dsh-live-voice 0.2.1 → 0.2.3

package/lib/client.js

@@ -114,7 +114,9 @@ var defaultSettings = Object.freeze({
   recognitionProcessLocally: true,
   recognitionAutoInstall: true,
   voiceDetectionPreset: "natural",
+  recognitionMaxUtteranceSeconds: 60,
   microphoneEnabled: true,
+  holdToTalkEnabled: true,
   announceAssistantMessages: true,
   interruptSpeechOnUserMessage: false,
   sendingMode: "manual",
@@ -150,7 +152,9 @@ function normalizeSettings(value) {
     recognitionProcessLocally: typeof source.recognitionProcessLocally === "boolean" ? source.recognitionProcessLocally : defaultSettings.recognitionProcessLocally,
     recognitionAutoInstall: typeof source.recognitionAutoInstall === "boolean" ? source.recognitionAutoInstall : defaultSettings.recognitionAutoInstall,
     voiceDetectionPreset: Object.hasOwn(voiceDetectionPresets, source.voiceDetectionPreset) ? source.voiceDetectionPreset : defaultSettings.voiceDetectionPreset,
+    recognitionMaxUtteranceSeconds: Number.isInteger(source.recognitionMaxUtteranceSeconds) && source.recognitionMaxUtteranceSeconds >= 10 && source.recognitionMaxUtteranceSeconds <= 300 ? source.recognitionMaxUtteranceSeconds : defaultSettings.recognitionMaxUtteranceSeconds,
     microphoneEnabled: typeof source.microphoneEnabled === "boolean" ? source.microphoneEnabled : defaultSettings.microphoneEnabled,
+    holdToTalkEnabled: typeof source.holdToTalkEnabled === "boolean" ? source.holdToTalkEnabled : defaultSettings.holdToTalkEnabled,
     announceAssistantMessages: typeof source.announceAssistantMessages === "boolean" ? source.announceAssistantMessages : defaultSettings.announceAssistantMessages,
     interruptSpeechOnUserMessage: typeof source.interruptSpeechOnUserMessage === "boolean" ? source.interruptSpeechOnUserMessage : defaultSettings.interruptSpeechOnUserMessage,
     sendingMode: source.sendingMode === "automatic" ? "queue" : ["manual", "queue", "steer"].includes(source.sendingMode) ? source.sendingMode : defaultSettings.sendingMode,
@@ -273,6 +277,7 @@ var VoiceCoordinator = class {
       conversation: false,
       listening: false,
       recognizing: false,
+      pendingTranscriptions: 0,
       muted: normalizeSettings(settings).microphoneEnabled === false,
       speaking: false,
       paused: false,
@@ -287,6 +292,7 @@ var VoiceCoordinator = class {
     };
     this.autoSendTimer = null;
     this.autoSendDraft = null;
+    this.holdToTalkRelease = false;
     this.interruptionTimer = null;
     this.interruptionTranscriptConfirmed = false;
     this.interruptionPausedSpeech = false;
@@ -404,6 +410,27 @@ var VoiceCoordinator = class {
   startDictation() {
     return this.startListening(false);
   }
+  startHoldToTalk() {
+    this.holdToTalkRelease = false;
+    return this.startListening(false);
+  }
+  async releaseHoldToTalk() {
+    if (this.disposed || !this.snapshot.listening && !this.snapshot.starting) return;
+    this.holdToTalkRelease = true;
+    if (this.snapshot.starting) return;
+    await this.recognition.finish?.();
+    if (this.snapshot.pendingTranscriptions === 0) this._finishHoldToTalk();
+  }
+  _finishHoldToTalk() {
+    if (!this.holdToTalkRelease) return;
+    this.holdToTalkRelease = false;
+    const draft = this.composer.getDraft();
+    if (!draft.trim()) {
+      void this.stopListening();
+      return;
+    }
+    this.scheduleAutoSend(draft, { force: true, stopAfter: true });
+  }
   startConversation() {
     return this.startListening(true);
   }
@@ -483,6 +510,14 @@ var VoiceCoordinator = class {
           onResult: (result) => {
             if (valid()) this.onResult(result);
           },
+          onProcessingChange: ({ pending = 0 } = {}) => {
+            if (!valid()) return;
+            const count = Number.isSafeInteger(pending) && pending >= 0 ? pending : 0;
+            this.patch({ pendingTranscriptions: count });
+            if (count > 0) this.cancelAutoSend();
+            else if (this.holdToTalkRelease) this._finishHoldToTalk();
+            else this.maybeScheduleAutoSend();
+          },
           onActivity: (active) => {
             if (!valid()) return;
             if (active) {
@@ -507,6 +542,7 @@ var VoiceCoordinator = class {
           return;
         }
         this.patch({ listening: true, starting: false });
+        if (this.holdToTalkRelease) void this.releaseHoldToTalk();
         this._drain();
       } catch (error) {
         try {
@@ -619,7 +655,7 @@ var VoiceCoordinator = class {
