@goodandready/dsh-voice 0.8.25 → 0.8.27

package/README.md

@@ -21,6 +21,16 @@
   <a href="README.zh.md"><b>🇨🇳 中文说明</b></a>
 </p>
 
+<table align="center">
+  <tr>
+    <td align="center">
+      ⭐ <strong>If you like this plugin, please star it on GitHub</strong> — it shows me that the plugin is useful to you and motivates me to keep developing it.
+      <br><br>
+      🐛 <strong>If you find a bug or would like to request a feature</strong>, open a GitHub issue in any language — I will review your proposal and implement useful suggestions in a future plugin version.
+    </td>
+  </tr>
+</table>
+
 </div>
 
 ---

package/README.ru.md

@@ -21,6 +21,16 @@
   <a href="README.zh.md"><b>🇨🇳 中文说明</b></a>
 </p>
 
+<table align="center">
+  <tr>
+    <td align="center">
+      ⭐ <strong>Если вам нравится этот плагин, поставьте ему звезду на GitHub</strong> — это покажет мне, что плагин вам полезен, и будет мотивировать меня развивать его дальше.
+      <br><br>
+      🐛 <strong>Если вы нашли баг или хотите предложить новый функционал</strong>, создайте issue на GitHub на любом языке — я рассмотрю ваше предложение и реализую полезные идеи в одной из следующих версий плагина.
+    </td>
+  </tr>
+</table>
+
 </div>
 
 ---

package/README.zh.md

@@ -21,6 +21,16 @@
   <a href="README.zh.md"><b>🇨🇳 中文说明</b></a>
 </p>
 
+<table align="center">
+  <tr>
+    <td align="center">
+      ⭐ <strong>如果您喜欢这个插件，请在 GitHub 上为它点亮 Star</strong> — 这能让我知道插件对您有用，并鼓励我继续开发和维护它。
+      <br><br>
+      🐛 <strong>如果您发现 Bug 或希望增加功能</strong>，请使用任意语言在 GitHub 上提交 Issue — 我会评估您的建议，并在后续版本中实现有价值的改进。
+    </td>
+  </tr>
+</table>
+
 </div>
 
