
4816
+ method: 'POST',
4817
+ headers: { 'content-type': 'application/json' },
4818
+ body: JSON.stringify(payload),
4819
+ })
4820
+ const data = await res.json()
4821
+ const latency = Date.now() - startMs
4822
+ setTestLatency(latency)
4823
+
4824
+ if (!res.ok || !data || !data.ok) {
4825
+ const msg = (data && data.error && data.error.message) || `HTTP ${res.status}`
4826
+ setTestError(msg)
4827
+ } else {
4828
+ const recognized = String(data.text || '').trim()
4829
+ if (recognized) {
4830
+ setTestResult(recognized)
4831
+ if (typeof setTop === 'function') {
4832
+ setTop('wizardCompleted', true)
4833
+ }
4834
+ } else {
4835
+ setTestError(t('wizardTestEmpty'))
4836
+ }
4837
+ }
4838
+ } catch (err) {
4839
+ setTestError(String(err && err.message ? err.message : err))
4840
+ } finally {
4841
+ if (testStream) {
4842
+ try { testStream.getTracks().forEach((trk) => trk.stop()) } catch (_) {}
4843
+ }
4844
+ setTestingStt(false)
4845
+ }
4846
+ }
4847
+
4848
+ // Determine active chain info
4849
+ const dictationChain = (draft && draft.dictation && draft.dictation.chain) || []
4850
+ const activeProvider = (dictationChain[0] && (dictationChain[0].provider || dictationChain[0])) || 'sensevoice'
4851
+ let areaLabel = t('wizardAreaServer')
4852
+ if (activeProvider === 'browser' || activeProvider === 'browser-webgpu') {
4853
+ areaLabel = t('wizardAreaBrowser')
4854
+ } else if (activeProvider !== 'sensevoice' && activeProvider !== 'local-whisper') {
4855
+ areaLabel = t('wizardAreaCloud')
4856
+ }
4857
+
4858
+ let nextStepHint = ''
4859
+ if (!isSecure) {
4860
+ nextStepHint = t('wizardHintHttps')
4861
+ } else if (micPermission === 'denied') {
4862
+ nextStepHint = t('wizardHintPermission')
4863
+ } else if (peakRms === 0 && signalTesting) {
4864
+ nextStepHint = t('wizardHintDevice')
4865
+ } else if (testError) {
4866
+ nextStepHint = t('wizardHintProvider')
4867
+ }
4868
+
4869
+ return React.createElement('div', { className: 'cb-section-card', style: { borderColor: 'var(--dsw-alias-state-brand-primary)' } },
4870
+ React.createElement('div', { className: 'cb-section-title' },
4871
+ React.createElement('span', null, '✨ ' + t('wizardTitle')),
4872
+ React.createElement('div', { style: { display: 'flex', gap: '8px' } },
4873
+ testResult ? React.createElement('span', { className: 'cb-badge cb-badge-ok' }, '✓ ' + t('wizardStatusCompleted')) : null,
4874
+ onClose ? React.createElement('button', {
4875
+ type: 'button', className: 'cb-btn cb-btn-mini', onClick: onClose, title: t('wizardClose'),
4876
+ }, '✕') : null,
4877
+ ),
4878
+ ),
4879
+ React.createElement('div', { className: 'cb-section-desc' }, t('wizardDesc')),
4880
+
4881
+ // Step 1: Environment & Permissions
4882
+ React.createElement('div', { className: 'dvs-card' },
4883
+ React.createElement('div', { className: 'dvs-h' }, t('wizardStep1')),
4884
+ React.createElement('div', { className: 'cb-row' },
4885
+ React.createElement('span', { className: 'cb-badge ' + (isSecure ? 'cb-badge-ok' : 'cb-badge-err') },
4886
+ isSecure ? ('✓ ' + t('wizardSecureContextOk')) : ('✗ ' + t('wizardSecureContextBad'))
4887
+ ),
4888
+ React.createElement('span', { className: 'cb-badge ' + (hasMediaDevices ? 'cb-badge-ok' : 'cb-badge-err') },
4889
+ hasMediaDevices ? ('✓ ' + t('wizardMediaDevicesOk')) : ('✗ ' + t('wizardMediaDevicesBad'))
4890
+ ),
4891
