dsh-live-voice 0.0.2 → 0.2.0

package/lib/client.js

@@ -35,6 +35,7 @@ __export(index_exports, {
 });
 module.exports = __toCommonJS(index_exports);
 var import_react = __toESM(require("react"), 1);
+var import_react_dom = require("react-dom");
 
 // src/core/transcript.ts
 var TranscriptDraft = class {
@@ -122,8 +123,24 @@ var defaultSettings = Object.freeze({
   lang: "pt-BR",
   recognitionLang: "pt-BR",
   voice: "",
+  inputDeviceId: "",
+  outputDeviceId: "",
+  recognitionFilterEnabled: true,
+  recognitionMinimumWords: 2,
+  voiceCommandsEnabled: true,
+  voiceCommandSend: "send, send message",
+  voiceCommandQueue: "queue, queue message",
+  voiceCommandEnd: "end, end conversation",
+  voiceCommandMute: "mute, stop listening",
+  voiceCommandResume: "resume, start listening",
+  voiceCommandStopSpeaking: "stop talking, stop speaking, shut up",
+  voiceCommandClear: "clear all, clear message",
+  outputCodeFilterEnabled: true,
+  outputCodeMaxLines: 5,
+  outputCodeNotice: "Look the code on out conversation",
   rate: 1
 });
+var normalizeCommandPhrases = (value, fallback) => typeof value === "string" && value.length <= 1e3 && !value.includes("\0") ? value.split(/[,\r\n]+/).map((phrase) => phrase.trim()).filter(Boolean).slice(0, 20).join(", ") : fallback;
 function normalizeSettings(value) {
   const source = value && typeof value === "object" && !Array.isArray(value) ? value : {};
   return {
@@ -134,17 +151,92 @@ function normalizeSettings(value) {
     voiceDetectionPreset: Object.hasOwn(voiceDetectionPresets, source.voiceDetectionPreset) ? source.voiceDetectionPreset : defaultSettings.voiceDetectionPreset,
     announceAssistantMessages: typeof source.announceAssistantMessages === "boolean" ? source.announceAssistantMessages : defaultSettings.announceAssistantMessages,
     interruptSpeechOnUserMessage: typeof source.interruptSpeechOnUserMessage === "boolean" ? source.interruptSpeechOnUserMessage : defaultSettings.interruptSpeechOnUserMessage,
-    sendingMode: ["manual", "automatic"].includes(source.sendingMode) ? source.sendingMode : defaultSettings.sendingMode,
+    sendingMode: source.sendingMode === "automatic" ? "queue" : ["manual", "queue", "steer"].includes(source.sendingMode) ? source.sendingMode : defaultSettings.sendingMode,
     autoSendDelaySeconds: Number.isInteger(source.autoSendDelaySeconds) && source.autoSendDelaySeconds >= 2 && source.autoSendDelaySeconds <= 10 ? source.autoSendDelaySeconds : defaultSettings.autoSendDelaySeconds,
     assistantSpeechDelaySeconds: Number.isInteger(source.assistantSpeechDelaySeconds) && source.assistantSpeechDelaySeconds >= 1 && source.assistantSpeechDelaySeconds <= 10 ? source.assistantSpeechDelaySeconds : defaultSettings.assistantSpeechDelaySeconds,
     mode: ["speaker", "headphones"].includes(source.mode) ? source.mode : defaultSettings.mode,
     lang: typeof source.lang === "string" && /^[a-z]{2,3}(?:-[A-Za-z0-9]{2,8})*$/.test(source.lang) ? source.lang : defaultSettings.lang,
     recognitionLang: source.recognitionLang === "auto" || typeof source.recognitionLang === "string" && /^[a-z]{2,3}(?:-[A-Za-z0-9]{2,8})*$/.test(source.recognitionLang) ? source.recognitionLang : typeof source.lang === "string" && /^[a-z]{2,3}(?:-[A-Za-z0-9]{2,8})*$/.test(source.lang) ? source.lang : defaultSettings.recognitionLang,
     voice: typeof source.voice === "string" && source.voice.length <= 200 && !source.voice.includes("\0") ? source.voice : "",
+    inputDeviceId: typeof source.inputDeviceId === "string" && source.inputDeviceId.length <= 500 && !source.inputDeviceId.includes("\0") ? source.inputDeviceId : "",
+    outputDeviceId: typeof source.outputDeviceId === "string" && source.outputDeviceId.length <= 500 && !source.outputDeviceId.includes("\0") ? source.outputDeviceId : "",
+    recognitionFilterEnabled: typeof source.recognitionFilterEnabled === "boolean" ? source.recognitionFilterEnabled : defaultSettings.recognitionFilterEnabled,
+    recognitionMinimumWords: Number.isInteger(source.recognitionMinimumWords) && source.recognitionMinimumWords >= 1 && source.recognitionMinimumWords <= 20 ? source.recognitionMinimumWords : defaultSettings.recognitionMinimumWords,
+    voiceCommandsEnabled: typeof source.voiceCommandsEnabled === "boolean" ? source.voiceCommandsEnabled : defaultSettings.voiceCommandsEnabled,
+    voiceCommandSend: normalizeCommandPhrases(
+      source.voiceCommandSend,
+      defaultSettings.voiceCommandSend
+    ),
+    voiceCommandQueue: normalizeCommandPhrases(
+      source.voiceCommandQueue,
+      defaultSettings.voiceCommandQueue
+    ),
+    voiceCommandEnd: normalizeCommandPhrases(
+      source.voiceCommandEnd,
+      defaultSettings.voiceCommandEnd
+    ),
+    voiceCommandMute: normalizeCommandPhrases(
+      source.voiceCommandMute,
+      defaultSettings.voiceCommandMute
+    ),
+    voiceCommandResume: normalizeCommandPhrases(
+      source.voiceCommandResume,
+      defaultSettings.voiceCommandResume
+    ),
+    voiceCommandStopSpeaking: normalizeCommandPhrases(
+      source.voiceCommandStopSpeaking,
+      defaultSettings.voiceCommandStopSpeaking
+    ),
+    voiceCommandClear: normalizeCommandPhrases(
+      source.voiceCommandClear,
+      defaultSettings.voiceCommandClear
+    ),
+    outputCodeFilterEnabled: typeof source.outputCodeFilterEnabled === "boolean" ? source.outputCodeFilterEnabled : defaultSettings.outputCodeFilterEnabled,
+    outputCodeMaxLines: Number.isInteger(source.outputCodeMaxLines) && source.outputCodeMaxLines >= 0 && source.outputCodeMaxLines <= 100 ? source.outputCodeMaxLines : defaultSettings.outputCodeMaxLines,
+    outputCodeNotice: typeof source.outputCodeNotice === "string" && source.outputCodeNotice.trim() && source.outputCodeNotice.length <= 300 && !source.outputCodeNotice.includes("\0") ? source.outputCodeNotice.trim() : defaultSettings.outputCodeNotice,
     rate: Number.isFinite(source.rate) && source.rate >= 0.1 && source.rate <= 3 ? source.rate : 1
   };
 }
 
+// src/core/filters.ts
+var words = (text) => String(text || "").trim().match(/[\p{L}\p{N}]+(?:['’_-][\p{L}\p{N}]+)*/gu) || [];
+function normalizeVoiceCommand(text) {
+  return String(text || "").normalize("NFD").replace(/[\u0300-\u036f]/g, "").toLocaleLowerCase("en-US").replace(/[^\p{L}\p{N}]+/gu, " ").trim().replace(/\s+/g, " ");
+}
+function matchVoiceCommand(text, commands) {
+  const candidate = normalizeVoiceCommand(text);
+  if (!candidate) return null;
+  for (const [action, phrases] of Object.entries(commands || {})) {
+    if (String(phrases || "").split(/[,\r\n]+/).some((phrase) => normalizeVoiceCommand(phrase) === candidate))
+      return action;
+  }
+  return null;
+}
+function hasMinimumWords(text, minimum = 2) {
+  return words(text).length >= minimum;
+}
+function splitSpeechOutput(text, options = {}) {
+  const filtered = filterSpeechOutput(text, options).trim();
+  if (!filtered) return [];
+  return filtered.split(/\r?\n+/u).map((segment) => segment.trim()).filter(Boolean);
+}
+function hasUnclosedCodeFence(text) {
+  return (String(text || "").match(/```/g) || []).length % 2 === 1;
+}
+function filterSpeechOutput(text, {
+  filterCodeBlocks = true,
+  codeBlockMaxLines = 5,
+  codeBlockNotice = "Look the code on out conversation"
+} = {}) {
+  const source = String(text || "");
+  if (!filterCodeBlocks) return source;
+  return source.replace(/```[^\n]*\n?([\s\S]*?)```/g, (block, body) => {
+    const normalized = body.replace(/\n$/, "");
+    const lineCount = normalized ? normalized.split(/\r?\n/).length : 0;
+    return lineCount <= codeBlockMaxLines ? normalized : codeBlockNotice;
+  });
+}
+
 // src/core/coordinator.ts
 var message = (error) => error?.message || String(error);
 function cancellable(promise, signal) {
@@ -179,20 +271,27 @@ var VoiceCoordinator = class {
       conversation: false,
       listening: false,
       recognizing: false,
+      muted: false,
       speaking: false,
       paused: false,
       starting: false,
       error: null,
       activeMessageId: null,
+      answeringQuestion: false,
       autoSendAt: null,
+      speechSegmentsRemaining: 0,
       capabilities: {},
       settings: normalizeSettings(settings)
     };
     this.autoSendTimer = null;
     this.autoSendDraft = null;
+    this.interruptionTimer = null;
+    this.interruptionTranscriptConfirmed = false;
+    this.interruptionPausedSpeech = false;
     this.assistantSpeechTimer = null;
     this.assistantSpeechNotBefore = 0;
     this.recognition.lang = this.snapshot.settings.recognitionLang;
+    this.meter.deviceId = this.snapshot.settings.inputDeviceId;
     this.getSnapshot = () => this.snapshot;
     this.subscribe = (listener) => {
       if (this.disposed) return () => {
