@olares/lares-mobile 0.15.10 → 0.15.12

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@olares/lares-mobile",
-  "version": "0.15.10",
+  "version": "0.15.12",
   "description": "Lares chat component for LarePass",
   "type": "module",
   "repository": {
@@ -28,7 +28,7 @@
     "vue": "^3.2.0"
   },
   "dependencies": {
-    "@olares/lares-core": "0.15.10",
+    "@olares/lares-core": "0.15.12",
     "markdown-it": "^14.3.0"
   }
 }

package/src/App.vue

@@ -89,6 +89,7 @@
       @files="attachFiles"
       @hold-start="onHoldStart"
       @hold-end="onHoldEnd"
+      @voice-action="voiceAction"
       @close-model="modelSheet = false"
       @choose-model="chooseModel"
       @close-effort="effortSheet = false"
@@ -108,11 +109,12 @@ import { pickHostPorts } from "@olares/lares-core/larepass/host";
 import { connectChat } from "./runtime.js";
 import { createT } from "./i18n.js";
 import { failText, withPendingUser } from "./chat/format.js";
-import { createVoiceCapture } from "./chat/voice.js";
+import { captureSource } from "@olares/lares-core/voice/input";
+import { createRealtimeConversation } from "@olares/lares-core/voice/realtime-conversation";
+import { createVoiceConversation } from "@olares/lares-core/voice/conversation";
 import { createAnswerPlayer } from "./chat/playback.js";
-import { createReplySpeaker } from "@olares/lares-core/voice/speech";
+import { createReplySpeaker, createSentencePlayer, speakable } from "@olares/lares-core/voice/speech";
 import { wavBlob } from "@olares/lares-core/voice/client";
-import { captureSource } from "@olares/lares-core/voice/input";
 import { browserPlays } from "@olares/lares-core/voice/playback";
 import { speakProblemText, speakToggleNotice } from "@olares/lares-core/voice/problem";
 import { appendDraftMentions } from "@olares/lares-core/files/mention";
@@ -218,11 +220,16 @@ export default {
       effortSheet: false,
       upload: { pending: 0, failures: [] },
       draftImages: [],
+      voiceCall: null,
+      voiceHistory: null,
+      voiceCaption: "",
       voicePhase: "idle",
       voiceElapsed: 0,
       voiceError: "",
       voiceLanguage: "",
       voiceSpeak: false,
+      voiceDefaultMode: "continuous",
+      voiceStreaming: false,
       speakBusy: false,
       speakNotice: "",
       intakeUnsub: null,
@@ -328,6 +335,8 @@ export default {
         effortDisabled: this.effortDisabled,
         upload: this.upload,
         draftImages: this.draftImages,
+        voiceDefaultMode: this.voiceDefaultMode,
+        voiceCall: this.voiceCall,
         voicePhase: this.voicePhase,
         voiceElapsed: this.voiceElapsed,
         voiceErrorText: this.voiceErrorText,
@@ -523,26 +532,21 @@ export default {
     this.imageUploadControllers = new Map();
     this.player = createAnswerPlayer();
     this.replySpeaker = createReplySpeaker();
-    this.speakGen = 0;
-    this.voiceCap = createVoiceCapture({
-      transcribe: (blob, signal) => this.runtime.transcribe(blob, this.voiceLanguage, signal),
-      onPhase: (phase, extra = {}) => {
-        this.voicePhase = phase;
-        if (extra.elapsed != null) this.voiceElapsed = extra.elapsed;
-        if (extra.error != null) this.voiceError = extra.error;
+    this.reader = createSentencePlayer({
+      send: async (request) => {
+        const result = await this.runtime.speak?.(request);
+        if (!result) throw Object.assign(new Error("voice_speak_failed"), { code: "voice_speak_failed" });
+        return result;
       },
-      onTrace: (event, fields) => {
-        void this.runtime.reportLog?.("voice", event, fields);
-      },
-      submit: (text) => {
-        void this.runtime.reportLog?.("voice", "submit", {
-          chars: String(text ?? "").trim().length,
-          running: Boolean(this.running),
-          session: Boolean(this.sessionId),
-        });
-        void this.sendTranscript(text);
+      play: (sentence) => {
+        if (!browserPlays("mobile", sentence.bluetooth)) return Promise.resolve(false);
+        return this.player?.play(wavBlob(sentence.audio));
       },
+      halt: () => this.player?.stop(),