 ---

package/lib/client.js

@@ -132,7 +132,7 @@ window.__ModuleLoader__.load({
       'whisperAutostart': 'Autostart the local whisper',
       'save': 'Save',
       'saved': 'Saved ✓',
-      'openrouterWarning': 'OpenRouter has no /audio/transcriptions \u2014 use the openai-chat-audio template there',
+      'openrouterWarning': 'OpenRouter has no /audio/transcriptions — use the openai-chat-audio template there',
       'noiseSuppression': 'Hardware noise suppression',
       'noiseSuppressionHint': 'Enable browser noise suppression, echo cancellation, and auto gain control',
       'contextGlossary': 'Context glossary injection',
@@ -180,13 +180,178 @@ window.__ModuleLoader__.load({
       'testMic': 'Test Mic',
       'stopTest': 'Stop Test',
       'micLevel': 'Mic level',
+      'noiseGate': 'Noise Gate (dB)',
+      'noiseGateHint': 'Mute background fan hum and keyboard clicks below threshold. Off = disabled',
+      'noiseGateOff': 'Off',
+    }
+
+    const zh = {
+      'saveFailed': '部分字段未保存 —',
+      'cardHint': '听写与语音消息：服务商、备用链、本地 whisper',
+      'expand': '展开',
+      'collapse': '收起',
+      'composerUnavailable': '输入框不可用',
+      'keySpace': '空格',
+      'keyUnset': '未设置',
+      'recognitionError': '识别失败',
+      'micUnavailable': '麦克风不可用：需要 HTTPS 或 localhost',
+      'noRecorder': '当前浏览器不支持 MediaRecorder',
+      'browserFailed': '浏览器未能识别：',
+      'nothingHeard': '未识别到任何内容',
+      'dictationBtn': '语音输入',
+      'messageBtn': '语音消息 — 点击或按住',
+      'dictationPill': '听写 — 文本将追加至输入框',
+      'messagePill': '正在录制语音消息',
+      'cancel': '取消',
+      'holdHint': '按住录音 — 松开即发送',
+      'listening': '浏览器正在倾听…',
+      'stop': '停止',
+      'transcribing': '正在转写…',
+      'sendingIn': '发送至智能体倒计时',
+      'secondsShort': ' 秒',
+      'keepPending': '不发送',
+      'hide': '隐藏',
+      'title': '语音',
+      'recordSlot': '语音录制',
+      'browserHint': '浏览器直接识别，无需密钥',
+      'openaiHint': 'whisper-1 · 密钥 OPENAI_API_KEY',
+      'siliconflowHint': 'FunAudioLLM/SenseVoiceSmall · 密钥 SILICONFLOW_API_KEY',
+      'deepinfraHint': 'openai/whisper-large-v3-turbo · 密钥 DEEPINFRA_API_KEY',
+      'fireworksHint': 'whisper-v3-turbo · 密钥 FIREWORKS_API_KEY',
+      'mistralHint': 'voxtral-mini-latest · 密钥 MISTRAL_API_KEY',
+      'openrouterHint': 'google/gemini-2.5-flash · 密钥 OPENROUTER_API_KEY',
+      'localHint': '服务器启动时配置',
+      'up': '上移',
+      'down': '下移',
+      'remove': '移除',
+      'addProvider': '添加服务商',
+      'chainHint': '自上而下为故障转移尝试顺序',
+      'customName': '备用链中使用的名称',
+      'customModel': '模型',
+      'customKeyName': '密钥名称',
+      'customModelHint': '请求的模型（留空使用内置默认值）',
+      'addCustom': '添加自定义服务商',
+      'loadingSettings': '正在加载设置…',
+      'notReady1': 'Harness 尚未同步此插件的设置。如果刚重启，',
+      'notReady2': '设置项将在数秒内自动呈现。',
+      'hotkey': '语音消息快捷键',
+      'pressKey': '请按下按键…',
+      'clearKey': '清除按键',
+      'hotkeyHint1': '按住录音，松开发送给智能体，按 Esc 取消。',
+      'hotkeyHint2': '支持任意按键：字母键、功能键 (F1-F12) 或修饰键。',
+      'language': '识别语言',
+      'dictationHint': '语音按停顿自动切句，文字实时追加至输入框。',
+      'pauseMs': '断句静音停顿时长（毫秒）',
+      'pauseHint': '数值越小切分越频繁、出字越快，但可能切断未说完的词',
+      'speaking': '正在说话…',
+      'silence': '停顿…',
+      'normalizeTranscript': '规范化转写文本',
+      'undo': '撤销上次插入',
+      'undone': '插入已撤销',
+      'nothingToUndo': '没有可撤销的内容',
+      'beep': '启停提示音',
+      'localOnly': '仅使用本地 whisper',
+      'localOnlyHint': '将双链限制为仅使用本地 whisper.cpp 服务：纯离线运行。',
+      'sendDelay': '听写延迟追加 (ms)',
+      'sendDelayHint': '追加听写文本前的缓冲等待时间，允许撤回。0 为关闭',
+      'mic': '麦克风设备',
+      'micDefault': '系统默认',
+      'vocabulary': '自定义词汇表（每行一个词）',
+      'polish': '模型润色转写文本',
+      'polishHint': '插入前通过模型修正标点符号并消除语气口头禅',