@@ -232,6 +331,7 @@ var VoiceCoordinator = class {
   updateSettings(next) {
     if (this.disposed) return;
     const settings = normalizeSettings({ ...this.snapshot.settings, ...next });
+    this.meter.deviceId = settings.inputDeviceId;
     Object.assign(this.recognition, {
       lang: settings.recognitionLang,
       processLocally: settings.recognitionProcessLocally,
@@ -239,7 +339,7 @@ var VoiceCoordinator = class {
       voiceDetectionPreset: settings.voiceDetectionPreset
     });
     this.patch({ settings, error: null });
-    if (settings.sendingMode !== "automatic") this.cancelAutoSend();
+    if (settings.sendingMode === "manual") this.cancelAutoSend();
     if (Object.hasOwn(next, "announceAssistantMessages") && !settings.announceAssistantMessages) {
       this.queue = [];
       this.assistantSpeechNotBefore = 0;
@@ -300,6 +400,21 @@ var VoiceCoordinator = class {
   startConversation() {
     return this.startListening(true);
   }
+  muteListening() {
+    if (!this.snapshot.settings.voiceCommandsEnabled) return this.stopListening();
+    this.cancelAutoSend();
+    this.composer.setDraft(this.transcript.update(this.composer.getDraft(), "", true));
+    this.transcript.reset();
+    this.patch({ muted: true, recognizing: false });
+  }
+  resumeListeningInput() {
+    if (this.snapshot.listening || this.snapshot.starting) {
+      this.transcript.reset();
+      this.patch({ muted: false, recognizing: false });
+      return;
+    }
+    return this.startListening(this.snapshot.conversation);
+  }
   startListening(conversation = this.snapshot.conversation) {
     if (this.disposed) return Promise.resolve();
     this.patch({ error: null });
@@ -369,7 +484,7 @@ var VoiceCoordinator = class {
             } else
               this.assistantSpeechNotBefore = Date.now() + this.snapshot.settings.assistantSpeechDelaySeconds * 1e3;
             this.patch({ recognizing: active });
-            if (active && this.snapshot.settings.mode === "headphones") void this.pauseSpeech();
+            this._handleSpeechInterruption(active);
             if (!active) this._drain();
           },
           onError: (error) => {
@@ -404,16 +519,99 @@ var VoiceCoordinator = class {
       }
     });
   }
+  _interruptionTranscriptQualifies(text) {
+    return hasMinimumWords(text, 1);
+  }
+  _handleSpeechInterruption(active) {
+    if (this.snapshot.settings.mode !== "headphones") return;
+    if (active) {
+      if (!this.interruptionTranscriptConfirmed || this.interruptionTimer !== null || !this.snapshot.speaking || this.snapshot.paused)
+        return;
+      this.interruptionTimer = setTimeout(() => {
+        this.interruptionTimer = null;
+        if (this.interruptionTranscriptConfirmed && this.snapshot.recognizing && this.snapshot.speaking && !this.snapshot.paused) {
+          this.interruptionPausedSpeech = true;
+          void this.pauseSpeech().then(() => {
+            if (this.interruptionPausedSpeech && !this.snapshot.recognizing && this.snapshot.paused) {
+              this.interruptionPausedSpeech = false;
+              void this.resumeSpeech();
+            }
+          });
+        }
+      }, this.snapshot.settings.assistantSpeechDelaySeconds * 1e3);
+      return;
+    }
+    if (this.interruptionTimer !== null) clearTimeout(this.interruptionTimer);
+    this.interruptionTimer = null;
+    this.interruptionTranscriptConfirmed = false;
+    if (this.interruptionPausedSpeech && this.snapshot.paused) {
+      this.interruptionPausedSpeech = false;
+      void this.resumeSpeech();
+    }
+  }
   onResult({ final = "", interim = "" }) {
     if (this.disposed || !this.snapshot.listening && !this.snapshot.starting) return;
-    if (this.snapshot.speaking && !this.snapshot.paused) {
-      if (this.snapshot.settings.mode === "speaker") return;
-      if (final || interim) void this.pauseSpeech();
+    if (this.snapshot.speaking && !this.snapshot.paused && this.snapshot.settings.mode === "speaker")
+      return;
+    if (this.snapshot.settings.mode === "headphones" && this.snapshot.speaking && this._interruptionTranscriptQualifies(final || interim)) {
+      this.interruptionTranscriptConfirmed = true;
+      this._handleSpeechInterruption(true);
+    }
+    if (typeof this.composer.handleQuestionResult === "function" && this.composer.handleQuestionResult({ final, interim })) {
+      this.cancelAutoSend();
+      this.transcript.reset();
+      this.patch({ recognizing: !!interim });
+      return;
     }
     if (final) {
+      const command = this.snapshot.settings.voiceCommandsEnabled && matchVoiceCommand(final, {
+        send: this.snapshot.settings.voiceCommandSend,
+        queue: this.snapshot.settings.voiceCommandQueue,
+        end: this.snapshot.settings.voiceCommandEnd,
+        mute: this.snapshot.settings.voiceCommandMute,
+        resume: this.snapshot.settings.voiceCommandResume,
+        stopSpeaking: this.snapshot.settings.voiceCommandStopSpeaking,
+        clear: this.snapshot.settings.voiceCommandClear
+      });
+      if (command) {
+        this.composer.setDraft(this.transcript.update(this.composer.getDraft(), "", true));
+        this.transcript.reset();
+        this.cancelAutoSend();
+        this.patch({ recognizing: false });
+        if (command === "end") void this.endConversation();
+        else if (command === "mute") this.muteListening();
+        else if (command === "resume") this.resumeListeningInput();
+        else if (command === "stopSpeaking") {
+          this.updateSettings({ announceAssistantMessages: false });
+          void this.stopSpeech();
+        } else if (command === "clear") {
+          this.cancelAutoSend();
+          this.transcript.reset();
+          this.composer.setDraft("");
+        } else if (this.snapshot.muted) return;
+        else {
+          const sendingMode = command === "send" ? "steer" : "queue";
+          this.updateSettings({ sendingMode });
+          if (this.composer.getDraft().trim() && typeof this.composer.submit === "function")
+            this.composer.submit(sendingMode);
+        }
+        return;
+      }
+      if (this.snapshot.muted) {
+        this.composer.setDraft(this.transcript.update(this.composer.getDraft(), "", true));
+        this.transcript.reset();
+        this.patch({ recognizing: false });
+        return;
+      }
+      if (this.snapshot.settings.recognitionFilterEnabled && !hasMinimumWords(final, this.snapshot.settings.recognitionMinimumWords)) {
+        this.composer.setDraft(this.transcript.update(this.composer.getDraft(), "", true));
+        this.transcript.reset();
+        this.patch({ recognizing: false });
+        return;
+      }
       const next = this.transcript.update(this.composer.getDraft(), final, true);
       this.composer.setDraft(next);
-      if (this.snapshot.settings.sendingMode === "automatic") this.scheduleAutoSend(next);
+      if (this.snapshot.settings.sendingMode !== "manual") this.scheduleAutoSend(next);
     }
     if (interim) {
       this.cancelAutoSend();
@@ -435,8 +633,15 @@ var VoiceCoordinator = class {
       const expected = this.autoSendDraft;
       this.autoSendDraft = null;
       this.patch({ autoSendAt: null });
-      if (!this.disposed && this.snapshot.settings.sendingMode === "automatic" && expected === this.composer.getDraft())
-        this.composer.submit();
+      if (!this.disposed && this.snapshot.settings.sendingMode !== "manual" && expected === this.composer.getDraft()) {
+        try {
+          this.composer.submit(this.snapshot.settings.sendingMode === "steer" ? "steer" : "queue");
+          this.transcript.reset();
+          this.recognition.reset?.();
+        } catch (error) {
+          this.patch({ error: message(error) });
+        }
+      }
     }, delay);
   }
   cancelAutoSend() {
@@ -474,6 +679,11 @@ var VoiceCoordinator = class {
     this.patch({ conversation: false });
     await Promise.all([this.stopListening(), this.stopSpeech(false)]);
   }
+  _syncSpeechSegments() {
+    const speechSegmentsRemaining = this.queue.length + (this.snapshot.speaking ? 1 : 0);
+    if (speechSegmentsRemaining !== this.snapshot.speechSegmentsRemaining)
+      this.patch({ speechSegmentsRemaining });
+  }
   _suppressPending() {
     for (const id2 of this.unfinished) this.suppressed.add(id2);
     if (this.snapshot.activeMessageId !== null) this.suppressed.add(this.snapshot.activeMessageId);
@@ -481,10 +691,15 @@ var VoiceCoordinator = class {
     this.queue = [];
   }
   async stopSpeech(resumeListening = true) {
+    if (this.interruptionTimer !== null) clearTimeout(this.interruptionTimer);
+    this.interruptionTimer = null;
+    this.interruptionTranscriptConfirmed = false;
+    this.interruptionPausedSpeech = false;