+ React.createElement('span', { className: 'cb-badge ' + (hasMediaRecorder ? 'cb-badge-ok' : 'cb-badge-err') },
4892
+ hasMediaRecorder ? ('✓ ' + t('wizardMediaRecorderOk')) : ('✗ ' + t('wizardMediaRecorderBad'))
4893
+ ),
4894
+ React.createElement('span', { className: 'cb-badge ' + (micPermission === 'granted' ? 'cb-badge-ok' : (micPermission === 'denied' ? 'cb-badge-err' : 'cb-badge-warn')) },
4895
+ micPermission === 'granted' ? ('✓ ' + t('wizardMicPermissionOk')) : (micPermission === 'denied' ? ('✗ ' + t('wizardMicPermissionBad')) : ('· ' + t('wizardMicPermissionPrompt')))
4896
+ ),
4897
+ ),
4898
+ React.createElement('div', { className: 'cb-field', style: { marginTop: '8px' } },
4899
+ React.createElement('label', null, t('wizardMicLabel')),
4900
+ React.createElement('div', { className: 'cb-row' },
4901
+ React.createElement('select', {
4902
+ value: selectedDev,
4903
+ style: { flex: 1, minWidth: '240px' },
4904
+ onChange: (e) => {
4905
+ setSelectedDev(e.target.value)
4906
+ if (typeof setTop === 'function') setTop('micDeviceId', e.target.value)
4907
+ stopSignalTest()
4908
+ },
4909
+ },
4910
+ React.createElement('option', { value: '' }, 'Default System Microphone'),
4911
+ devices.map((d) => React.createElement('option', { key: d.deviceId, value: d.deviceId }, d.label || `Microphone ${d.deviceId.slice(0, 8)}…`))
4912
+ ),
4913
+ React.createElement('button', {
4914
+ type: 'button',
4915
+ className: 'cb-btn',
4916
+ onClick: requestPermissionAndTestSignal,
4917
+ }, t('wizardRequestMic')),
4918
+ ),
4919
+ ),
4920
+ ),
4921
+
4922
+ // Step 2: Live Audio Signal Test
4923
+ React.createElement('div', { className: 'dvs-card' },
4924
+ React.createElement('div', { className: 'dvs-h' }, t('wizardStep2')),
4925
+ React.createElement('div', { className: 'cb-row', style: { alignItems: 'center' } },
4926
+ React.createElement('button', {
4927
+ type: 'button',
4928
+ className: 'cb-btn ' + (signalTesting ? 'cb-btn-danger' : 'cb-btn-primary'),
4929
+ onClick: signalTesting ? stopSignalTest : requestPermissionAndTestSignal,
4930
+ }, signalTesting ? t('wizardStopSignalCheck') : t('wizardStartSignalCheck')),
4931
+ signalTesting ? React.createElement('span', { className: 'dvs-sub' }, t('wizardSignalTesting')) : null,
4932
+ ),
4933
+ signalTesting ? React.createElement('div', { style: { marginTop: '8px', display: 'flex', flexDirection: 'column', gap: '6px' } },
4934
+ React.createElement('div', {
4935
+ style: {
4936
+ height: '14px',
4937
+ borderRadius: '7px',
4938
+ background: 'var(--dsw-alias-bg-layer-1)',
4939
+ border: '1px solid var(--dsw-alias-border-l1)',
4940
+ overflow: 'hidden',
4941
+ position: 'relative',
4942
+ },
4943
+ },
4944
+ React.createElement('div', {
4945
+ style: {
4946
+ width: `${currentRms}%`,
4947
+ height: '100%',
4948
+ background: currentRms > 5 ? 'var(--dsw-alias-state-success-primary)' : 'var(--dsw-alias-label-dimmed)',
4949
+ transition: 'width 0.05s linear',
4950
+ },
4951
+ }),
4952
+ ),
4953
+ React.createElement('span', { className: currentRms > 5 ? 'dvs-ok' : 'dvs-sub' },
4954
+ currentRms > 5
4955
+ ? t('wizardSignalOk').replace('{level}', currentRms)
4956
+ : t('wizardSignalZero')
4957
+ ),
4958
+ ) : null,
4959
+ ),
4960
+
4961
+ // Step 3: Transcription Round-Trip Test
4962
+ React.createElement('div', { className: 'dvs-card' },
4963
+ React.createElement('div', { className: 'dvs-h' }, t('wizardStep3')),