+      'stream': '连续流式听写',
+      'streamHint': '说话期间按定时器切分短句，无需等待长时间停顿',
+      'streamChunkMs': '流式切片时长 (ms)',
+      'vadAdapt': '自适应静音阈值',
+      'vadAdaptHint': '根据语速动态调整停顿静音阈值。0 为固定阈值',
+      'wakeWord': '语音唤醒词',
+      'wakeWordHint': '说话以此短语开头时自动触发录音。留空为关闭',
+      'bargeIn': '语音插话打断 (Barge-in)',
+      'polishSend': '发送前润色整篇草稿',
+      'polishSendHint': '发送给智能体前对整个输入框草稿进行一次整体润色',
+      'sessionCommands': '会话语音控制指令',
+      'sessionCommandsHint': '"send"、"cancel"、"stop"、"continue" 执行操作而非转为文本',
+      'polishBaseUrl': '离线润色服务端点',
+      'polishBaseUrlHint': '兼容 OpenAI /chat/completions 的基础地址（如本地 Ollama）。留空使用 harness 模型',
+      'polishModel': '离线润色模型',
+      'polishKeyEnv': '离线润色密钥凭据名',
+      'voiceCommandsLabel': '语音编辑指令（"换行"、"段落"）',
+      'normalizeTranscriptHint': 'transcribe_audio：口语数字转阿拉伯数字，整理标点符号',
+      'messageTitle': '语音消息',
+      'messageHint': '录制单条完整语音，转写完成后自动发送给智能体。',
+      'undoMs': '撤回窗口时长（毫秒）',
+      'undoHint': '转写完成后多久以内可以随时取消发送',
+      'customTitle': '自定义语音服务商',
+      'customHint': '兼容 OpenAI 的任意音频 API。配置名称将出现在上方备用链中。',
+      'general': '通用设置',
+      'whisperEndpoint': '本地 whisper：端点地址',
+      'whisperEndpointHint': 'whisper.cpp 服务器的 POST /inference 接口',
+      'deepgramEndpoint': 'Deepgram：基础地址',
+      'deepgramEndpointHint': 'Deepgram 或私有化部署的基础地址（默认 https://api.deepgram.com）',
+      'whisperBin': '本地 whisper：可执行文件',
+      'whisperBinHint': '开启自动启动时使用',
+      'whisperModel': '本地 whisper：模型文件',
+      'whisperModelHint': 'ggml 模型文件的绝对路径',
+      'whisperAutostart': '自动启动本地 whisper',
+      'save': '保存',
+      'saved': '已保存 ✓',
+      'openrouterWarning': 'OpenRouter 无 /audio/transcriptions 端点 — 请选用 openai-chat-audio 模板',
+      'noiseSuppression': '硬件级降噪',
+      'noiseSuppressionHint': '启用浏览器硬件降噪、回声消除与自动增益控制',
+      'contextGlossary': '上下文专业词汇注入',
+      'contextGlossaryHint': '自动从输入草稿提取代码变量名与术语以提升语音识别准确率',
+      'providerDashboard': '服务商延迟与健康监控',
+      'avgLatency': '平均延迟',
+      'successRate': '成功率',
+      'fast': '极速',
+      'normal': '正常',
+      'slow': '较慢',
+      'error': '异常',
+      'idle': '无调用记录',
+      'play': '播放',
+      'pause': '暂停',
+      'listenBack': '试听回放',
+      'lastRecording': '最近一条语音记录',
+      'sensevoiceHint': 'SenseVoice-ONNX / Sherpa-ONNX · 极速本地语音识别 (~50ms)',
+      'sensevoiceEndpoint': 'SenseVoice：端点地址',
+      'sensevoiceEndpointHint': 'sherpa-onnx 或兼容服务器的 POST 端点',
+      'sensevoiceBin': 'SenseVoice：可执行文件',
+      'sensevoiceBinHint': '开启自动启动时使用',
+      'sensevoiceModel': 'SenseVoice：模型路径',
+      'sensevoiceModelHint': 'SenseVoice / sherpa-onnx 模型的绝对路径或标识符',
+      'sensevoiceAutostart': '自动启动 SenseVoice 服务',
+      'visualizerStyle': '音频动态波形样式',
+      'visualizerStyleHint': '录音胶囊面板内的动态音频波形动画效果',
+      'visLiquidWave': '流体水波 (Liquid Wave)',
+      'visDynamicOrb': '动态脉冲球体 (Dynamic Orb)',
+      'visBars': '经典频条 (Classic Bars)',
+      'visOff': '关闭',
+      'statusTitle': '连接与引擎状态看板',
+      'statusDesc': '当前语音后端与识别服务的实时网络与进程就绪状态。',
+      'badgeHostOnline': '服务端在线 ({ms} ms)',
+      'badgeHostOffline': '服务端不可达',
+      'badgeWhisperActive': 'Whisper 就绪',
+      'badgeWhisperInactive': 'Whisper 未就绪',
+      'badgeSenseVoiceActive': 'SenseVoice 就绪',
+      'badgeSenseVoiceInactive': 'SenseVoice 未就绪',
+      'badgeProviders': '{count} 个服务商已就绪',
+      'refreshStats': '刷新状态',
+      'hardwareTitle': '硬件与本地引擎',
+      'hardwareDesc': '麦克风输入设备、本地 whisper.cpp 与 SenseVoice 运行时配置。',
+      'uiError': '语音界面错误：',
+      'retry': '重试',
+      'testMic': '测试麦克风',
+      'stopTest': '停止测试',
+      'micLevel': '麦克风电平',
+      'noiseGate': '噪声门 (dB)',
+      'noiseGateHint': '静音切除低于阈值的风扇嗡嗡声与机械键盘敲击声。Off = 关闭',
+      'noiseGateOff': '关闭',
     }
 