     const epoch = ++this.speechEpoch;
     ++this.controlEpoch;
     this._suppressPending();
     this.patch({ speaking: false, paused: false, activeMessageId: null });
+    this._syncSpeechSegments();
     const stopped = this.speechBarrier.then(async () => {
       const results = await Promise.allSettled(
         Object.values(this.engines).map((engine) => Promise.resolve().then(() => engine.stop()))
@@ -538,13 +753,23 @@ var VoiceCoordinator = class {
     const stopped = this.stopSpeech(false);
     const epoch = this.speechEpoch;
     await stopped;
-    if (!this.disposed && epoch === this.speechEpoch && !this.speechStopError)
-      return this.play(text, messageId);
+    if (this.disposed || epoch !== this.speechEpoch || this.speechStopError) return;
+    const segments = splitSpeechOutput(text, {
+      filterCodeBlocks: this.snapshot.settings.outputCodeFilterEnabled,
+      codeBlockMaxLines: this.snapshot.settings.outputCodeMaxLines,
+      codeBlockNotice: this.snapshot.settings.outputCodeNotice
+    });
+    const first = segments.shift();
+    if (!first) return;
+    this.queue.push(...segments.map((segment) => ({ text: segment, id: messageId, manual: true })));
+    this._syncSpeechSegments();
+    return this.play(first, messageId, true);
   }
-  async play(text, messageId) {
+  async play(text, messageId, manual = false) {
     if (this.disposed || !text.trim()) return;
     if (this.snapshot.speaking) {
-      this.queue.push({ text, id: messageId });
+      this.queue.push({ text, id: messageId, manual });
+      this._syncSpeechSegments();
       return;
     }
     const epoch = ++this.speechEpoch;
@@ -555,6 +780,7 @@ var VoiceCoordinator = class {
       return;
     }
     this.patch({ speaking: true, paused: false, activeMessageId: messageId, error: null });
+    this._syncSpeechSegments();
     let failed = false;
     try {
       await this.speechBarrier;
@@ -567,7 +793,8 @@ var VoiceCoordinator = class {
       if (epoch !== this.speechEpoch || this.disposed) return;
       await engine.speak(text, {
         voice: this.snapshot.settings.voice || void 0,
-        rate: this.snapshot.settings.rate
+        rate: this.snapshot.settings.rate,
+        outputDeviceId: this.snapshot.settings.outputDeviceId
       });
     } catch (error) {
       if (epoch === this.speechEpoch) {
@@ -578,6 +805,7 @@ var VoiceCoordinator = class {
     } finally {
       if (epoch === this.speechEpoch && !this.disposed) {
         this.patch({ speaking: false, paused: false, activeMessageId: null });
+        this._syncSpeechSegments();
         if (!failed && this.queue.length) this._drain();
         else if (this.snapshot.conversation && !this.snapshot.listening && !this.snapshot.starting)
           await this._startInput(true);
@@ -589,7 +817,7 @@ var VoiceCoordinator = class {
     this.assistantSpeechTimer = null;
   }
   _drain() {
-    if (this.disposed || !this.snapshot.conversation || !this.snapshot.settings.announceAssistantMessages || this.snapshot.speaking || this.snapshot.recognizing || this.snapshot.starting || !this.queue.length)
+    if (this.disposed || !this.snapshot.conversation && !this.queue[0]?.manual || !this.snapshot.settings.announceAssistantMessages && !this.queue[0]?.manual || this.snapshot.speaking || this.snapshot.recognizing || this.snapshot.starting || !this.queue.length)
       return;
     const wait = this.assistantSpeechNotBefore - Date.now();
     if (wait > 0) {
@@ -603,7 +831,7 @@ var VoiceCoordinator = class {
     this._cancelAssistantSpeechTimer();
     this.assistantSpeechNotBefore = 0;
     const next = this.queue.shift();
-    void this.play(next.text, next.id);
+    void this.play(next.text, next.id, next.manual === true);
   }
   /** Baselines survive conversation toggles; cancelled message IDs remain suppressed. */
   observeMessage(id2, text, { complete = false, baseline = false } = {}) {
@@ -620,17 +848,19 @@ var VoiceCoordinator = class {
       return;
     }
     const remaining = text.slice(offset);
-    const boundary = complete ? remaining.length : Math.max(
-      remaining.lastIndexOf(". "),
-      remaining.lastIndexOf("? "),
-      remaining.lastIndexOf("! "),
-      remaining.lastIndexOf("\n")
-    ) + 1;
+    const boundary = complete ? remaining.length : remaining.lastIndexOf("\n") + 1;
     if (boundary <= 0) return;
-    const chunk = remaining.slice(0, boundary).trim();
+    if (this.snapshot.settings.outputCodeFilterEnabled && hasUnclosedCodeFence(remaining.slice(0, boundary)) && !complete)
+      return;
+    const chunk = filterSpeechOutput(remaining.slice(0, boundary).trim(), {
+      filterCodeBlocks: this.snapshot.settings.outputCodeFilterEnabled,
+      codeBlockMaxLines: this.snapshot.settings.outputCodeMaxLines,
+      codeBlockNotice: this.snapshot.settings.outputCodeNotice
+    }).trim();
     this.consumed.set(id2, offset + boundary);
     if (chunk) {
       this.queue.push({ text: chunk, id: id2 });
+      this._syncSpeechSegments();
       this._drain();
     }
   }
@@ -640,7 +870,7 @@ var VoiceCoordinator = class {
     this._cancelAssistantSpeechTimer();
     this.disposed = true;
     this.listeners.clear();
-    await this.endConversation();
+    await this.endConversation(true);
   }
 };
 
@@ -657,7 +887,7 @@ var VoiceOwnership = class {
     const acquired = this.tail.then(async () => {
       if (this.closed || epoch !== this.epoch || owner.disposed) return;
       await Promise.all(
-        owners.filter((other) => other !== owner).map((other) => other.endConversation())
+        owners.filter((other) => other !== owner).map((other) => other.endConversation(true))
       );
       if (this.closed || epoch !== this.epoch || owner.disposed) return;
       outcome = Promise.resolve(action());
@@ -681,6 +911,7 @@ var VoiceOwnership = class {
 var MicrophoneMeter = class {
   constructor(globals = globalThis) {
     this.g = globals;
+    this.deviceId = "";
     this.epoch = 0;
     this.current = null;
     this.stream = this.context = this.source = this.analyser = this.samples = null;
@@ -745,7 +976,12 @@ var MicrophoneMeter = class {
     const capture = async () => {
       try {
         const stream = await this.g.navigator.mediaDevices.getUserMedia({
-          audio: { echoCancellation: true, noiseSuppression: true, autoGainControl: true }
+          audio: {
+            ...this.deviceId ? { deviceId: { exact: this.deviceId } } : {},
+            echoCancellation: true,
+            noiseSuppression: true,
+            autoGainControl: true
+          }
         });
         job.stream = stream;
         if (!valid()) {
@@ -1271,6 +1507,26 @@ var BrowserRecognitionEngine = class {
     };
     recognition.start();
   }
+  reset() {
+    const session = this.session;
+    const recognition = session?.recognition;
+    if (!session || !recognition) return false;
+    this._detach(recognition);
+    session.recognition = null;
+    try {
+      recognition.abort();
+    } catch {
+    }
+    this._activity(session, false);
+    if (this.session !== session || session.signal?.aborted) return false;
+    try {
+      this._begin(session);
+      return true;
+    } catch (error) {
+      this._fail(session, error);
+      return false;
+    }
+  }
   async stop() {
     const session = this.session;
     if (!session) return;
@@ -1522,7 +1778,7 @@ var QwenHttpSpeakingEngine = class {
       };
     }
   }
-  async speak(text, { rate = 1, signal, lang = this.lang, voice = "aiden" } = {}) {
+  async speak(text, { rate = 1, signal, lang = this.lang, voice = "aiden", outputDeviceId = "" } = {}) {
     if (typeof text !== "string") throw new TypeError("Speech text must be a string.");
     if (!Number.isFinite(rate) || rate < 0.1 || rate > 3)
       throw new RangeError("Speech rate must be between 0.1 and 3.");
@@ -1555,6 +1811,8 @@ var QwenHttpSpeakingEngine = class {
       operation.url = this.g.URL.createObjectURL(blob);
       const audio = operation.audio = new this.g.Audio(operation.url);
       audio.playbackRate = rate;
+      if (outputDeviceId && typeof audio.setSinkId === "function")
+        await audio.setSinkId(outputDeviceId);
       await new Promise((resolve, reject) => {
         const done = () => {
           cleanup();
@@ -1854,7 +2112,7 @@ function createQwenSettings(React2) {
       h(
         "p",
         null,
-        "Host-wide settings for the resident Qwen3 ASR + TTS server. Only unauthenticated loopback HTTP is allowed; the browser reaches it through authenticated DSH routes."
+        "Host-wide settings for the Qwen3 ASR + TTS server. Enter any HTTP or HTTPS base URL reachable from the DSH host. The browser accesses it through authenticated DSH routes."