4964
+ React.createElement('div', { className: 'cb-row' },
4965
+ React.createElement('span', { className: 'dvs-sub' }, `${t('wizardProviderLabel')} ${activeProvider}`),
4966
+ React.createElement('span', { className: 'cb-badge' }, `${t('wizardProcessingArea')} ${areaLabel}`),
4967
+ ),
4968
+ React.createElement('div', { className: 'cb-row', style: { marginTop: '8px' } },
4969
+ React.createElement('button', {
4970
+ type: 'button',
4971
+ className: 'cb-btn cb-btn-primary',
4972
+ disabled: testingStt || micPermission === 'denied',
4973
+ onClick: runSttTest,
4974
+ }, testingStt
4975
+ ? (testCountdown > 0 ? t('wizardRecordingTest').replace('{left}', testCountdown) : t('wizardTranscribingTest'))
4976
+ : t('wizardStartTestRecord')
4977
+ ),
4978
+ ),
4979
+ testResult ? React.createElement('div', { style: { marginTop: '8px', display: 'flex', flexDirection: 'column', gap: '4px' } },
4980
+ React.createElement('div', { className: 'dvs-ok' },
4981
+ '✓ ' + t('wizardTestSuccess').replace('{latency}', testLatency || 0)
4982
+ ),
4983
+ React.createElement('blockquote', {
4984
+ style: {
4985
+ margin: '4px 0 0 0',
4986
+ padding: '8px 12px',
4987
+ borderRadius: '6px',
4988
+ background: 'var(--dsw-alias-bg-layer-3)',
4989
+ borderLeft: '4px solid var(--dsw-alias-state-success-primary)',
4990
+ color: 'var(--dsw-alias-label-primary)',
4991
+ fontStyle: 'italic',
4992
+ },
4993
+ }, `“${testResult}”`),
4994
+ ) : null,
4995
+ testError ? React.createElement('div', { className: 'cb-alert-err', style: { marginTop: '8px' } },
4996
+ t('wizardTestFailed').replace('{error}', testError)
4997
+ ) : null,
4998
+ ),
4999
+
5000
+ // Next Step Hint / Feedback
5001
+ nextStepHint ? React.createElement('div', { className: 'dvs-wait', style: { fontStyle: 'italic' } },
5002
+ t('wizardNextStepHint').replace('{hint}', nextStepHint)
5003
+ ) : null,
5004
+
5005
+ // Wizard Bottom Actions
5006
+ React.createElement('div', { className: 'cb-row', style: { justifyContent: 'space-between', marginTop: '6px' } },
5007
+ React.createElement('button', {
5008
+ type: 'button',
5009
+ className: 'cb-btn',
5010
+ onClick: () => {
5011
+ if (typeof setTop === 'function') setTop('wizardCompleted', true)
5012
+ if (onClose) onClose()
5013
+ },
5014
+ }, t('wizardSkip')),
5015
+ testResult ? React.createElement('button', {
5016
+ type: 'button',
5017
+ className: 'cb-btn cb-btn-primary',
5018
+ onClick: () => {
5019
+ if (typeof setTop === 'function') setTop('wizardCompleted', true)
5020
+ if (onClose) onClose()
5021
+ },
5022
+ }, '✓ ' + t('wizardFinishSuccess')) : null,
5023
+ ),
5024
+ )
5025
+ }
4454
5026
  // Card in Settings → Plugins → Plugin settings (#18)
4455
5027
  function PluginCard(props) {
4456
5028
  const page = !!(props && props.view === 'page')
package/lib/host-routes.js CHANGED
@@ -96,6 +96,7 @@ export function registerHostRoutes(ctx, deps) {
96
96
  dictationPrompts: cfg.dictationPrompts || {},
97
97
  voiceMacros: Array.isArray(cfg.voiceMacros) ? cfg.voiceMacros : [],
98
98
  providerStats: statsTracker.get(),
99
+ recentDiagnostics: typeof statsTracker.getDiagnostics === 'function' ? statsTracker.getDiagnostics() : [],
99
100
  })
100
101
  },
101
102
  }), 'dsh-voice: /status route')
@@ -133,6 +134,19 @@ export function registerHostRoutes(ctx, deps) {
133
134
  },
134
135
  }), 'dsh-voice: /polish route')
135
136
 