     // Strings are also needed outside components — in recording handlers and
     // slot labels — so the translator is module-level, not only via props.
     let moduleT = (key) => key
     const t = (key) => moduleT(key)
-
     // ------------------------------------------------------------------ css
     const CSS =
       '.dvo-btn{display:flex;align-items:center;justify-content:center;width:30px;height:30px;border-radius:50%;border:1px solid var(--dsw-alias-border-l1);background:transparent;color:var(--dsw-alias-label-secondary);cursor:pointer;padding:0;box-sizing:border-box;user-select:none;-webkit-user-select:none;touch-action:manipulation}' +
@@ -219,7 +384,8 @@ window.__ModuleLoader__.load({
       '.dvo-badge-idle{background:var(--dsw-alias-bg-layer-2);color:var(--dsw-alias-label-tertiary)}' +
       '.dvo-mic-test{display:flex;align-items:center;gap:10px;margin-top:8px;padding:6px 12px;background:var(--dsw-alias-bg-layer-2);border-radius:8px;border:1px solid var(--dsw-alias-border-l1)}' +
       '.dvo-meter-bar{flex:1;height:8px;background:var(--dsw-alias-bg-layer-3);border-radius:4px;overflow:hidden;position:relative}' +
-      '.dvo-meter-fill{height:100%;background:var(--dsw-alias-state-success-primary);border-radius:4px;transition:width 60ms ease-out}'
+      '.dvo-meter-fill{height:100%;background:var(--dsw-alias-state-success-primary);border-radius:4px;transition:width 60ms ease-out}' +
+      '.dvo-meter-gate{position:absolute;top:0;bottom:0;width:2px;background:var(--dsw-alias-state-warning-primary);box-shadow:0 0 4px var(--dsw-alias-state-warning-primary);z-index:2;pointer-events:none}'
     const cssId = 'dsh-voice/client.module.css'
     if (typeof document !== 'undefined' && !document.querySelector('style[data-dsh-plugin="dsh-voice"][data-plugin-css="' + cssId + '"]')) {
       const tag = document.createElement('style')
@@ -228,7 +394,6 @@ window.__ModuleLoader__.load({
       tag.dataset.pluginCss = cssId
       document.head.appendChild(tag)
     }
-
     // ---------------------------------------------------------------- store
     const voice = {
       phase: 'idle',      // idle | recording | processing | pending | error
@@ -300,8 +465,14 @@ window.__ModuleLoader__.load({
       if (!voice.settings || voice.settings.contextGlossary === false) return []
       const text = (voice.input && typeof voice.input.draft === 'string') ? voice.input.draft : ''
       if (!text || text.length < 3) return []
-      const matches = text.match(/\b[A-Za-z_][A-Za-z0-9_]{2,29}\b/g) || []
-      const stop = new Set(['the', 'and', 'for', 'are', 'but', 'not', 'you', 'all', 'any', 'can', 'her', 'was', 'one', 'our', 'out', 'day', 'get', 'has', 'him', 'his', 'how', 'man', 'new', 'now', 'old', 'see', 'two', 'way', 'who', 'boy', 'did', 'its', 'let', 'put', 'say', 'she', 'too', 'use'])
+      const matches = text.match(/[A-Za-zА-Яа-яЁё_][A-Za-zА-Яа-яЁё0-9_]{2,29}/g) || []
+      const stop = new Set([
+        'the', 'and', 'for', 'are', 'but', 'not', 'you', 'all', 'any', 'can', 'her', 'was',
+        'one', 'our', 'out', 'day', 'get', 'has', 'him', 'his', 'how', 'man', 'new', 'now',
+        'old', 'see', 'two', 'way', 'who', 'boy', 'did', 'its', 'let', 'put', 'say', 'she',
+        'too', 'use', 'это', 'как', 'что', 'для', 'или', 'если', 'все', 'при', 'так', 'уже',
+        'был', 'быть', 'только', 'тоже', 'под', 'над', 'без', 'нет', 'даже', 'где', 'чем',
+      ])
       const words = []
       const seen = new Set()
       for (const m of matches) {
@@ -541,10 +712,13 @@ window.__ModuleLoader__.load({
       if (!rec) return
       try { rec.stream.getTracks().forEach((t) => t.stop()) } catch (e) { /* already stopped */ }
       if (rec.srcNode) { try { rec.srcNode.disconnect() } catch (e) { /* already disconnected */ } }
+      if (rec.filterNode) { try { rec.filterNode.disconnect() } catch (e) { /* already disconnected */ } }
+      if (rec.analyser) { try { rec.analyser.disconnect() } catch (e) { /* already disconnected */ } }
       if (rec.audioCtx) { try { rec.audioCtx.close() } catch (e) { /* already closed */ } }
     }
 