       }
       const next = this.transcript.update(this.composer.getDraft(), final, true);
       this.composer.setDraft(next);
-      if (this.snapshot.settings.sendingMode !== "manual") this.scheduleAutoSend(next);
+      this.maybeScheduleAutoSend(next);
     }
     if (interim) {
       this.cancelAutoSend();
@@ -630,7 +666,12 @@ var VoiceCoordinator = class {
     this.patch({ recognizing: !!interim });
     this._drain();
   }
-  scheduleAutoSend(draft) {
+  maybeScheduleAutoSend(draft = this.composer.getDraft()) {
+    if (this.snapshot.settings.sendingMode === "manual" || this.snapshot.recognizing || this.snapshot.pendingTranscriptions > 0)
+      return;
+    this.scheduleAutoSend(draft);
+  }
+  scheduleAutoSend(draft, { force = false, stopAfter = false } = {}) {
     this.cancelAutoSend();
     if (!draft.trim() || typeof this.composer.submit !== "function") return;
     const delay = this.snapshot.settings.autoSendDelaySeconds * 1e3;
@@ -641,11 +682,14 @@ var VoiceCoordinator = class {
       const expected = this.autoSendDraft;
       this.autoSendDraft = null;
       this.patch({ autoSendAt: null });
-      if (!this.disposed && this.snapshot.settings.sendingMode !== "manual" && expected === this.composer.getDraft()) {
+      if (!this.disposed && (force || this.snapshot.settings.sendingMode !== "manual") && expected === this.composer.getDraft()) {
         try {
-          this.composer.submit(this.snapshot.settings.sendingMode === "steer" ? "steer" : "queue");
+          this.composer.submit(
+            this.snapshot.settings.sendingMode === "steer" && !force ? "steer" : "queue"
+          );
           this.transcript.reset();
           this.recognition.reset?.();
+          if (stopAfter) void this.stopListening();
         } catch (error) {
           this.patch({ error: message(error) });
         }
@@ -662,10 +706,11 @@ var VoiceCoordinator = class {
     if (this.autoSendDraft !== null && draft !== this.autoSendDraft) this.cancelAutoSend();
   }
   stopListening() {
+    this.holdToTalkRelease = false;
     this.cancelAutoSend();
     ++this.epoch;
     this.inputController?.abort();
-    this.patch({ listening: false, recognizing: false, starting: false });
+    this.patch({ listening: false, recognizing: false, pendingTranscriptions: 0, starting: false });
     return this._input(async () => {
       try {
         await this._releaseInput();
@@ -1389,7 +1434,10 @@ var BrowserRecognitionEngine = class {
       restarts: 0,
       timer: null,
       recognition: null,
-      speaking: false
+      speaking: false,
+      finishing: false,
+      finishPromise: null,
+      resolveFinish: null
     };
     this.session = session;
     const cancelled3 = new Promise((resolve, reject) => {
@@ -1487,6 +1535,12 @@ var BrowserRecognitionEngine = class {
       session.recognition = null;
       this._activity(session, false);
       if (this.session !== session) return;
+      if (session.finishing) {
+        this.session = null;
+        session.signal?.removeEventListener("abort", session.cancel);
+        session.resolveFinish?.();
+        return;
+      }
       if (session.restarts >= this.maxRestarts) {
         this._fail(
           session,
@@ -1515,6 +1569,33 @@ var BrowserRecognitionEngine = class {
     };
     recognition.start();
   }
+  finish() {
+    const session = this.session;
+    if (!session) return Promise.resolve();
+    if (session.finishPromise) return session.finishPromise;
+    session.finishing = true;
+    session.finishPromise = new Promise((resolve) => {
+      session.resolveFinish = resolve;
+    });
+    if (session.timer !== null) {
+      (this.globals.clearTimeout ?? globalThis.clearTimeout)(session.timer);
+      session.timer = null;
+      this.session = null;
+      session.resolveFinish();
+      return session.finishPromise;
+    }
+    try {
+      session.recognition?.stop();
+      if (!session.recognition) {
+        this.session = null;
+        session.resolveFinish();
+      }
+    } catch {
+      this.session = null;
+      session.resolveFinish();
+    }
+    return session.finishPromise;
+  }
   reset() {
     const session = this.session;
     const recognition = session?.recognition;
@@ -1549,6 +1630,7 @@ var BrowserRecognitionEngine = class {
     } catch {
     }
     this._activity(session, false);
+    session.resolveFinish?.();
   }
 };
 
@@ -1585,12 +1667,18 @@ function encodeMonoPcm16Wav(samples, inputRate) {
   return buffer;
 }