       ),
       field("Qwen API base URL", "baseUrl"),
       field("Request timeout (ms)", "timeoutMs", "number"),
@@ -1906,12 +2164,14 @@ function createComponents(React2) {
     };
     const paths = {
       mic: "M9 5a3 3 0 0 1 6 0v7a3 3 0 0 1-6 0V5M6 10v2a6 6 0 0 0 12 0v-2M12 18v4M8 22h8",
+      micOff: "M9 9v3a3 3 0 0 0 5.12 2.12M15 9V5a3 3 0 0 0-5.64-1.42M6 10v2a6 6 0 0 0 9.5 4.88M18 10v2a6 6 0 0 1-.5 2.4M12 18v4M8 22h8M3 3l18 18",
       speaker: "M3 9h4l6-5v16l-6-5H3V9M17 8a6 6 0 0 1 0 8M20 5a10 10 0 0 1 0 14",
       close: "M6 6l12 12M18 6L6 18",
       stop: "M6 6h12v12H6z",
       pause: "M8 5v14M16 5v14",
       play: "M7 4l13 8-13 8z",
       send: "M3 11.5L21 3l-8.5 18-2-7.5L3 11.5zm7.5 2L21 3",
+      queue: "M5 6h14M5 12h10M5 18h6M18 15v6M15 18h6",
       speakerOff: "M3 9h4l6-5v16l-6-5H3V9M17 9l5 6M22 9l-5 6"
     };
     return h("svg", common, h("path", { d: paths[name] || paths.mic }));
@@ -2052,7 +2312,7 @@ function createComponents(React2) {
     }, [controller, enabled]);
     return h("canvas", { ref, className: "dlv-wave", "aria-hidden": true });
   }
-  function RecordingBar({ controller }) {
+  function RecordingBar({ controller, questionOnly = false, overlay = false, overlayStyle }) {
     const state = useController(controller);
     const [now, setNow] = React2.useState(Date.now());
     React2.useEffect(() => {
@@ -2063,13 +2323,17 @@ function createComponents(React2) {
     }, [state.autoSendAt]);
     const [invoke, error, clearError] = useActions(controller);
     const capture = state.starting || state.listening || state.recognizing;
+    if (questionOnly && !state.answeringQuestion) return null;
     if (!state.conversation && !capture && !state.speaking && !state.paused && !state.error && !error)
       return null;
     const remaining = state.autoSendAt ? Math.max(1, Math.ceil((state.autoSendAt - now) / 1e3)) : null;
-    const status = remaining ? `Sending in ${remaining}\u2026` : state.starting ? "Starting microphone\u2026" : state.paused ? "Speech paused" : state.speaking ? "Speaking" : state.recognizing ? "Recognizing speech\u2026" : state.listening ? "Listening \u2014 waiting for speech" : state.conversation ? "Conversation idle" : "Voice ready";
+    const status = state.answeringQuestion && state.recognizing ? "Recognizing answer\u2026" : state.answeringQuestion && state.listening ? "Listening for your answer\u2026" : remaining ? `Sending in ${remaining}\u2026` : state.starting ? "Starting microphone\u2026" : state.paused ? "Speech paused" : state.speaking ? "Speaking" : state.recognizing ? "Recognizing speech\u2026" : state.listening ? "Listening \u2014 waiting for speech" : state.conversation ? "Conversation idle" : "Voice ready";
     return h(
       "div",
-      { className: "dlv-bar-wrap" },
+      {
+        className: overlay ? "dlv-bar-wrap dlv-question-overlay" : "dlv-bar-wrap",
+        style: overlay ? overlayStyle : void 0
+      },
       h(
         "div",
         { className: "dlv-pill", role: "group", "aria-label": "Voice controls" },
@@ -2087,22 +2351,23 @@ function createComponents(React2) {
         h("span", { className: "dlv-status", role: "status", "aria-live": "polite" }, status),
         h(Button, {
           className: "dlv-live-toggle",
-          label: "Automatic sending",
-          title: `Automatic sending: ${state.settings.sendingMode === "automatic" ? "on" : "off"}`,
-          icon: "send",
-          visibleLabel: state.settings.sendingMode === "automatic" ? "ON" : "OFF",
-          role: "switch",
-          "aria-checked": state.settings.sendingMode === "automatic",
+          label: "Automatic delivery mode",
+          title: `Automatic delivery: ${state.settings.sendingMode === "steer" ? "send to the running agent" : state.settings.sendingMode === "queue" ? "queue" : "off"}`,
+          icon: state.settings.sendingMode === "queue" ? "queue" : "send",
+          visibleLabel: state.settings.sendingMode === "steer" ? "SEND" : state.settings.sendingMode === "queue" ? "QUEUE" : "OFF",
+          "aria-label": `Automatic delivery: ${state.settings.sendingMode || "manual"}`,
+          "aria-pressed": ["queue", "steer"].includes(state.settings.sendingMode),
+          "data-mode": state.settings.sendingMode || "manual",
           onClick: () => invoke("updateSettings", {
-            sendingMode: state.settings.sendingMode === "automatic" ? "manual" : "automatic"
+            sendingMode: !["queue", "steer"].includes(state.settings.sendingMode) ? "queue" : state.settings.sendingMode === "queue" ? "steer" : "manual"
           })
         }),
         h(Button, {
-          className: "dlv-live-toggle",
+          className: `dlv-live-toggle${state.speechSegmentsRemaining > 0 ? " dlv-live-toggle-expanded" : ""}`,
           label: "Automatic assistant speech",
-          title: `Automatic assistant speech: ${state.settings.announceAssistantMessages !== false ? "on" : "off"}`,
+          title: `Automatic assistant speech: ${state.settings.announceAssistantMessages !== false ? "on" : "off"}${state.speechSegmentsRemaining > 0 ? ` \u2014 ${state.speechSegmentsRemaining} speech segment${state.speechSegmentsRemaining === 1 ? "" : "s"} remaining` : ""}`,
           icon: state.settings.announceAssistantMessages !== false ? "speaker" : "speakerOff",
-          visibleLabel: state.settings.announceAssistantMessages !== false ? "ON" : "OFF",
+          visibleLabel: state.settings.announceAssistantMessages === false ? "OFF" : state.speechSegmentsRemaining > 0 ? String(state.speechSegmentsRemaining) : "ON",
           role: "switch",
           "aria-checked": state.settings.announceAssistantMessages !== false,
           onClick: () => invoke("updateSettings", {
@@ -2120,9 +2385,14 @@ function createComponents(React2) {
           onClick: () => invoke("startConversation")
         }) : null,
         capture ? h(Button, {
-          label: "Stop listening",
-          icon: "stop",
-          onClick: () => invoke("stopListening")
+          className: "dlv-live-toggle dlv-mic-state",
+          label: state.muted ? "Resume listening" : "Ignore composer input",
+          title: `Microphone input: ${state.muted ? "ignoring" : "listening"}`,
+          icon: state.muted ? "micOff" : "mic",
+          visibleLabel: state.muted ? "IGNORING" : "LISTENING",
+          "aria-pressed": Boolean(state.muted),
+          "data-muted": state.muted ? "true" : "false",
+          onClick: () => invoke(state.muted ? "resumeListeningInput" : "muteListening")
         }) : null,
         state.speaking && !state.paused && state.capabilities[state.settings.engine]?.pause ? h(Button, {
           label: "Pause speech",
@@ -2164,6 +2434,57 @@ function createComponents(React2) {
     const [invoke, error, clearError] = useActions(controller);
     const settings = state.settings || {};
     const capabilities = state.capabilities || {};
+    const tabsId = React2.useId();
+    const tabs = [
+      { id: "conversation", label: "Conversation" },
+      { id: "speech", label: "Speech" },
+      { id: "recognition", label: "Speech recognition" }
+    ];
+    const tabRefs = React2.useRef([]);
+    const [activeTab, setActiveTab] = React2.useState("conversation");
+    const activateTab = (index) => {
+      const tab = tabs[index];
+      if (!tab) return;
+      setActiveTab(tab.id);
+      tabRefs.current[index]?.focus();
+    };
+    const handleTabKeyDown = (event, index) => {
+      let nextIndex;
+      if (event.key === "ArrowRight") nextIndex = (index + 1) % tabs.length;
+      else if (event.key === "ArrowLeft") nextIndex = (index - 1 + tabs.length) % tabs.length;
+      else if (event.key === "Home") nextIndex = 0;
+      else if (event.key === "End") nextIndex = tabs.length - 1;
+      else return;
+      event.preventDefault();
+      activateTab(nextIndex);
+    };
+    const [audioDevices, setAudioDevices] = React2.useState([]);
+    React2.useEffect(() => {
+      let active = true;
+      const refresh = async () => {
+        try {
+          const devices = await globalThis.navigator?.mediaDevices?.enumerateDevices?.();
+          if (active) setAudioDevices(Array.from(devices || []));
+        } catch {
+          if (active) setAudioDevices([]);
+        }
+      };
+      void refresh();
+      globalThis.navigator?.mediaDevices?.addEventListener?.("devicechange", refresh);
+      return () => {
+        active = false;
+        globalThis.navigator?.mediaDevices?.removeEventListener?.("devicechange", refresh);
+      };
+    }, []);
+    const deviceOptions = (kind, fallback) => [
+      { value: "", label: "System default" },
+      ...audioDevices.filter(
+        (device) => device.kind === kind && device.deviceId && device.deviceId !== "default"
+      ).map((device, index) => ({
+        value: device.deviceId,
+        label: device.label || `${fallback} ${index + 1}`
+      }))
+    ];
     const field = (label, key, options) => h(
       "label",
       { key },
@@ -2186,11 +2507,16 @@ function createComponents(React2) {
         )
       )
     );
-    const card = (title, children, open = false) => h(
-      "details",
-      { className: "dlv-settings-card", open },
-      h("summary", null, title),
-      h("div", { className: "dlv-settings-card-body" }, ...children)
+    const panel = (id2, children) => h(
+      "div",
+      {
+        id: `${tabsId}-panel-${id2}`,