     function waitStop(recorder) {
+      if (!recorder || recorder.state === 'inactive') return Promise.resolve()
       return new Promise((resolve) => recorder.addEventListener('stop', resolve, { once: true }))
     }
 
@@ -581,9 +755,21 @@ window.__ModuleLoader__.load({
         if (rec.audioCtx.state === 'suspended') { rec.audioCtx.resume().catch(() => {}) }
         const src = rec.audioCtx.createMediaStreamSource(stream)
         rec.srcNode = src
+
+        let lastNode = src
+        try {
+          const filter = rec.audioCtx.createBiquadFilter()
+          filter.type = 'highpass'
+          filter.frequency.value = 80
+          filter.Q.value = 0.707
+          src.connect(filter)
+          lastNode = filter
+          rec.filterNode = filter
+        } catch (e) {}
+
         rec.analyser = rec.audioCtx.createAnalyser()
         rec.analyser.fftSize = 128
-        src.connect(rec.analyser)
+        lastNode.connect(rec.analyser)
       }
       // No timeslice: only then each stop() yields a standalone webm file.
       recorder.start()
@@ -596,7 +782,14 @@ window.__ModuleLoader__.load({
       rec.analyser.getByteFrequencyData(data)
       let sum = 0
       for (let i = 0; i < data.length; i++) sum += data[i]
-      return Math.min(1, (sum / data.length / 255) * 2.2)
+      const raw = Math.min(1, (sum / data.length / 255) * 2.2)
+
+      const gateDb = Number(voice.settings && voice.settings.noiseGateDb !== undefined ? voice.settings.noiseGateDb : -45)
+      if (gateDb > -90) {
+        const gateAmp = Math.pow(10, gateDb / 20) * 2.2
+        if (raw < gateAmp) return 0
+      }
+      return raw
     }
 
     // Cut the current phrase: stop the recorder, send the finished file and
@@ -608,7 +801,14 @@ window.__ModuleLoader__.load({
       if (!rec || rec.cutting || rec.closing) return
       rec.cutting = true
       const stopped = waitStop(rec.recorder)
-      try { rec.recorder.stop() } catch (e) { /* already stopped */ }
+      try {
+        if (rec.recorder && rec.recorder.state !== 'inactive') {
+          rec.recorder.stop()
+        }
+      } catch (e) {
+        rec.cutting = false
+        return
+      }
       stopped.then(async () => {
         const blob = new Blob(rec.chunks, { type: rec.mime })
         rec.chunks = []
@@ -616,7 +816,9 @@ window.__ModuleLoader__.load({
         rec.hadSpeech = false
         rec.streamMs = 0
         if (!rec.closing) {
-          try { rec.recorder.start() } catch (e) { /* stream already closed */ }
+          try {
+            if (rec.recorder && rec.recorder.state === 'inactive') rec.recorder.start()
+          } catch (e) { /* stream already closed */ }
         }
         rec.cutting = false
         if (blob.size < 600) return       // too short — not speech
@@ -797,7 +999,9 @@ window.__ModuleLoader__.load({
       if (!rec) { voice.set({ phase: 'idle', error: '' }); return }
       rec.closing = true
       const stopped = waitStop(rec.recorder)
-      try { rec.recorder.stop() } catch (e) { /* already stopped */ }
+      try {
+        if (rec.recorder && rec.recorder.state !== 'inactive') rec.recorder.stop()
+      } catch (e) { /* already stopped */ }
       stopped.then(() => { teardown(rec); voice.rec = null; voice.set({ phase: 'idle', error: '' }) })
     }
 
@@ -827,7 +1031,9 @@ window.__ModuleLoader__.load({
       rec.closing = true
       const mode = rec.mode
       const stopped = waitStop(rec.recorder)
-      try { rec.recorder.stop() } catch (e) { /* already stopped */ }
+      try {
+        if (rec.recorder && rec.recorder.state !== 'inactive') rec.recorder.stop()
+      } catch (e) { /* already stopped */ }
       stopped.then(async () => {
         const blob = new Blob(rec.chunks, { type: rec.mime })
         teardown(rec)
@@ -1344,6 +1550,7 @@ window.__ModuleLoader__.load({
                 polishSend: !!(data && data.modes && data.modes.message && data.modes.message.polishSend),
                 sessionCommands: !!(data && data.modes && data.modes.message && data.modes.message.sessionCommands),
                 noiseSuppression: data && data.noiseSuppression !== false,
+                noiseGateDb: Number(data && data.noiseGateDb !== undefined ? data.noiseGateDb : -45),
                 contextGlossary: data && data.contextGlossary !== false,
                 visualizerStyle: (data && data.visualizerStyle) || 'liquid-wave',
               })
@@ -1641,7 +1848,7 @@ window.__ModuleLoader__.load({
         const timer = setInterval(() => {
           if (tries >= 15) { clearInterval(timer); return }
           tries += 1
-          try { ((ctx.get && ctx.get('lanSettings')) || ctx.settingsScope).describe().load() } catch (e) { /* service not up yet */ }
+          try { ctx.settingsScope.describe().load() } catch (e) { /* service not up yet */ }
         }, 1000)
         return () => clearInterval(timer)
       }, [ready])
@@ -1674,12 +1881,13 @@ window.__ModuleLoader__.load({
           sessionCommands: !!(value && value.message && value.message.sessionCommands),
           polishBaseUrl: String((value && value.polishBaseUrl) || ''),
           noiseSuppression: value.noiseSuppression !== false,
+          noiseGateDb: Number((value && value.noiseGateDb) !== undefined ? value.noiseGateDb : -45),
           contextGlossary: value.contextGlossary !== false,
           visualizerStyle: (value && value.visualizerStyle) || 'liquid-wave',
         })
       }, [ready, value])
 