 var WhisperHttpRecognitionEngine = class {
-  constructor({ globals = globalThis, meter, voiceDetectionPreset = "natural" } = {}) {
+  constructor({
+    globals = globalThis,
+    meter,
+    voiceDetectionPreset = "natural",
+    maxUtteranceSeconds = 60
+  } = {}) {
     this.g = globals;
     this.meter = meter;
     this.session = null;
     this.lang = "pt-BR";
     this.voiceDetectionPreset = voiceDetectionPreset;
+    this.maxUtteranceSeconds = maxUtteranceSeconds;
     this.route = ROUTE;
   }
   get segmentation() {
@@ -1616,7 +1704,14 @@ var WhisperHttpRecognitionEngine = class {
       };
     }
   }
-  async start({ lang = this.lang, signal, onResult, onActivity, onError } = {}) {
+  async start({
+    lang = this.lang,
+    signal,
+    onResult,
+    onActivity,
+    onError,
+    onProcessingChange
+  } = {}) {
     await this.stop();
     if (signal?.aborted) throw abortError3();
     const context = this.meter?.context, source = this.meter?.source;
@@ -1636,16 +1731,24 @@ var WhisperHttpRecognitionEngine = class {
       samples: 0,
       voiced: false,
       silence: 0,
-      inflight: /* @__PURE__ */ new Set(),
+      transcriptionQueue: [],
+      activeRequest: null,
+      draining: false,
       onResult,
       onActivity,
       onError,
+      onProcessingChange,
       lang,
       signal
     };
     this.session = session;
     const valid = () => this.session === session && !signal?.aborted;
-    const submit = async () => {
+    const notifyProcessing = () => session.onProcessingChange?.({
+      queued: session.transcriptionQueue.length,
+      active: !!session.activeRequest,
+      pending: session.transcriptionQueue.length + (session.activeRequest ? 1 : 0)
+    });
+    const enqueue = () => {
       if (!session.voiced || session.samples < context.sampleRate * 0.25) {
         session.chunks = [];
         session.samples = 0;
@@ -1663,29 +1766,46 @@ var WhisperHttpRecognitionEngine = class {
       session.samples = 0;
       session.voiced = false;
       session.silence = 0;
-      const request = new AbortController();
-      session.inflight.add(request);
+      session.transcriptionQueue.push(samples);
+      notifyProcessing();
+      void drain();
+    };
+    const drain = async () => {
+      if (session.draining || !valid()) return;
+      session.draining = true;
       try {
-        const response = await this.g.fetch(this.route + "/transcribe", {
-          method: "POST",
-          credentials: "same-origin",
-          headers: {
-            "content-type": "audio/wav",
-            "x-dlv-client-id": session.operation,
-            "x-dlv-operation-id": id(),
-            "x-dlv-language": lang
-          },
-          body: encodeMonoPcm16Wav(samples, context.sampleRate),
-          signal: request.signal
-        });
-        const json = await response.json();
-        if (!response.ok || !json?.ok)
-          throw new Error(json?.error?.message || "HTTP transcription failed.");
-        if (valid() && json.value.text) onResult?.({ final: json.value.text, interim: "" });
-      } catch (error) {
-        if (error.name !== "AbortError" && valid()) onError?.(error);
+        while (valid() && session.transcriptionQueue.length) {
+          const samples = session.transcriptionQueue.shift();
+          const request = new AbortController();
+          session.activeRequest = request;
+          notifyProcessing();
+          try {
+            const response = await this.g.fetch(this.route + "/transcribe", {
+              method: "POST",
+              credentials: "same-origin",
+              headers: {
+                "content-type": "audio/wav",
+                "x-dlv-client-id": session.operation,
+                "x-dlv-operation-id": id(),
+                "x-dlv-language": lang
+              },
+              body: encodeMonoPcm16Wav(samples, context.sampleRate),
+              signal: request.signal
+            });
+            const json = await response.json();
+            if (!response.ok || !json?.ok)
+              throw new Error(json?.error?.message || "HTTP transcription failed.");
+            if (valid() && json.value.text) onResult?.({ final: json.value.text, interim: "" });
+          } catch (error) {
+            if (error.name !== "AbortError" && valid()) onError?.(error);
+          } finally {
+            if (session.activeRequest === request) session.activeRequest = null;
+            notifyProcessing();
+          }
+        }