137
+ ctx.effect(() => ctx.webServer.register({
138
+ kind: 'exact',
139
+ path: '/dsh-voice/diagnostics/export',
140
+ handler: async (req, res) => {
141
+ if (req.method !== 'GET') { writeJson(res, 405, { ok: false, error: { code: 'method', message: 'GET only' } }); return }
142
+ if (!isTrustedCaller(req)) { writeJson(res, 403, { ok: false, error: { code: 'forbidden', message: 'forbidden source origin' } }); return }
143
+ const report = typeof statsTracker.exportAnonymizedReport === 'function'
144
+ ? statsTracker.exportAnonymizedReport({ version: '0.9.34', environment: 'production' })
145
+ : { summary: statsTracker.get(), recentDiagnostics: [] }
146
+ writeJson(res, 200, { ok: true, report })
147
+ },
148
+ }), 'dsh-voice: /diagnostics/export route')
149
+
136
150
  ctx.effect(() => ctx.webServer.register({
137
151
  kind: 'exact',
138
152
  path: '/dsh-voice/transcribe',
@@ -178,13 +192,31 @@ export function registerHostRoutes(ctx, deps) {
178
192
  await startSensevoice()
179
193
  }
180
194
 
195
+ const reqStart = Date.now()
181
196
  const controller = new AbortController()
182
197
  const timer = setTimeout(() => controller.abort(), cfg.timeoutMs)
198
+ let out = null
199
+ let polishTookMs = 0
183
200
  try {
184
- const out = await transcribe(modeCfg, bytes, mime, controller.signal, payload.contextWords)
201
+ out = await transcribe(modeCfg, bytes, mime, controller.signal, payload.contextWords)
185
202
  if (payload.mode === 'message' && modeCfg.sessionCommands === true) {
186
203
  const cmd = sessionCommand(out.text)
187
- if (cmd) { writeJson(res, 200, { ok: true, command: cmd, provider: out.provider, tookMs: out.tookMs }); return }
204
+ if (cmd) {
205
+ if (typeof statsTracker.recordDiagnostic === 'function') {
206
+ statsTracker.recordDiagnostic({
207
+ mode: payload.mode || 'dictation',
208
+ subMode: 'command',
209
+ mimeType: mime,
210
+ audioBytes: bytes.length,
211
+ provider: out.provider,
212
+ success: true,
213
+ totalTookMs: Date.now() - reqStart,
214
+ sttMs: out.tookMs || 0,
215
+ vadMs: Number(payload.vadMs) || 0,
216
+ })
217
+ }
218
+ writeJson(res, 200, { ok: true, command: cmd, provider: out.provider, tookMs: out.tookMs }); return
219
+ }
188
220
  }
189
221
  const subMode = payload.subMode || payload.dictationMode || cfg.dictationMode || 'clean'
190
222
  const rawText = out.text
@@ -192,7 +224,9 @@ export function registerHostRoutes(ctx, deps) {
192
224
 