-            // Interactive microphone tester
+      // Interactive microphone tester
       const [testingMic, setTestingMic] = React.useState(false)
       const [testLevel, setTestLevel] = React.useState(0)
       const testRef = React.useRef(null)
@@ -1706,9 +1914,18 @@ window.__ModuleLoader__.load({
           const audioCtx = new AC()
           if (audioCtx.state === 'suspended') await audioCtx.resume().catch(() => {})
           const src = audioCtx.createMediaStreamSource(stream)
+          let lastNode = src
+          try {
+            const filter = audioCtx.createBiquadFilter()
+            filter.type = 'highpass'
+            filter.frequency.value = 80
+            filter.Q.value = 0.707
+            src.connect(filter)
+            lastNode = filter
+          } catch (e) {}
           const analyser = audioCtx.createAnalyser()
           analyser.fftSize = 128
-          src.connect(analyser)
+          lastNode.connect(analyser)
           testRef.current = { stream, audioCtx, analyser, active: true }
           setTestingMic(true)
 
@@ -1807,6 +2024,7 @@ window.__ModuleLoader__.load({
           streamChunkMs: Number(draft.dictation && draft.dictation.streamChunkMs) || 1200,
           vadAdapt: Number(draft.dictation && draft.dictation.vadAdapt) || 0,
           noiseSuppression: draft.noiseSuppression !== false,
+          noiseGateDb: Number(draft.noiseGateDb !== undefined ? draft.noiseGateDb : -45),
           contextGlossary: draft.contextGlossary !== false,
           visualizerStyle: draft.visualizerStyle || 'liquid-wave',
         })
@@ -1843,6 +2061,9 @@ window.__ModuleLoader__.load({
         React.createElement('span', { className: 'dvs-sub' }, t('hotkeyHint1') + t('hotkeyHint2')),
       )
 
+      const gateDbVal = Number((draft && draft.noiseGateDb) !== undefined ? draft.noiseGateDb : (value && value.noiseGateDb) !== undefined ? value.noiseGateDb : -45)
+      const gateThresholdPercent = gateDbVal <= -90 ? 0 : Math.min(100, Math.max(0, Math.round(((gateDbVal + 60) / 40) * 100)))
+
       const micField = () => React.createElement('div', { className: 'cb-field' },
         React.createElement('label', null, t('mic')),
         React.createElement('div', { className: 'cb-row' },
@@ -1866,9 +2087,25 @@ window.__ModuleLoader__.load({
           React.createElement('span', { style: { fontSize: '11px', color: 'var(--dsw-alias-label-secondary)' } }, t('micLevel') + ':'),
           React.createElement('div', { className: 'dvo-meter-bar' },
             React.createElement('div', { className: 'dvo-meter-fill', style: { width: `${testLevel}%` } }),
+            gateThresholdPercent > 0 ? React.createElement('div', { className: 'dvo-meter-gate', style: { left: `${gateThresholdPercent}%` }, title: `${gateDbVal} dB` }) : null,
           ),
           React.createElement('span', { style: { fontSize: '11px', fontVariantNumeric: 'tabular-nums', width: '32px', textAlign: 'right' } }, `${testLevel}%`),
         ) : null,
+        React.createElement('div', { className: 'cb-row', style: { marginTop: '6px' } },
+          React.createElement('span', { style: { fontSize: '12px', color: 'var(--dsw-alias-label-secondary)', minWidth: '120px' } }, t('noiseGate') + ':'),
+          React.createElement('select', {
+            value: String(gateDbVal), disabled: !writable,
+            onChange: (e) => setTop('noiseGateDb', Number(e.target.value)),
+            style: { flex: 1 },
+          },
+            React.createElement('option', { value: '-999' }, t('noiseGateOff')),
+            React.createElement('option', { value: '-50' }, '-50 dB (Low / Gentle)'),
+            React.createElement('option', { value: '-45' }, '-45 dB (Standard Default)'),
+            React.createElement('option', { value: '-35' }, '-35 dB (Medium Room)'),
+            React.createElement('option', { value: '-25' }, '-25 dB (High / Mechanical Clicks)'),
+          ),
+        ),
+        React.createElement('span', { className: 'dvs-sub' }, t('noiseGateHint')),
       )
 