       } finally {
-        session.inflight.delete(request);
+        session.draining = false;
+        notifyProcessing();
       }
     };
     processor.onaudioprocess = (event) => {
@@ -1701,13 +1821,24 @@ var WhisperHttpRecognitionEngine = class {
       }
       session.chunks.push(data);
       session.samples += data.length;
-      if (session.voiced && session.silence > context.sampleRate * (this.segmentation.silenceMs / 1e3) || session.samples > context.sampleRate * 20)
-        void submit();
+      if (session.voiced && session.silence > context.sampleRate * (this.segmentation.silenceMs / 1e3) || session.samples > context.sampleRate * this.maxUtteranceSeconds)
+        enqueue();
     };
     source.connect(processor);
+    session.finish = enqueue;
     session.abort = () => this.stop();
     signal?.addEventListener("abort", session.abort, { once: true });
   }
+  finish() {
+    const session = this.session;
+    if (!session) return;
+    session.processor.onaudioprocess = null;
+    try {
+      this.meter?.source?.disconnect(session.processor);
+    } catch {
+    }
+    session.finish?.();
+  }
   async stop() {
     const session = this.session;
     if (!session) return;
@@ -1723,8 +1854,10 @@ var WhisperHttpRecognitionEngine = class {
       session.gain?.disconnect();
     } catch {
     }
-    for (const request of session.inflight) request.abort();
-    session.inflight.clear();
+    session.transcriptionQueue = [];
+    session.activeRequest?.abort();
+    session.activeRequest = null;
+    session.onProcessingChange?.({ queued: 0, active: false, pending: 0 });
   }
 };
 
@@ -2150,6 +2283,100 @@ function createQwenSettings(React2) {
   };
 }
 
+// src/client/releases.ts
+var CURRENT_VERSION = "0.2.3";
+var TESTED_DSH_VERSION = "0.1.6-alpha.2";
+var REPOSITORY_URL = "https://github.com/victorwads/dsh-live-voice";
+var RELEASES_URL = `${REPOSITORY_URL}/releases`;
+var LATEST_RELEASE_API_URL = "https://api.github.com/repos/victorwads/dsh-live-voice/releases/latest";
+var RELEASE_CHECK_STORAGE_KEY = "dsh-live-voice.latest-release-check";
+var RELEASE_CHECK_INTERVAL_MS = 24 * 60 * 60 * 1e3;
+var LIVE_VOICE_BADGE_URL = "https://cdn.simpleicons.org/npm/white";
+var DSH_BADGE_URL = "https://cdn.simpleicons.org/deepseek/white";
+var TESTED_DSH_RELEASE_URL = `https://github.com/deepseek-ai/deepseek-harness/releases/tag/v${TESTED_DSH_VERSION}`;
+function versionParts(version) {
+  const normalized = String(version || "").trim().replace(/^v/i, "").split("+", 1)[0];
+  const [core = "", prerelease = ""] = normalized.split("-", 2);
+  const numbers = core.split(".").map((part) => {
+    const match = part.match(/^\d+/);
+    return match ? Number(match[0]) : 0;
+  });
+  while (numbers.length < 3) numbers.push(0);
+  return { numbers, prerelease: prerelease ? prerelease.split(".") : [] };
+}
+function compareIdentifier(left, right) {
+  const leftNumber = /^\d+$/.test(left) ? Number(left) : null;
+  const rightNumber = /^\d+$/.test(right) ? Number(right) : null;
+  if (leftNumber !== null && rightNumber !== null) return Math.sign(leftNumber - rightNumber);
+  if (leftNumber !== null) return -1;
+  if (rightNumber !== null) return 1;
+  return left.localeCompare(right);
+}
+function compareVersions(left, right) {
+  const a = versionParts(left);
+  const b = versionParts(right);
+  for (let index = 0; index < Math.max(a.numbers.length, b.numbers.length); index++) {
+    const difference = (a.numbers[index] || 0) - (b.numbers[index] || 0);
+    if (difference) return Math.sign(difference);
+  }
+  if (!a.prerelease.length && !b.prerelease.length) return 0;
+  if (!a.prerelease.length) return 1;
+  if (!b.prerelease.length) return -1;
+  for (let index = 0; index < Math.max(a.prerelease.length, b.prerelease.length); index++) {
+    if (a.prerelease[index] === void 0) return -1;
+    if (b.prerelease[index] === void 0) return 1;
+    const difference = compareIdentifier(a.prerelease[index], b.prerelease[index]);
+    if (difference) return difference;
+  }
+  return 0;
+}
+function readCached(storage, now) {
+  try {
+    const cached = JSON.parse(storage?.getItem(RELEASE_CHECK_STORAGE_KEY) || "null");
+    if (cached && Number.isFinite(cached.checkedAt) && now - cached.checkedAt >= 0 && now - cached.checkedAt < RELEASE_CHECK_INTERVAL_MS)
+      return cached;
+  } catch {
+  }
+  return null;
+}
+function writeCached(storage, value) {
+  try {
+    storage?.setItem(RELEASE_CHECK_STORAGE_KEY, JSON.stringify(value));