+        className: "dlv-settings-panel",
+        role: "tabpanel",
+        "aria-labelledby": `${tabsId}-tab-${id2}`,
+        hidden: activeTab !== id2
+      },
+      ...children
     );
     const subcard = (title, children, open = false) => h(
       "details",
@@ -2204,9 +2530,41 @@ function createComponents(React2) {
       h("h3", null, "Live Voice"),
       onClose ? h(Button, { label: "Close voice settings", icon: "close", onClick: onClose }) : null,
       h(
-        "details",
-        { className: "dlv-settings-card" },
-        h("summary", null, "Speech output"),
+        "div",
+        { className: "dlv-settings-tabs", role: "tablist", "aria-label": "Live Voice settings" },
+        tabs.map((tab, index) => {
+          const selected = activeTab === tab.id;
+          return h(
+            "button",
+            {
+              key: tab.id,
+              ref: (element) => {
+                tabRefs.current[index] = element;
+              },
+              id: `${tabsId}-tab-${tab.id}`,
+              type: "button",
+              role: "tab",
+              className: "dlv-settings-tab",
+              "aria-selected": selected,
+              "aria-controls": `${tabsId}-panel-${tab.id}`,
+              "data-active": selected ? "true" : void 0,
+              tabIndex: selected ? 0 : -1,
+              onClick: () => setActiveTab(tab.id),
+              onKeyDown: (event) => handleTabKeyDown(event, index)
+            },
+            tab.label
+          );
+        })
+      ),
+      h(
+        "div",
+        {
+          id: `${tabsId}-panel-speech`,
+          className: "dlv-settings-panel",
+          role: "tabpanel",
+          "aria-labelledby": `${tabsId}-tab-speech`,
+          hidden: activeTab !== "speech"
+        },
         h(
           "div",
           { className: "dlv-settings-card-body" },
@@ -2227,22 +2585,22 @@ function createComponents(React2) {
               disabled: capabilities.browser?.supported === false
             }
           ]),
-          settings.engine === "browser" ? field("Local browser voice", "voice", [
-            { value: "", label: "Automatic local voice" },
-            ...(capabilities.browser?.voices || []).map((voice) => ({
-              value: voice.voiceURI || voice.name,
-              label: `${voice.name} \u2014 ${voice.lang || "unknown language"}`
-            }))
-          ]) : settings.engine === "say" ? h(
-            "label",
+          settings.engine !== "say" ? field("Output device", "outputDeviceId", deviceOptions("audiooutput", "Audio output")) : h("small", null, "macOS say uses the output selected on the DSH host."),
+          settings.engine === "browser" ? h(
+            React2.Fragment,
             null,
-            "macOS say voice (empty uses system default)",
-            h("input", {
-              type: "text",
-              value: settings.voice || "",
-              onChange: (event) => invoke("updateSettings", { voice: event.target.value }),
-              autoComplete: "off"
-            })
+            h(
+              "small",
+              null,
+              "Browser speech synthesis may ignore the selected output device; this browser API normally follows the system default."
+            ),
+            field("Local browser voice", "voice", [
+              { value: "", label: "Automatic local voice" },
+              ...(capabilities.browser?.voices || []).map((voice) => ({
+                value: voice.voiceURI || voice.name,
+                label: `${voice.name} \u2014 ${voice.lang || "unknown language"}`
+              }))
+            ])
           ) : null,
           settings.engine === "qwen-http" ? h(
             React2.Fragment,
@@ -2257,6 +2615,47 @@ function createComponents(React2) {
               h(QwenSettings, { key: "qwen-output-settings", controller })
             ])
           ) : null,
+          subcard("Filtering", [
+            h(
+              "label",
+              { key: "output-code-filter", className: "dlv-check" },
+              h("input", {
+                type: "checkbox",
+                checked: settings.outputCodeFilterEnabled !== false,
+                onChange: (event) => invoke("updateSettings", { outputCodeFilterEnabled: event.target.checked })
+              }),
+              " Filter Markdown code blocks before speaking"
+            ),
+            h(
+              "label",
+              { key: "output-code-lines" },
+              "Read code blocks up to this many lines",
+              h("input", {
+                type: "number",
+                min: 0,
+                max: 100,
+                value: settings.outputCodeMaxLines ?? 5,
+                disabled: settings.outputCodeFilterEnabled === false,
+                onChange: (event) => {
+                  const value = Number(event.target.value);
+                  if (Number.isInteger(value) && value >= 0 && value <= 100)
+                    invoke("updateSettings", { outputCodeMaxLines: value });
+                }
+              })
+            ),
+            h(
+              "label",
+              { key: "output-code-notice" },
+              "Replacement phrase for larger code blocks",
+              h("input", {
+                type: "text",
+                maxLength: 300,
+                value: settings.outputCodeNotice || "Look the code on out conversation",
+                disabled: settings.outputCodeFilterEnabled === false,
+                onChange: (event) => invoke("updateSettings", { outputCodeNotice: event.target.value })
+              })
+            )
+          ]),
           h(
             "label",
             null,
@@ -2279,13 +2678,48 @@ function createComponents(React2) {
             "p",
             null,
             "Qwen synthesis runs on the DSH host and the generated WAV plays in this browser. macOS say plays on the host; Browser speech plays on this device."
+          ),
+          ...["qwen-http", "say", "browser"].filter((id2) => capabilities[id2]?.supported === false).map(
+            (id2) => h(
+              "p",
+              { key: id2, role: "status" },
+              `${id2 === "qwen-http" ? "Qwen3 local" : id2 === "say" ? "macOS say" : "Browser speech"}: ${capabilities[id2].reason}`
+            )
+          ),
+          h(
+            "div",
+            { className: "dlv-settings-actions" },
+            h(
+              "button",
+              {
+                type: "button",
+                disabled: capabilities[settings.engine]?.supported !== true || state.speaking,
+                onClick: () => invoke("speak", "DSH Live Voice. The selected speech output is working.")
+              },
+              state.speaking ? "Testing speech\u2026" : "Test selected speech output"
+            ),
+            state.speaking || state.paused ? h(
+              "button",
+              { type: "button", onClick: () => invoke("stopSpeech") },
+              "Stop speech test"
+            ) : null,
+            h(
+              "button",
+              { type: "button", onClick: () => invoke("refreshCapabilities") },
+              "Refresh available engines"
+            )
           )
         )
       ),
       h(
-        "details",
-        { className: "dlv-settings-card" },
-        h("summary", null, "Speech recognition"),
+        "div",
+        {
+          id: `${tabsId}-panel-recognition`,
+          className: "dlv-settings-panel",
+          role: "tabpanel",
+          "aria-labelledby": `${tabsId}-tab-recognition`,
+          hidden: activeTab !== "recognition"
+        },
         h(
           "div",
           { className: "dlv-settings-card-body" },
@@ -2307,6 +2741,12 @@ function createComponents(React2) {
               h("option", { value: "whisper-http" }, "Whisper HTTP \u2014 DSH host")
             )
           ),
+          field("Input device", "inputDeviceId", deviceOptions("audioinput", "Microphone")),
+          settings.recognitionEngine === "browser" ? h(
+            "small",
+            null,
+            "Browser SpeechRecognition may use the browser or system default microphone instead of this selection."
+          ) : null,
           field("Recognition language", "recognitionLang", [
             ...settings.recognitionEngine !== "browser" ? [{ value: "auto", label: "Automatic \u2014 detect language" }] : [],
             { value: "pt-BR", label: "Portugu\xEAs (Brasil)" },
@@ -2351,6 +2791,89 @@ function createComponents(React2) {
             h(QwenSettings, { key: "qwen-recognition-settings", controller })
           ]) : null,
           h("p", null, "Provider settings change with the selected recognition engine."),
+          subcard("Voice commands", [
+            h(
+              "label",
+              { key: "voice-commands-enabled", className: "dlv-check" },
+              h("input", {
+                type: "checkbox",
+                checked: settings.voiceCommandsEnabled !== false,
+                onChange: (event) => invoke("updateSettings", { voiceCommandsEnabled: event.target.checked })
+              }),
+              " Enable exact voice commands"
+            ),
+            h(
+              "small",
+              { key: "voice-command-help" },
+              "Separate phrases with commas. Matching ignores capitalization, accents, punctuation, and extra spaces. The entire final chunk must match."