193
225
  // Dictation modes & polish (#264): verbatim skips LLM entirely (0 LLM calls)
194
226
  if (subMode !== 'verbatim' && (modeCfg.polish === true || payload.polish === true)) {
227
+ const polishStart = Date.now()
195
228
  text = await polishText(rawText, { ...modeCfg, subMode, prompt: payload.prompt }, controller.signal)
229
+ polishTookMs = Date.now() - polishStart
196
230
  }
197
231
 
198
232
  // Voice action macros (#167)
@@ -211,8 +245,37 @@ export function registerHostRoutes(ctx, deps) {
211
245
  action = actionRes.action
212
246
  }
213
247
  }
214
- writeJson(res, 200, { ok: true, text, rawText, subMode, action, provider: out.provider, tookMs: out.tookMs })
248
+ const totalTookMs = Date.now() - reqStart
249
+ if (typeof statsTracker.recordDiagnostic === 'function') {
250
+ statsTracker.recordDiagnostic({
251
+ mode: payload.mode || 'dictation',
252
+ subMode,
253
+ mimeType: mime,
254
+ audioBytes: bytes.length,
255
+ provider: out.provider,
256
+ success: true,
257
+ totalTookMs,
258
+ sttMs: out.tookMs || 0,
259
+ polishMs: polishTookMs,
260
+ vadMs: Number(payload.vadMs) || 0,
261
+ wordCount: text ? text.split(/\s+/).filter(Boolean).length : 0,
262
+ })
263
+ }
264
+ writeJson(res, 200, { ok: true, text, rawText, subMode, action, provider: out.provider, tookMs: out.tookMs, polishTookMs, totalTookMs })
215
265
  } catch (e) {
266
+ if (typeof statsTracker.recordDiagnostic === 'function') {
267
+ statsTracker.recordDiagnostic({
268
+ mode: payload.mode || 'dictation',
269
+ subMode: payload.subMode || 'clean',
270
+ mimeType: mime,
271
+ audioBytes: bytes.length,
272
+ provider: (out && out.provider) || 'unknown',
273
+ success: false,
274
+ totalTookMs: Date.now() - reqStart,
275
+ vadMs: Number(payload.vadMs) || 0,
276
+ error: String(e && e.message || e),
277
+ })
278
+ }
216
279
  writeJson(res, 502, { ok: false, error: { code: 'chain', message: String(e && e.message || e) } })
217
280
  } finally {
218
281
  clearTimeout(timer)
package/lib/stats.js CHANGED
@@ -3,6 +3,55 @@
3
3
 
4
4
  export function createStatsTracker() {
5
5
  const stats = new Map()
6
+ const diagnostics = []
7
+ const MAX_DIAGNOSTICS = 50
8
+
9
+ function recordDiagnostic(entry = {}) {
10
+ diagnostics.push({
11
+ id: Math.random().toString(36).slice(2, 10),
12
+ timestamp: Date.now(),
13
+ mode: entry.mode || 'dictation',
14
+ subMode: entry.subMode || 'clean',
15
+ mimeType: entry.mimeType || 'audio/webm',
16
+ audioBytes: Number(entry.audioBytes) || 0,
17
+ provider: entry.provider || 'unknown',
18
+ success: !!entry.success,
19
+ totalTookMs: Number(entry.totalTookMs) || 0,
20
+ breakdown: {
21
+ vadMs: Number(entry.vadMs) || 0,
22
+ sttMs: Number(entry.sttMs) || 0,
23
+ polishMs: Number(entry.polishMs) || 0,
24
+ },
25
+ chainAttempts: Array.isArray(entry.chainAttempts) ? entry.chainAttempts.map((a) => ({
26
+ provider: a.provider || 'unknown',
27
+ ok: !!a.ok,
28
+ tookMs: Number(a.tookMs) || 0,
29
+ reason: a.reason ? String(a.reason).slice(0, 150) : undefined,
30
+ })) : [],
31
+ error: entry.error ? String(entry.error).slice(0, 200) : undefined,
32
+ wordCount: typeof entry.wordCount === 'number' ? entry.wordCount : undefined,
33
+ })
34
+ if (diagnostics.length > MAX_DIAGNOSTICS) {
35
+ diagnostics.shift()
36
+ }
37
+ }
38
+
39
+ function getDiagnostics() {
40
+ return diagnostics.slice()
41
+ }
42
+
43
+ function exportAnonymizedReport(appInfo = {}) {
44
+ return {
45
+ exportedAt: new Date().toISOString(),
46
+ app: {
47
+ name: '@goodandready/dsh-voice',
48
+ version: appInfo.version || '0.9.34',
49
+ environment: appInfo.environment || 'production',
50
+ },
51
+ summary: get(),
52
+ recentDiagnostics: getDiagnostics(),
53
+ }
54
+ }
6
55
 
7
56
  function record(key, res = {}) {
8
57
  if (!key) return
@@ -51,9 +100,10 @@ export function createStatsTracker() {
51
100
 
52
101
  function clear() {
53
102
  stats.clear()
103
+ diagnostics.length = 0
54
104
  }
55
105
 
56
- return { record, get, clear }
106
+ return { record, get, clear, recordDiagnostic, getDiagnostics, exportAnonymizedReport }
57
107
  }
58
108
 
59
109
  export function mergeContextVocabulary(baseVocab = [], contextWords = [], maxWords = 50) {
package/lib/updater.js CHANGED
@@ -18,15 +18,82 @@ function header(request, name) {
18
18
  return Array.isArray(value) ? value[0] : value;
19
19
  }
20
20
 