       const langField = (mode) => React.createElement('div', { className: 'cb-field' },
@@ -2180,7 +2417,6 @@ window.__ModuleLoader__.load({
         ),
       )
     }
-
     // Card in Settings → Plugins → Plugin settings (#18)
     function PluginCard(props) {
       const [open, setOpen] = React.useState(false)
@@ -2221,9 +2457,9 @@ window.__ModuleLoader__.load({
 
     exports.inject = ['timer', 'slots', 'settingsScope', 'locale']
     exports.apply = function apply(ctx) {
-      // English is the source language. Other languages come from the
-      // separate translation plugin at runtime. Re-registering the same
-      // namespace+language throws, so tolerate a pre-registered dictionary.
+      // English is the source and fallback language; Chinese is the built-in
+      // localized user dictionary. Other languages (like Russian) come from
+      // the separate translation plugin at runtime.
       const addLocale = (locale, dictionary) => {
         try {
           return ctx.locale.register(NS, locale, dictionary)
@@ -2232,7 +2468,7 @@ window.__ModuleLoader__.load({
         }
       }
       ctx.effect(() => {
-        const undo = [addLocale('en', en)]
+        const undo = [addLocale('en', en), addLocale('zh', zh)]
         return () => { for (const off of undo) off() }
       }, 'dsh-voice: locale dictionaries')
       moduleT = ctx.locale.bind(NS)

package/lib/index.js

@@ -126,6 +126,8 @@ export const Config = z.object({
     .description('Restrict both chains to local-whisper only: fully offline, no cloud providers.'),
   micDeviceId: z.string().default('')
     .description('Microphone device id for recording. Empty means the system default.'),
+  noiseGateDb: z.number().default(-45)
+    .description('Audio noise gate threshold in dB for client-side recording (-50 to -25 dB, or <= -90 to disable). Silence below this threshold cuts ambient background noise and prevents false VAD triggers.'),
   historyLimit: z.number().default(20)
     .description('How many recent dictation inserts to keep for undo in the browser. 0 disables history.'),
   vocabulary: z.array(z.string()).default([])
@@ -301,7 +303,7 @@ export function apply(ctx, baseConfig) {
       : (Array.isArray(cfg.vocabulary) ? cfg.vocabulary : [])
 
     const providers = makeProviders(
-      { resolveKey, fetchImpl: fetch, cfg, toWav: (b) => toWav16k(b, cfg.ffmpegBin) },
+      { resolveKey, fetchImpl: fetch, cfg, toWav: (b, sig) => toWav16k(b, cfg.ffmpegBin, sig || signal) },
       { bytes, mime, lang: modeCfg.language, signal, models, vocabulary: vocab },
     )
     return runChain(order, providers, statsTracker.record)
@@ -354,6 +356,7 @@ export function apply(ctx, baseConfig) {
         beep: cfg.beep,
         localOnly: cfg.localOnly,
         micDeviceId: cfg.micDeviceId,
+        noiseGateDb: cfg.noiseGateDb !== undefined ? cfg.noiseGateDb : -45,
         historyLimit: cfg.historyLimit,
         voiceCommands: cfg.voiceCommands,
         wakeWord: String(cfg.wakeWord || ''),