+            ),
+            ...[
+              ["Send to running agent", "voiceCommandSend", "send, send message"],
+              ["Put in queue", "voiceCommandQueue", "queue, queue message"],
+              ["End conversation", "voiceCommandEnd", "end, end conversation"],
+              ["Mute composer input", "voiceCommandMute", "mute, stop listening"],
+              ["Resume composer input", "voiceCommandResume", "resume, start listening"],
+              [
+                "Stop assistant speech",
+                "voiceCommandStopSpeaking",
+                "stop talking, stop speaking, shut up"
+              ],
+              ["Clear composer", "voiceCommandClear", "clear all, clear message"]
+            ].map(
+              ([label, key, fallback]) => h(
+                "label",
+                { key },
+                label,
+                h("textarea", {
+                  rows: 2,
+                  maxLength: 1e3,
+                  defaultValue: settings[key] || fallback,
+                  disabled: settings.voiceCommandsEnabled === false,
+                  onBlur: (event) => invoke("updateSettings", { [key]: event.target.value })
+                })
+              )
+            )
+          ]),
+          subcard("Filtering", [
+            h(
+              "label",
+              { key: "recognition-filter", className: "dlv-check" },
+              h("input", {
+                type: "checkbox",
+                checked: settings.recognitionFilterEnabled !== false,
+                onChange: (event) => invoke("updateSettings", { recognitionFilterEnabled: event.target.checked })
+              }),
+              " Ignore short final transcription chunks"
+            ),
+            h(
+              "label",
+              { key: "recognition-minimum-words" },
+              "Minimum words per final chunk",
+              h("input", {
+                type: "number",
+                min: 1,
+                max: 20,
+                value: settings.recognitionMinimumWords ?? 2,
+                disabled: settings.recognitionFilterEnabled === false,
+                onChange: (event) => {
+                  const value = Number(event.target.value);
+                  if (Number.isInteger(value) && value >= 1 && value <= 20)
+                    invoke("updateSettings", { recognitionMinimumWords: value });
+                }
+              }),
+              h(
+                "small",
+                null,
+                "Final chunks with fewer words are ignored before they reach the composer or automatic delivery."
+              )
+            )
+          ]),
+          capabilities.capture?.supported === false ? h("p", { role: "status" }, `Microphone: ${capabilities.capture.reason}`) : capabilities.capture?.permission === "prompt" ? h(
+            "p",
+            { role: "status" },
+            "Microphone permission will be requested only when you start dictation or a voice conversation."
+          ) : null,
+          capabilities.recognition?.supported === false ? h("p", { role: "status" }, capabilities.recognition.reason) : null,
           usesPluginVoiceDetection(settings.recognitionEngine) ? h(
             "details",
             { className: "dlv-settings-subcard", "aria-label": "Silence detection settings" },
@@ -2399,135 +2922,100 @@ function createComponents(React2) {
           ) : null
         )
       ),
-      card(
-        "Conversation",
-        [
-          h(
-            "label",
-            { key: "announce", className: "dlv-check" },
-            h("input", {
-              type: "checkbox",
-              checked: settings.announceAssistantMessages !== false,
-              onChange: (event) => invoke("updateSettings", { announceAssistantMessages: event.target.checked })
-            }),
-            " Automatically speak new assistant messages"
-          ),
-          h(
-            "p",
-            { key: "policy" },
-            "During a voice conversation, assistant phrases are announced automatically. Playback waits while you are speaking."
-          ),
-          h(
-            "label",
-            { key: "interrupt-message", className: "dlv-check" },
-            h("input", {
-              type: "checkbox",
-              checked: settings.interruptSpeechOnUserMessage === true,
-              onChange: (event) => invoke("updateSettings", { interruptSpeechOnUserMessage: event.target.checked })
-            }),
-            " Stop assistant speech when I send a message"
-          ),
+      panel("conversation", [
+        h(
+          "label",
+          { key: "announce", className: "dlv-check" },
+          h("input", {
+            type: "checkbox",
+            checked: settings.announceAssistantMessages !== false,
+            onChange: (event) => invoke("updateSettings", { announceAssistantMessages: event.target.checked })
+          }),
+          " Automatically speak new assistant messages"
+        ),
+        h(
+          "p",
+          { key: "policy" },
+          "During a voice conversation, assistant phrases are announced automatically. Playback waits while you are speaking."
+        ),
+        h(
+          "label",
+          { key: "interrupt-message", className: "dlv-check" },
+          h("input", {
+            type: "checkbox",
+            checked: settings.interruptSpeechOnUserMessage === true,
+            onChange: (event) => invoke("updateSettings", { interruptSpeechOnUserMessage: event.target.checked })
+          }),
+          " Stop assistant speech when I send a message"
+        ),
+        h(
+          "p",
+          { key: "interrupt-message-description", className: "dlv-setting-description" },
+          settings.interruptSpeechOnUserMessage ? "Sending or steering a new user message stops current or paused assistant speech." : "Sending another message does not stop the assistant audio you are already hearing."
+        ),
+        field("Listening mode", "mode", [
+          { value: "speaker", label: "Speakers \u2014 gated listening" },
+          { value: "headphones", label: "Headphones \u2014 open microphone" }
+        ]),
+        h(
+          "p",
+          { key: "mode-description", className: "dlv-setting-description" },
+          settings.mode === "headphones" ? "Open microphone keeps listening while responses play. When your speech is detected, playback pauses and resumes only when you choose." : "Gated listening releases the microphone while responses play, preventing speaker audio from being recognized. Use Take microphone to interrupt."
+        ),
+        h(
+          "label",
+          { key: "speech-delay" },
+          "Assistant response delay",
           h(
-            "p",
-            { key: "interrupt-message-description", className: "dlv-setting-description" },
-            settings.interruptSpeechOnUserMessage ? "Sending or steering a new user message stops current or paused assistant speech." : "Sending another message does not stop the assistant audio you are already hearing."
+            "select",
+            {
+              value: String(settings.assistantSpeechDelaySeconds || 3),
+              onChange: (event) => invoke("updateSettings", {
+                assistantSpeechDelaySeconds: Number(event.target.value)
+              })
+            },
+            [1, 2, 3, 4, 5, 6, 8, 10].map(
+              (seconds) => h("option", { key: seconds, value: String(seconds) }, seconds + " seconds")
+            )
           ),
-          field("Listening mode", "mode", [
-            { value: "speaker", label: "Speakers \u2014 gated listening" },
-            { value: "headphones", label: "Headphones \u2014 open microphone" }
-          ]),
           h(
-            "p",
-            { key: "mode-description", className: "dlv-setting-description" },
-            settings.mode === "headphones" ? "Open microphone keeps listening while responses play. When your speech is detected, playback pauses and resumes only when you choose." : "Gated listening releases the microphone while responses play, preventing speaker audio from being recognized. Use Take microphone to interrupt."
-          ),
+            "small",
+            null,
+            "After you stop speaking, automatic assistant playback waits for this much continuous silence. Speaking again restarts the wait."
+          )
+        ),
+        field("Sending mode", "sendingMode", [
+          { value: "manual", label: "Off \u2014 review and send manually" },
+          { value: "queue", label: "Queue \u2014 automatically add after silence" },
+          { value: "steer", label: "Steer \u2014 automatically send to the running agent" }
+        ]),
+        settings.sendingMode !== "manual" ? h(
+          "label",
+          { key: "delay" },
+          "Send after silence",
           h(
-            "label",
-            { key: "speech-delay" },
-            "Assistant response delay",
-            h(
-              "select",
-              {
-                value: String(settings.assistantSpeechDelaySeconds || 3),
-                onChange: (event) => invoke("updateSettings", {
-                  assistantSpeechDelaySeconds: Number(event.target.value)
-                })
-              },
-              [1, 2, 3, 4, 5, 6, 8, 10].map(
-                (seconds) => h("option", { key: seconds, value: String(seconds) }, seconds + " seconds")
-              )
-            ),
-            h(
-              "small",
-              null,
-              "After you stop speaking, automatic assistant playback waits for this much continuous silence. Speaking again restarts the wait."
+            "select",
+            {
+              value: String(settings.autoSendDelaySeconds || 4),
+              onChange: (event) => invoke("updateSettings", {
+                autoSendDelaySeconds: Number(event.target.value)
+              })
+            },
+            [2, 3, 4, 5, 6, 8, 10].map(
+              (seconds) => h("option", { key: seconds, value: String(seconds) }, seconds + " seconds")
             )
           ),
-          field("Sending mode", "sendingMode", [
-            { value: "manual", label: "Manual \u2014 review and send" },
-            { value: "automatic", label: "Automatic \u2014 send after silence" }
-          ]),
-          settings.sendingMode === "automatic" ? h(
-            "label",
-            { key: "delay" },
-            "Send after silence",
-            h(
-              "select",
-              {
-                value: String(settings.autoSendDelaySeconds || 4),
-                onChange: (event) => invoke("updateSettings", {
-                  autoSendDelaySeconds: Number(event.target.value)
-                })
-              },
-              [2, 3, 4, 5, 6, 8, 10].map(
-                (seconds) => h("option", { key: seconds, value: String(seconds) }, seconds + " seconds")
-              )
-            ),
-            h(
-              "small",
-              null,
-              "Countdown starts after a final recognized phrase. New speech or edits cancel it."
-            )
-          ) : h(
-            "p",
-            { key: "manual", className: "dlv-setting-description" },
-            "Recognized text stays in the composer until you use the normal DSH Send control."
+          h(
+            "small",
+            null,
+            "Countdown starts after a final recognized phrase. New speech or edits cancel it."
           )
-        ],
-        false
-      ),
-      capabilities.capture?.supported === false ? h("p", { role: "status" }, `Microphone: ${capabilities.capture.reason}`) : capabilities.capture?.permission === "prompt" ? h(
-        "p",
-        { role: "status" },
-        "Microphone permission will be requested only when you start dictation or a voice conversation."