+  } catch {
+  }
+}
+async function checkLatestRelease({
+  fetchImpl = globalThis.window?.fetch?.bind(globalThis.window),
+  storage = globalThis.window?.localStorage,
+  now = Date.now()
+} = {}) {
+  const cached = readCached(storage, now);
+  if (cached) return cached;
+  const attempted = { checkedAt: now, release: null };
+  writeCached(storage, attempted);
+  if (typeof fetchImpl !== "function") return attempted;
+  try {
+    const response = await fetchImpl(LATEST_RELEASE_API_URL, {
+      headers: { Accept: "application/vnd.github+json" }
+    });
+    if (!response.ok) return attempted;
+    const body = await response.json();
+    if (typeof body?.tag_name !== "string" || !body.tag_name.trim()) return attempted;
+    const release = {
+      tag: body.tag_name.trim(),
+      url: typeof body.html_url === "string" && body.html_url.startsWith("https://github.com/") ? body.html_url : RELEASES_URL
+    };
+    const result = { checkedAt: now, release };
+    writeCached(storage, result);
+    return result;
+  } catch {
+    return attempted;
+  }
+}
+function hasNewerRelease(release, currentVersion = CURRENT_VERSION) {
+  return Boolean(release?.tag && compareVersions(release.tag, currentVersion) > 0);
+}
+
 // src/client/components.ts
 function createComponents(React2) {
   const h = React2.createElement;
@@ -2443,6 +2670,17 @@ function createComponents(React2) {
     const settings = state.settings || {};
     const capabilities = state.capabilities || {};
     const tabsId = React2.useId();
+    const [latestRelease, setLatestRelease] = React2.useState(null);
+    React2.useEffect(() => {
+      let active = true;
+      void checkLatestRelease().then((result) => {
+        if (active) setLatestRelease(result.release);
+      });
+      return () => {
+        active = false;
+      };
+    }, []);
+    const updateAvailable = hasNewerRelease(latestRelease);
     const tabs = [
       { id: "conversation", label: "Conversation" },
       { id: "speech", label: "Speech" },
@@ -2535,7 +2773,69 @@ function createComponents(React2) {
     return h(
       "section",
       { className: "dlv-settings", "aria-label": "Live Voice settings" },
-      h("h3", null, "Live Voice"),
+      h(
+        "div",
+        { className: "dlv-settings-heading" },
+        h("h3", null, "Live Voice"),
+        h(
+          "div",
+          { className: "dlv-version-badges", "aria-label": "Version information" },
+          h(
+            "a",
+            {
+              className: "dlv-shields-badge",
+              href: RELEASES_URL,
+              target: "_blank",
+              rel: "noreferrer",
+              "aria-label": `DSH Live Voice v${CURRENT_VERSION}. Open releases`,
+              title: `DSH Live Voice v${CURRENT_VERSION}`
+            },
+            h("img", { src: LIVE_VOICE_BADGE_URL, alt: "" }),
+            h("span", null, `v${CURRENT_VERSION}`)
+          ),
+          h(
+            "a",
+            {
+              className: "dlv-shields-badge",
+              href: TESTED_DSH_RELEASE_URL,
+              target: "_blank",
+              rel: "noreferrer",
+              "aria-label": `Compatible with DSH v${TESTED_DSH_VERSION}. Open release`,
+              title: `Compatible with DSH v${TESTED_DSH_VERSION}`
+            },
+            h("img", { src: DSH_BADGE_URL, alt: "" }),
+            h("span", null, `v${TESTED_DSH_VERSION}`)
+          )
+        ),
+        h("span", { className: "dlv-heading-divider", "aria-hidden": true }),
+        updateAvailable ? h(
+          "a",
+          {
+            className: "dlv-version-badge dlv-update-badge",
+            href: latestRelease.url,
+            target: "_blank",
+            rel: "noreferrer",
+            "aria-label": `Update available: ${latestRelease.tag}. Open release`,
+            title: `Update available: ${latestRelease.tag}`
+          },
+          h("span", { className: "dlv-update-icon", "aria-hidden": true }, "\u2191"),
+          "Update available"
+        ) : null,
+        updateAvailable ? h("span", { className: "dlv-heading-divider", "aria-hidden": true }) : null,
+        h(
+          "a",
+          {
+            className: "dlv-version-badge dlv-star-badge",
+            href: REPOSITORY_URL,
+            target: "_blank",
+            rel: "noreferrer",
+            "aria-label": "Star DSH Live Voice on GitHub",
+            title: "Star DSH Live Voice on GitHub"
+          },