package/lib/providers.js

@@ -158,6 +158,7 @@ export function makeProviders(deps, req) {
     form.append('file', new Blob([bytes], { type: mime }), fileName(mime))
     form.append('model', pickModel(models, 'groq'))
     if (!isAutoLang(lang)) form.append('language', lang)
+    if (vocab) form.append('prompt', vocab)
     form.append('response_format', 'json')
     const res = await fetchImpl('https://api.groq.com/openai/v1/audio/transcriptions', {
       method: 'POST',
@@ -205,7 +206,7 @@ export function makeProviders(deps, req) {
         return { ok: false, provider: 'local-whisper', reason: 'local whisper needs WAV, no converter configured' }
       }
       try {
-        sendBytes = await deps.toWav(bytes)
+        sendBytes = await deps.toWav(bytes, signal)
         sendMime = 'audio/wav'
       } catch (e) {
         return { ok: false, provider: 'local-whisper', reason: `local whisper: ${String(e && e.message || e)}` }
@@ -249,7 +250,7 @@ export function makeProviders(deps, req) {
         return { ok: false, provider: 'sensevoice', reason: 'sensevoice needs WAV, no converter configured' }
       }
       try {
-        sendBytes = await deps.toWav(bytes)
+        sendBytes = await deps.toWav(bytes, signal)
         sendMime = 'audio/wav'
       } catch (e) {
         return { ok: false, provider: 'sensevoice', reason: `sensevoice: ${String(e && e.message || e)}` }
@@ -259,12 +260,11 @@ export function makeProviders(deps, req) {
     form.append('file', new Blob([sendBytes], { type: sendMime }), fileName(sendMime))
     const isOpenAI = url.includes('/transcriptions')
     if (isOpenAI) {
-      const model = pickModel(models, 'sensevoice') || 'SenseVoiceSmall'
-      form.append('model', model)
+      form.append('model', pickModel(models, 'sensevoice'))
+      if (!isAutoLang(lang)) form.append('language', lang)
+      if (vocab) form.append('prompt', vocab)
       form.append('response_format', 'json')
     }
-    if (!isAutoLang(lang)) form.append('language', lang)
-    if (vocab) form.append('prompt', vocab)
 
     let res
     try {
@@ -316,6 +316,7 @@ export function makeProviders(deps, req) {
       form.append('file', new Blob([bytes], { type: mime }), fileName(mime))
       form.append('model', model)
       if (lang && lang !== 'auto') form.append('language', lang)
+      if (vocab) form.append('prompt', vocab)
       form.append('response_format', 'json')
       const res = await fetchImpl(`${base}/audio/transcriptions`, {
         method: 'POST', headers, body: form, signal,
@@ -330,9 +331,9 @@ export function makeProviders(deps, req) {
       let format = chatAudioFormat(mime)
       if (!format) {
         if (typeof deps.toWav !== 'function') {
-          throw new Error(`${label} needs wav or mp3, no converter configured`)
+          throw new Error(`${label}: chat template needs wav or mp3, unsupported format ${mime}`)
         }
-        sendBytes = await deps.toWav(bytes)
+        sendBytes = await deps.toWav(bytes, signal)
         format = 'wav'
       }
       const ask = (spec.prompt || CHAT_AUDIO_PROMPT)

package/lib/wav.js

@@ -11,8 +11,12 @@ import { spawn } from 'node:child_process'
  * @param ffmpegBin {string} ffmpeg binary path
  * @returns {Promise<Buffer>} 16 kHz mono WAV
  */
-export function toWav16k(bytes, ffmpegBin = 'ffmpeg') {
+export function toWav16k(bytes, ffmpegBin = 'ffmpeg', signal = null) {
   return new Promise((resolve, reject) => {
+    if (signal?.aborted) {
+      reject(new Error('ffmpeg conversion aborted before start'))
+      return
+    }
     const proc = spawn(ffmpegBin, [
       '-hide_banner', '-loglevel', 'error',
       '-i', 'pipe:0',
@@ -21,10 +25,37 @@ export function toWav16k(bytes, ffmpegBin = 'ffmpeg') {
     ])
     const out = []
     const err = []
+    let done = false
+
+    const timeout = setTimeout(() => {
+      cleanup()
+      try { proc.kill('SIGKILL') } catch {}
+      reject(new Error('ffmpeg conversion timed out after 30s'))
+    }, 30000)
+
+    const onAbort = () => {
+      cleanup()
+      try { proc.kill('SIGKILL') } catch {}
+      reject(new Error('ffmpeg conversion aborted'))
+    }
+
+    if (signal) signal.addEventListener('abort', onAbort, { once: true })
+
+    function cleanup() {
+      if (done) return
+      done = true
+      clearTimeout(timeout)
+      if (signal) signal.removeEventListener('abort', onAbort)
+    }
+
     proc.stdout.on('data', (c) => out.push(c))
     proc.stderr.on('data', (c) => err.push(c))
-    proc.on('error', (e) => reject(new Error(`ffmpeg unavailable: ${e.message}`)))
+    proc.on('error', (e) => {
+      cleanup()
+      reject(new Error(`ffmpeg unavailable: ${e.message}`))
+    })
     proc.on('close', (code) => {
+      cleanup()
       if (code !== 0) {
         reject(new Error(`ffmpeg exit ${code}: ${Buffer.concat(err).toString('utf8').slice(0, 200)}`))
         return