21
- function isPrivateLan(value) {
22
- const address = value?.toLowerCase().replace(/^\[|\]$/g, '');
21
+ function parseIPv4(ip) {
22
+ const parts = ip.split('.');
23
+ if (parts.length !== 4) return null;
24
+ let num = 0;
25
+ for (const p of parts) {
26
+ if (!/^\d+$/.test(p)) return null;
27
+ const n = Number(p);
28
+ if (n < 0 || n > 255) return null;
29
+ num = (num << 8) | n;
30
+ }
31
+ return (num >>> 0);
32
+ }
33
+
34
+ function parseIPv6(ip) {
35
+ const str = ip.toLowerCase();
36
+ if (str.includes('.')) {
37
+ const lastColon = str.lastIndexOf(':');
38
+ if (lastColon === -1) return null;
39
+ const v4Part = str.slice(lastColon + 1);
40
+ const v4Num = parseIPv4(v4Part);
41
+ if (v4Num === null) return null;
42
+ const v6Prefix = str.slice(0, lastColon);
43
+ const hiHex = ((v4Num >>> 16) & 0xffff).toString(16);
44
+ const loHex = (v4Num & 0xffff).toString(16);
45
+ return parseIPv6(v6Prefix + ':' + hiHex + ':' + loHex);
46
+ }
47
+ const parts = str.split('::');
48
+ if (parts.length > 2) return null;
49
+ let groups = [];
50
+ if (parts.length === 2) {
51
+ const left = parts[0] ? parts[0].split(':') : [];
52
+ const right = parts[1] ? parts[1].split(':') : [];
53
+ const missing = 8 - (left.length + right.length);
54
+ if (missing < 1) return null;
55
+ groups = [...left, ...Array(missing).fill('0'), ...right];
56
+ } else {
57
+ groups = str.split(':');
58
+ }
59
+ if (groups.length !== 8) return null;
60
+ let hi = 0n;
61
+ for (let i = 0; i < 8; i++) {
62
+ if (!/^[0-9a-f]{1,4}$/.test(groups[i])) return null;
63
+ hi = (hi << 16n) | BigInt(parseInt(groups[i], 16));
64
+ }
65
+ return hi;
66
+ }
67
+
68
+ export function isPrivateLan(value) {
69
+ const address = value?.toLowerCase().replace(/^\[|\]$/g, '').trim();
23
70
  if (!address) return false;
24
- const ipv4 = address.startsWith('::ffff:') ? address.slice(7) : address;
25
- return (
26
- ipv4.startsWith('192.168.') ||
27
- ipv4.startsWith('10.') ||
28
- /^172\.(1[6-9]|2[0-9]|3[0-1])\./.test(ipv4)
29
- );
71
+ const slashIdx = address.indexOf('/');
72
+ const ipPart = slashIdx !== -1 ? address.slice(0, slashIdx) : address;
73
+
74
+ const ipv4 = ipPart.startsWith('::ffff:') ? ipPart.slice(7) : ipPart;
75
+
76
+ const v4Num = parseIPv4(ipv4);
77
+ if (v4Num !== null) {
78
+ if ((v4Num >>> 24) === 10) return true;
79
+ if ((v4Num >>> 20) === ((172 << 4) | 1)) return true;
80
+ if ((v4Num >>> 16) === ((192 << 8) | 168)) return true;
81
+ if ((v4Num >>> 16) === ((169 << 8) | 254)) return true;
82
+ return false;
83
+ }
84
+
85
+ const v6Num = parseIPv6(ipPart);
86
+ if (v6Num !== null) {
87
+ // fc00::/7 ULA (covers fc00::/8 and fd00::/8)
88
+ const first7Bits = Number(v6Num >> 121n);
89
+ if (first7Bits === 0b1111110) return true;
90
+ // fe80::/10 Link-Local
91
+ const first10Bits = Number(v6Num >> 118n);
92
+ if (first10Bits === 0b1111111010) return true;
93
+ return false;
94
+ }
95
+
96
+ return false;
30
97
  }
31
98
 
32
99
  function isLoopback(value) {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@goodandready/dsh-voice",
3
- "version": "0.9.32",
3
+ "version": "0.9.34",
4
4
  "description": "Voice input for DeepSeek Harness: dictation chunked by pauses and voice messages, each with its own provider fallback chain (Deepgram, Groq, HuggingFace, local whisper.cpp, plus any OpenAI-compatible API of your own).",
5
5
  "license": "MIT",
6
6
  "type": "module",