-      ) : null,
-      capabilities.recognition?.supported === false ? h("p", { role: "status" }, capabilities.recognition.reason) : null,
-      ...["qwen-http", "say", "browser"].filter((id2) => capabilities[id2]?.supported === false).map(
-        (id2) => h(
+        ) : h(
           "p",
-          { key: id2, role: "status" },
-          `${id2 === "qwen-http" ? "Qwen3 local" : id2 === "say" ? "macOS say" : "Browser speech"}: ${capabilities[id2].reason}`
-        )
-      ),
-      h(
-        "div",
-        { className: "dlv-settings-actions" },
-        h(
-          "button",
-          {
-            type: "button",
-            disabled: capabilities[settings.engine]?.supported !== true || state.speaking,
-            onClick: () => invoke("speak", "DSH Live Voice. The selected speech output is working.")
-          },
-          state.speaking ? "Testing speech\u2026" : "Test selected speech output"
-        ),
-        state.speaking || state.paused ? h("button", { type: "button", onClick: () => invoke("stopSpeech") }, "Stop speech test") : null,
-        h(
-          "button",
-          { type: "button", onClick: () => invoke("refreshCapabilities") },
-          "Refresh available engines"
+          { key: "manual", className: "dlv-setting-description" },
+          "Recognized text stays in the composer until you use the normal DSH Send control."
         )
-      ),
+      ]),
       h(ErrorText, {
         error: error || state.error,
         onDismiss: () => {
@@ -2552,23 +3040,27 @@ var styles = `
 .dlv-icon-button:disabled{opacity:.4;cursor:default}
 .dlv-icon-button:focus-visible,.dlv-settings :is(input,select):focus-visible{outline:2px solid var(--dsw-alias-label-primary);outline-offset:3px}
 .dlv-bar-wrap{width:100%;min-width:0}
+.dlv-question-overlay{position:fixed;z-index:10000;bottom:8px;transform:translateX(-50%);max-width:calc(100vw - 32px);pointer-events:none}
+.dlv-question-overlay .dlv-pill{pointer-events:auto}
 .dlv-pill{display:flex;align-items:center;box-sizing:border-box;gap:10px;min-height:52px;border-radius:26px;background:var(--dsw-alias-bg-layer-1);border:1px solid var(--dsw-alias-border-l1);padding:0 14px;width:100%;max-width:720px;margin:0 auto;box-shadow:0 8px 24px rgba(0,0,0,.18)}
 .dlv-pill-button{width:34px;height:34px;border-radius:50%;border:1px solid var(--dsw-alias-border-l2);color:var(--dsw-alias-label-primary)}
 .dlv-pill-button:hover{background:var(--dsw-alias-bg-layer-2)}
-.dlv-live-toggle{width:auto;min-width:58px;height:34px;padding:0 9px;gap:5px;border:1px solid var(--dsw-alias-border-l2);border-radius:17px;color:var(--dsw-alias-label-primary)}
-.dlv-live-toggle svg{width:18px;height:18px}
-.dlv-toggle-state{min-width:22px;font-size:10px;font-weight:700;line-height:1;letter-spacing:.04em;text-align:left}
-.dlv-live-toggle:hover{background:var(--dsw-alias-bg-layer-2)}
-.dlv-live-toggle[aria-checked=false]{color:var(--dsw-alias-label-tertiary);border-color:var(--dsw-alias-border-l1);opacity:.72}
+.dlv-live-toggle{width:34px;height:34px;padding:0 7px;gap:0;overflow:hidden;border:1px solid var(--dsw-alias-border-l2);border-radius:17px;color:var(--dsw-alias-label-primary);transition:width .16s ease,gap .16s ease}
+.dlv-live-toggle svg{width:18px;height:18px;flex:none}
+.dlv-toggle-state{max-width:0;opacity:0;overflow:hidden;white-space:nowrap;font-size:10px;font-weight:700;line-height:1;letter-spacing:.04em;text-align:left;transition:max-width .16s ease,opacity .12s ease}
+.dlv-live-toggle:hover,.dlv-live-toggle:focus-visible,.dlv-live-toggle-expanded{width:78px;gap:5px;background:var(--dsw-alias-bg-layer-2)}
+.dlv-live-toggle:hover .dlv-toggle-state,.dlv-live-toggle:focus-visible .dlv-toggle-state,.dlv-live-toggle-expanded .dlv-toggle-state{max-width:38px;opacity:1}
+.dlv-live-toggle[aria-checked=false],.dlv-live-toggle[data-mode=manual],.dlv-mic-state[data-muted=true]{color:var(--dsw-alias-label-tertiary);border-color:var(--dsw-alias-border-l1);opacity:.55}
+.dlv-mic-state:hover,.dlv-mic-state:focus-visible{width:102px}.dlv-mic-state:hover .dlv-toggle-state,.dlv-mic-state:focus-visible .dlv-toggle-state{max-width:64px}
 .dlv-live-toggle[aria-checked=true]{background:var(--dsw-alias-bg-layer-2)}
 .dlv-wave{display:block;flex:1 1 180px;min-width:30px;width:100%;height:40px;color:var(--dsw-alias-label-primary)}
 .dlv-status{flex:1 1 120px;min-width:0;font-size:13px;line-height:1.4;color:var(--dsw-alias-label-secondary)}
 .dlv-error{color:var(--dsw-alias-state-error-primary);overflow-wrap:anywhere;font-size:13px;max-width:720px;margin:8px auto}
 .dlv-settings{box-sizing:border-box;padding:18px;width:100%;display:grid;gap:16px;max-width:640px;color:var(--dsw-alias-label-primary)}
 .dlv-settings h3,.dlv-settings p{margin:0}
-.dlv-settings-group,.dlv-settings-card{margin:0;border:1px solid var(--dsw-alias-border-l1);border-radius:10px;min-width:0}.dlv-settings-group{display:grid;gap:14px;padding:16px}.dlv-settings-group legend{padding:0 6px;font-weight:600;color:var(--dsw-alias-label-primary)}.dlv-settings-card>summary,.dlv-settings-subcard>summary{cursor:pointer;font-weight:650;list-style:none;display:flex;align-items:center;justify-content:space-between;padding:12px}.dlv-settings-card>summary::-webkit-details-marker,.dlv-settings-subcard>summary::-webkit-details-marker{display:none}.dlv-settings-card>summary:after,.dlv-settings-subcard>summary:after{content:"\u203A";transform:rotate(90deg);transition:transform .15s}.dlv-settings-card:not([open])>summary:after,.dlv-settings-subcard:not([open])>summary:after{transform:rotate(0)}.dlv-settings-card-body{display:grid;gap:12px;padding:0 12px 12px}.dlv-settings-subcard{border:1px solid var(--dsw-alias-border-l2);border-radius:8px;min-width:0}.dlv-settings-subcard>summary{padding:10px}.dlv-settings-subcard-body{display:grid;gap:10px;padding:0 10px 10px}
+.dlv-settings-tabs{border-bottom:.5px solid var(--dsw-alias-border-l2);display:flex;align-items:flex-end;gap:22px;overflow-x:auto}.dlv-settings-tab{position:relative;flex:none;padding:7px 1px 9px;border:0;background:transparent;color:var(--dsw-alias-label-tertiary);font:inherit;font-size:13px;line-height:20px;cursor:pointer}.dlv-settings-tab:hover,.dlv-settings-tab[data-active=true]{color:var(--dsw-alias-label-primary)}.dlv-settings-tab[data-active=true]:after,.dlv-settings-tab:focus-visible:after{content:"";position:absolute;right:0;bottom:-1px;left:0;height:2px;border-radius:2px 2px 0 0;background:var(--dsw-alias-label-primary)}.dlv-settings-tab:focus-visible{border-radius:2px;outline:2px solid var(--dsw-alias-state-business-primary);outline-offset:2px;color:var(--dsw-alias-label-primary)}.dlv-settings-panel{min-width:0;padding-top:2px}.dlv-settings-panel[hidden]{display:none}.dlv-settings-group{margin:0;border:1px solid var(--dsw-alias-border-l1);border-radius:10px;min-width:0;display:grid;gap:14px;padding:16px}.dlv-settings-group legend{padding:0 6px;font-weight:600;color:var(--dsw-alias-label-primary)}.dlv-settings-card-body{display:grid;gap:12px;padding:12px 0}.dlv-settings-subcard{border:1px solid var(--dsw-alias-border-l2);border-radius:8px;min-width:0}.dlv-settings-subcard>summary{cursor:pointer;font-weight:650;list-style:none;display:flex;align-items:center;justify-content:space-between;padding:10px}.dlv-settings-subcard>summary::-webkit-details-marker{display:none}.dlv-settings-subcard>summary:after{content:"\u203A";transform:rotate(90deg);transition:transform .15s}.dlv-settings-subcard:not([open])>summary:after{transform:rotate(0)}.dlv-settings-subcard-body{display:grid;gap:10px;padding:0 10px 10px}
 .dlv-settings label{display:grid;gap:6px;font-size:14px}.dlv-settings label.dlv-check{display:flex;align-items:center;gap:8px}.dlv-settings label.dlv-check input{width:auto}
-.dlv-settings :is(input,select){box-sizing:border-box;width:100%;padding:8px;border:1px solid var(--dsw-alias-border-l2);border-radius:8px;background:var(--dsw-alias-bg-layer-1);color:inherit;font:inherit}
+.dlv-settings :is(input,select,textarea){box-sizing:border-box;width:100%;padding:8px;border:1px solid var(--dsw-alias-border-l2);border-radius:8px;background:var(--dsw-alias-bg-layer-1);color:inherit;font:inherit}
 .dlv-settings small,.dlv-settings p{font-size:13px;color:var(--dsw-alias-label-secondary);line-height:1.5}
 .dlv-preset-group{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:8px}.dlv-settings label.dlv-preset{display:flex;align-items:flex-start;gap:8px;padding:10px;border:1px solid var(--dsw-alias-border-l1);border-radius:8px;cursor:pointer}.dlv-settings label.dlv-preset:has(input:checked){border-color:var(--dsw-alias-label-primary);background:var(--dsw-alias-bg-layer-1)}.dlv-settings label.dlv-preset input{width:auto;margin-top:3px}.dlv-preset span{display:grid;gap:3px}.dlv-vad-summary{font-weight:500}