+          h("span", { className: "dlv-star-icon", "aria-hidden": true }, "\u2605"),
+          "Star Us on GitHub"
+        )
+      ),
       onClose ? h(Button, { label: "Close voice settings", icon: "close", onClick: onClose }) : null,
       h(
         "div",
@@ -2744,11 +3044,28 @@ function createComponents(React2) {
                   event.target.value === "browser" && settings.recognitionLang === "auto" ? { recognitionEngine: "browser", recognitionLang: "pt-BR" } : { recognitionEngine: event.target.value }
                 )
               },
-              h("option", { value: "qwen-http" }, "Qwen3 ASR \u2014 local MLX server"),
-              h("option", { value: "browser" }, "Browser SpeechRecognition"),
-              h("option", { value: "whisper-http" }, "Whisper HTTP \u2014 DSH host")
+              h("option", { value: "browser" }, "Browser SpeechRecognition \u2014 Default option"),
+              h("option", { value: "qwen-http" }, "Qwen3 ASR \u2014 HTTP API"),
+              h("option", { value: "whisper-http" }, "Whisper \u2014 HTTP API"),
+              h(
+                "optgroup",
+                { label: "Coming Soon \u2014 vote on repo issues" },
+                h("option", { value: "webgpu", disabled: true }, "Browser WebGPU Inference \u2014 Soon"),
+                h(
+                  "option",
+                  { value: "sherpa-onnx", disabled: true },
+                  "sherpa-onnx Streaming \u2014 Soon"
+                ),
+                h("option", { value: "parakeet", disabled: true }, "NVIDIA Parakeet \u2014 Soon"),
+                h("option", { value: "voxtral", disabled: true }, "Voxtral Realtime \u2014 Soon")
+              )
             )
           ),
+          h(
+            "small",
+            { className: "dlv-recognition-engine-description" },
+            settings.recognitionEngine === "qwen-http" ? "HTTP API at the configured base URL (default: http://127.0.0.1:8080/). Compatible with POST /v1/audio/transcriptions." : settings.recognitionEngine === "whisper-http" ? "HTTP API at the configured DSH host URL (default: http://127.0.0.1:8080/inference). Audio uses the authenticated DSH host transcription route." : "Uses the browser SpeechRecognition API. This is the default option."
+          ),
           field("Input device", "inputDeviceId", deviceOptions("audioinput", "Microphone")),
           settings.recognitionEngine === "browser" ? h(
             "small",
@@ -2894,6 +3211,28 @@ function createComponents(React2) {
                 null,
                 "Controls how long a pause must last before captured speech is sent for recognition."
               ),
+              h(
+                "label",
+                { className: "dlv-setting-field" },
+                "Maximum continuous speech (seconds)",
+                h("input", {
+                  type: "number",
+                  min: 10,
+                  max: 300,
+                  step: 1,
+                  value: settings.recognitionMaxUtteranceSeconds ?? 60,
+                  onChange: (event) => {
+                    const value = Number(event.target.value);
+                    if (Number.isInteger(value) && value >= 10 && value <= 300)
+                      invoke("updateSettings", { recognitionMaxUtteranceSeconds: value });
+                  }
+                }),
+                h(
+                  "small",
+                  null,
+                  "If speech never pauses, start a new transcription chunk after this duration. Default: 60 seconds."
+                )
+              ),
               h(
                 "div",
                 {
@@ -2961,6 +3300,21 @@ function createComponents(React2) {
           { 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."
         ),
+        h(
+          "label",
+          { key: "hold-to-talk", className: "dlv-check" },
+          h("input", {
+            type: "checkbox",
+            checked: settings.holdToTalkEnabled !== false,
+            onChange: (event) => invoke("updateSettings", { holdToTalkEnabled: event.target.checked })
+          }),
+          " Hold Control to talk"
+        ),
+        h(
+          "p",
+          { key: "hold-to-talk-description", className: "dlv-setting-description" },
+          "While a composer is open, hold Control anywhere on the page to capture speech. Release it to flush queued transcription, wait the configured send delay, queue the message, and close voice capture. Press Escape while holding to cancel."