 .dlv-settings-actions{display:flex;flex-wrap:wrap;gap:8px}.dlv-settings-actions button{padding:8px 12px;border:1px solid var(--dsw-alias-border-l2);border-radius:8px;background:var(--dsw-alias-bg-layer-1);color:inherit;cursor:pointer}.dlv-settings-actions button:disabled{opacity:.45;cursor:default}
@@ -2593,6 +3085,11 @@ function assistantMessages(snapshot) {
     };
   }).sort((a, b) => a.turn - b.turn || a.step - b.step);
 }
+function pendingQuestionSpeech(interaction, index = 0) {
+  if (!interaction || interaction.kind !== "question" || !Array.isArray(interaction.questions)) return "";
+  const question = interaction.questions[index]?.question;
+  return typeof question === "string" ? question.trim() : "";
+}
 function latestUserSequence(snapshot) {
   return Math.max(
     -1,
@@ -2606,13 +3103,14 @@ function addressedTurn(messages, messageId) {
 }
 
 // src/client/index.ts
-var inject = ["slots", "connection", "uiConversation"];
+var inject = ["slots", "connection", "uiConversation", "uiSession"];
 function apply(ctx) {
   const e = import_react.default.createElement;
   const { MicrophoneButtons, RecordingBar, SpeakButton, SettingsPanel } = createComponents(import_react.default);
   const controllers = /* @__PURE__ */ new Map();
   const retiring = /* @__PURE__ */ new Set();
   let disposed = false;
+  let voiceModeActive = false;
   const ownership = new VoiceOwnership();
   const recognitionFor = (settings, meter) => settings.recognitionEngine === "qwen-http" ? new QwenHttpRecognitionEngine({
     meter,
@@ -2635,6 +3133,10 @@ function apply(ctx) {
     entry.composers.clear();
     entry.unsubscribe?.();
     entry.unsubscribe = null;
+    entry.unsubscribePendingQuestion?.();
+    entry.unsubscribePendingQuestion = null;
+    entry.questionCapture = null;
+    entry.controller.patch({ answeringQuestion: false });
     entry.chatListeners.clear();
     if (controllers.get(entry.key) === entry) controllers.delete(entry.key);
     const done = run(entry.controller, entry.controller.dispose());
@@ -2674,9 +3176,32 @@ function apply(ctx) {
       meter,
       composer: {
         getDraft: () => entry.draft,
-        submit: () => {
+        submit: (mode = "queue") => {
           const owner = [...entry.composers.values()].at(-1);
-          owner?.actions.submit?.();
+          owner?.actions.submit?.(mode);
+        },
+        handleQuestionResult: ({ final, interim }) => {
+          const capture = entry.questionCapture;
+          if (!capture || capture.answering || !final && !interim) return false;
+          if (final?.trim()) {
+            capture.answering = true;
+            capture.answers.push({
+              id: capture.interaction.questions[capture.index].id,
+              selected: [],
+              custom: final.trim()
+            });
+            capture.index++;
+            if (capture.index < capture.interaction.questions.length) {
+              capture.answering = false;
+              const text = pendingQuestionSpeech(capture.interaction, capture.index);
+              if (text) run(entry.controller, entry.controller.speak(text, capture.interaction.key));
+            } else {
+              entry.questionCapture = null;
+              entry.controller.patch({ answeringQuestion: false });
+              run(entry.controller, capture.interaction.answer({ answers: capture.answers }));
+            }
+          }
+          return true;
         },
         setDraft: (text) => {
           if (disposed || entry.closed) return;
@@ -2717,6 +3242,30 @@ function apply(ctx) {
       refresh();
       for (const listener of entry.chatListeners) listener();
     });
+    const pendingInteractions = ctx.uiSession.pendingInteractions;
+    const refreshPendingQuestion = (baseline = false) => {
+      if (disposed || entry.closed) return;
+      const interaction = pendingInteractions.getSnapshot().get(sessionId);
+      const key2 = interaction?.kind === "question" ? interaction.key : null;
+      if (!key2 || entry.questionCapture && entry.questionCapture.interaction.key !== key2) {
+        entry.questionCapture = null;
+        controller.patch({ answeringQuestion: false });
+      }
+      if (baseline) {
+        entry.pendingQuestionKey = key2;
+        return;
+      }
+      if (!key2 || key2 === entry.pendingQuestionKey) return;
+      entry.pendingQuestionKey = key2;
+      const text = pendingQuestionSpeech(interaction);
+      if (text && controller.getSnapshot().conversation) {
+        entry.questionCapture = { interaction, index: 0, answers: [], answering: false };
+        controller.patch({ answeringQuestion: true });
+        run(controller, controller.speak(text, key2));
+      }
+    };
+    refreshPendingQuestion(true);
+    entry.unsubscribePendingQuestion = pendingInteractions.subscribe(refreshPendingQuestion);
     const update = controller.updateSettings.bind(controller);
     controller.updateSettings = (next) => {
       if (disposed || entry.closed) return;
@@ -2736,6 +3285,7 @@ function apply(ctx) {
       const original = controller[method].bind(controller);
       controller[method] = (...args) => {
         entry.request++;
+        if (method === "endConversation" && args[0] !== true) voiceModeActive = false;
         return original(...args);
       };
     }
@@ -2744,6 +3294,7 @@ function apply(ctx) {
       controller[method] = (...args) => {
         if (disposed || entry.closed || !entry.refs) return Promise.resolve();
         if (method !== "speak" && !entry.composers.size) return Promise.resolve();
+        if (method === "startConversation") voiceModeActive = true;
         const request = ++entry.request;
         return ownership.run(
           controller,
@@ -2787,19 +3338,14 @@ function apply(ctx) {
       if (!entry || entry.closed || disposed) return;
       return () => {
         entry.composers.delete(token.current);
-        if (!entry.composers.size) {
-          entry.request++;
-          queueMicrotask(() => {
-            if (!entry.closed && !entry.composers.size)
-              run(entry.controller, entry.controller.endConversation());
-          });
-        }
       };
     }, [entry]);
     import_react.default.useLayoutEffect(() => {
       if (!entry || entry.closed || disposed) return;
       if (!input || typeof props.inputActions?.setDraft !== "function") return;
       entry.composers.set(token.current, { actions: props.inputActions });
+      if (voiceModeActive && !entry.controller.getSnapshot().conversation)
+        run(entry.controller, entry.controller.startConversation());
     }, [entry, input, props.inputActions]);
     import_react.default.useLayoutEffect(() => {
       if (!entry || entry.closed || disposed || !input) return;
@@ -2900,6 +3446,47 @@ function apply(ctx) {
     useComposer(entry, props);
     return entry ? e(RecordingBar, { controller: entry.controller }) : null;
   }
+  function QuestionStatusView({ entry }) {
+    const snapshot = import_react.default.useSyncExternalStore(
+      entry.controller.subscribe,
+      entry.controller.getSnapshot
+    );
+    const [overlayStyle, setOverlayStyle] = import_react.default.useState();
+    import_react.default.useLayoutEffect(() => {
+      if (!snapshot.answeringQuestion) return;
+      const seat = document.querySelector("[data-composer-seat]");
+      if (!seat) return;
+      const update = () => {
+        const rect = seat.getBoundingClientRect();
+        setOverlayStyle({
+          left: rect.left + rect.width / 2,
+          width: Math.min(720, Math.max(0, rect.width - 32))
+        });
+      };
+      update();
+      const observer = typeof ResizeObserver === "function" ? new ResizeObserver(update) : null;
+      observer?.observe(seat);
+      window.addEventListener("resize", update);
+      return () => {
+        observer?.disconnect();
+        window.removeEventListener("resize", update);
+      };
+    }, [snapshot.answeringQuestion]);
+    const target = typeof document === "undefined" ? null : document.body;
+    return snapshot.answeringQuestion && target && overlayStyle ? (0, import_react_dom.createPortal)(
+      e(RecordingBar, {
+        controller: entry.controller,
+        questionOnly: true,
+        overlay: true,
+        overlayStyle
+      }),
+      target
+    ) : null;
+  }
+  function QuestionStatus(props) {
+    const entry = useEntry(props.sessionId, "question-status");
+    return entry ? e(QuestionStatusView, { entry }) : null;
+  }
   function ActionView({ entry, messageId }) {
     const snapshot = import_react.default.useSyncExternalStore(
       entry.controller.subscribe,
@@ -2934,6 +3521,7 @@ function apply(ctx) {
   for (const [name, id2, order, component] of [
     ["conversation.input.right", "live-voice-controls", 6, Buttons],
     ["conversation.input.dock", "live-voice-status", -100, Dock],
+    ["conversation.session.header.utilities", "live-voice-question-status", 100, QuestionStatus],
     ["conversation.chat.assistant-actions", "live-voice-speak", 5, Action]
   ])
     ctx.slots.inject(