+        ),
         field("Listening mode", "mode", [
           { value: "speaker", label: "Speakers \u2014 gated listening" },
           { value: "headphones", label: "Headphones \u2014 open microphone" }
@@ -3066,13 +3420,15 @@ var styles = `
 .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-heading{display:flex;align-items:center;gap:8px;min-width:0}.dlv-settings-heading h3{white-space:nowrap}.dlv-heading-divider{height:1px;min-width:12px;flex:1;background:var(--dsw-alias-border-l1)}.dlv-settings-heading .dlv-star-badge{flex:none}
+.dlv-version-badges{display:flex;flex:none;gap:5px;align-items:center}.dlv-shields-badge{display:inline-flex;align-items:center;justify-content:center;box-sizing:border-box;height:20px;gap:4px;padding:3px 5px;border:1px solid var(--dsw-alias-border-l1);border-radius:5px;background:var(--dsw-alias-bg-layer-1);color:var(--dsw-alias-label-secondary);font-size:10px;font-weight:650;line-height:12px;text-decoration:none}.dlv-shields-badge img{display:block;width:12px;height:12px;max-width:100%}.dlv-shields-badge:focus-visible{border-radius:4px;outline:2px solid var(--dsw-alias-label-primary);outline-offset:3px}.dlv-version-badge{display:inline-flex;align-items:center;min-height:24px;box-sizing:border-box;padding:3px 8px;border:1px solid var(--dsw-alias-border-l1);border-radius:999px;background:var(--dsw-alias-bg-layer-1);color:var(--dsw-alias-label-secondary);font-size:11px;font-weight:650;line-height:16px;text-decoration:none}.dlv-version-badges a.dlv-version-badge:hover{border-color:var(--dsw-alias-border-l2);color:var(--dsw-alias-label-primary)}.dlv-star-badge{padding:4px 10px;color:var(--dsw-alias-label-primary)}.dlv-update-icon{margin-right:4px;font-size:14px;line-height:1}.dlv-star-icon{margin-right:4px;color:#f59e0b;font-size:14px;line-height:1;text-shadow:0 0 8px rgba(245,158,11,.28)}.dlv-version-badges a.dlv-star-badge:hover{border-color:#f59e0b;background:rgba(245,158,11,.08)}.dlv-update-badge{border-color:var(--dsw-alias-state-business-primary);color:var(--dsw-alias-state-business-primary)}
 .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,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}
-@media(max-width:480px){.dlv-preset-group{grid-template-columns:1fr}.dlv-pill{flex-wrap:wrap}.dlv-wave{flex-basis:90px}.dlv-status{flex-basis:100px}}
+@media(max-width:480px){.dlv-settings-heading{gap:6px}.dlv-heading-divider{min-width:4px}.dlv-settings-heading .dlv-star-badge{padding:3px 6px;font-size:10px}.dlv-preset-group{grid-template-columns:1fr}.dlv-pill{flex-wrap:wrap}.dlv-wave{flex-basis:90px}.dlv-status{flex-basis:100px}}
 `;
 
 // src/client/chat.ts
@@ -3120,12 +3476,22 @@ function apply(ctx) {
   let disposed = false;
   let voiceModeActive = false;
   const ownership = new VoiceOwnership();
+  const recognitionSettingKeys = /* @__PURE__ */ new Set([
+    "recognitionEngine",
+    "recognitionProcessLocally",
+    "recognitionAutoInstall",
+    "voiceDetectionPreset",
+    "recognitionMaxUtteranceSeconds"
+  ]);
+  const changesRecognition = (next) => Object.keys(next).some((key) => recognitionSettingKeys.has(key));
   const recognitionFor = (settings, meter) => settings.recognitionEngine === "qwen-http" ? new QwenHttpRecognitionEngine({
     meter,
-    voiceDetectionPreset: settings.voiceDetectionPreset
+    voiceDetectionPreset: settings.voiceDetectionPreset,
+    maxUtteranceSeconds: settings.recognitionMaxUtteranceSeconds
   }) : settings.recognitionEngine === "whisper-http" ? new WhisperHttpRecognitionEngine({
     meter,
-    voiceDetectionPreset: settings.voiceDetectionPreset
+    voiceDetectionPreset: settings.voiceDetectionPreset,
+    maxUtteranceSeconds: settings.recognitionMaxUtteranceSeconds
   }) : new BrowserRecognitionEngine({
     processLocally: settings.recognitionProcessLocally,
     autoInstallLocalPack: settings.recognitionAutoInstall
@@ -3202,7 +3568,8 @@ function apply(ctx) {
             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));
+              if (text)
+                run(entry.controller, entry.controller.speak(text, capture.interaction.key));
             } else {
               entry.questionCapture = null;
               entry.controller.patch({ answeringQuestion: false });
@@ -3303,7 +3670,7 @@ function apply(ctx) {
         return original(...args);
       };
     }
-    for (const method of ["startDictation", "startConversation", "speak"]) {
+    for (const method of ["startDictation", "startHoldToTalk", "startConversation", "speak"]) {
       const original = controller[method].bind(controller);
       controller[method] = (...args) => {
         if (disposed || entry.closed || !entry.refs) return Promise.resolve();
@@ -3414,13 +3781,11 @@ function apply(ctx) {
       let settingsRevision = 0;
       c.updateSettings = async (next) => {
         const revision = ++settingsRevision;
-        const previousEngine = c.getSnapshot().settings.recognitionEngine;
         update(next);
         browser.lang = c.getSnapshot().settings.lang;
         qwen.lang = c.getSnapshot().settings.lang;
         const settings2 = c.getSnapshot().settings;
-        if (previousEngine !== settings2.recognitionEngine)
-          c.replaceRecognition(recognitionFor(settings2, c.meter));
+        if (changesRecognition(next)) c.replaceRecognition(recognitionFor(settings2, c.meter));
         localStorage.setItem("dsh-live-voice.settings", JSON.stringify(settings2));
         run(c, c.refreshCapabilities());
         const active = [...controllers.values()];
@@ -3431,7 +3796,7 @@ function apply(ctx) {
         if (revision !== settingsRevision || disposed || c.disposed) return;
         for (const entry of controllers.values()) {
           if (entry.closed) continue;
-          if (entry.controller.getSnapshot().settings.recognitionEngine !== settings2.recognitionEngine)
+          if (changesRecognition(next))
             entry.controller.replaceRecognition(recognitionFor(settings2, entry.controller.meter));
           entry.applySettings(settings2);
           run(entry.controller, entry.controller.refreshCapabilities());
@@ -3548,19 +3913,47 @@ function apply(ctx) {
       for (const entry of controllers.values())
         run(entry.controller, entry.controller.endConversation());
     };
-    const onKey = (event) => {
-      if (disposed || event.defaultPrevented || !event.ctrlKey || !event.shiftKey || event.code !== "Space" || event.repeat)
-        return;
+    let holdToTalk = null;
+    const candidate = () => {
       const candidates = [...controllers.values()].filter(
         (entry) => entry.buttons > 0 && entry.composers.size > 0
       );
-      if (candidates.length !== 1) return;
-      event.preventDefault();
-      const c = candidates[0].controller;
-      run(
-        c,
-        c.getSnapshot().listening || c.getSnapshot().starting ? c.stopListening() : c.startDictation()
-      );
+      return candidates.length === 1 ? candidates[0] : null;
+    };
+    const releaseHoldToTalk = () => {
+      const entry = holdToTalk;
+      holdToTalk = null;
+      if (!entry || entry.closed) return;
+      run(entry.controller, entry.controller.releaseHoldToTalk());
+    };
+    const onKeyDown = (event) => {
+      if (disposed || event.defaultPrevented || event.repeat) return;
+      if (event.key === "Escape" && holdToTalk) {
+        const entry2 = holdToTalk;
+        holdToTalk = null;
+        event.preventDefault();
+        run(entry2.controller, entry2.controller.cancelDictation());
+        return;
+      }
+      if (event.ctrlKey && event.shiftKey && event.code === "Space") {
+        const entry2 = candidate();
+        if (!entry2) return;
+        event.preventDefault();
+        const c = entry2.controller;
+        run(
+          c,
+          c.getSnapshot().listening || c.getSnapshot().starting ? c.stopListening() : c.startDictation()
+        );
+        return;
+      }
+      if (event.key !== "Control" || event.altKey || event.metaKey || event.shiftKey) return;
+      const entry = candidate();
+      if (!entry || !entry.controller.getSnapshot().settings.holdToTalkEnabled) return;
+      holdToTalk = entry;
+      run(entry.controller, entry.controller.startHoldToTalk());
+    };
+    const onKeyUp = (event) => {
+      if (event.key === "Control") releaseHoldToTalk();
     };
     const refreshCapabilities = () => {
       for (const entry of controllers.values())
@@ -3573,7 +3966,9 @@ function apply(ctx) {
     window.speechSynthesis?.addEventListener?.("voiceschanged", refreshCapabilities);
     navigator.mediaDevices?.addEventListener?.("devicechange", refreshCapabilities);
     document.addEventListener("visibilitychange", visibilityChanged);
-    document.addEventListener("keydown", onKey);
+    document.addEventListener("keydown", onKeyDown);
+    document.addEventListener("keyup", onKeyUp);
+    window.addEventListener("blur", releaseHoldToTalk);
     window.addEventListener("pagehide", stop);
     return () => {
       disposed = true;
@@ -3581,7 +3976,9 @@ function apply(ctx) {
       window.speechSynthesis?.removeEventListener?.("voiceschanged", refreshCapabilities);
       navigator.mediaDevices?.removeEventListener?.("devicechange", refreshCapabilities);
       document.removeEventListener("visibilitychange", visibilityChanged);
-      document.removeEventListener("keydown", onKey);
+      document.removeEventListener("keydown", onKeyDown);
+      document.removeEventListener("keyup", onKeyUp);
+      window.removeEventListener("blur", releaseHoldToTalk);
       window.removeEventListener("pagehide", stop);
       for (const entry of controllers.values()) retire(entry);
     };