+      silence: () => void this.runtime.interruptSpeech?.(),
+      fail: (err) => this.showSpeakProblem(err),
     });
+
     this.bindIntake(this.sessionId);
     this.bindRuntime();
     try {
@@ -558,6 +562,13 @@ export default {
     this.conversationUnsub = subscribeConversationSettings((section) => {
       this.busyEnter = section.busyEnter;
     });
+    this.voiceVisibilityHandler = () => {
+      if (document.hidden) void this.voiceConversation?.controller.disconnect("voice_background");
+    };
+    document.addEventListener("visibilitychange", this.voiceVisibilityHandler);
+    this.voiceClock = setInterval(() => {
+      if (this.voicePhase === "recording") this.voiceElapsed += 1;
+    }, 1000);
     this.retry();
   },
   activated() {
@@ -567,11 +578,14 @@ export default {
     this.captureLog();
   },
   beforeUnmount() {
+    document.removeEventListener("visibilitychange", this.voiceVisibilityHandler);
+    clearInterval(this.voiceClock);
     this.captureLog();
     this.noteSeen(this.sessionId);
     for (const controller of this.imageUploadControllers?.values?.() ?? []) controller.abort();
     this.imageUploadControllers?.clear?.();
-    this.voiceCap?.dispose();
+    void this.voiceConversation?.stop().catch(() => {});
+    this.voiceUnsub?.();
     this.conversationUnsub?.();
     this.intakeUnsub?.();
     if (this.sessionId) this.intake?.cancelSession(this.sessionId);
@@ -628,21 +642,24 @@ export default {
         this.commands = null;
         this.commandsError = "";
       }
-      if (sessionChanged) {
-        this.speakGen += 1;
+      if (sessionChanged && snap.sessionId !== this.voiceAdoptingSession) {
+        void this.voiceConversation?.stop().catch(() => {});
+        this.reader.stop();
         this.player?.stop();
       }
-      const answer = this.replySpeaker.observe({
+      if (sessionChanged) this.voiceAdoptingSession = null;
+      const utterance = this.replySpeaker.observe({
         sessionChanged,
         wasRunning: this.running,
         running: snap.running,
-        reply: snap.reply,
+        utterance: snap.utterance,
       });
       this.sessionId = snap.sessionId;
       this.items = snap.items;
       this.running = snap.running;
       this.runningSince = snap.runningSince || 0;
-      if (answer) void this.speakReply(answer, ++this.speakGen);
+      const target = this.voiceSpeak && !this.voiceCall ? speakable(utterance, this.voiceStreaming) : null;
+      if (target) this.reader.follow({ ...target, key: `${snap.sessionId}:${target.key}` });
       this.failed = snap.failed;
       this.error = snap.error;
       this.sessions = snap.sessions ?? this.sessions;
@@ -936,7 +953,7 @@ export default {
       this.unsub = this.runtime.subscribe((snap) => this.applySnap(snap));
     },
     switchHost() {
-      this.voiceCap?.stop?.(true);
+      void this.voiceConversation?.stop().catch(() => {});
       for (const controller of this.imageUploadControllers.values()) controller.abort();
       this.imageUploadControllers.clear();
       this.releaseDraftImages();
@@ -1015,10 +1032,13 @@ export default {
       if (settings) {
         jobs.push(settings.voice().then((voice) => {
           this.voiceLanguage = voice.config?.language || "";
+          this.voiceDefaultMode = voice.config?.mobileMode || "continuous";
           this.voiceSpeak = voice.config?.speak === true;
+          this.voiceStreaming = voice.config?.streaming === true;
         }).catch(() => {
           this.voiceLanguage = "";
           this.voiceSpeak = false;
+          this.voiceStreaming = false;
         }));
       }
       await Promise.all(jobs);
@@ -1169,24 +1189,90 @@ export default {
       this.draftImages = [];
       this.imagesBySession = {};
     },
-    onHoldStart(detail) {
-      void this.runtime.reportLog?.("voice", "hold-start", {
-        via: detail?.via || "",
-        phase: this.voicePhase,
+    async onHoldStart({ mode: selectedMode } = {}) {
+      this.voiceReleased = false;
+      if (this.voiceCall?.connection === "open") {
+        await this.voiceConversation.stop(); return;
+      }
+      const target = this.sessionId;
+      this.voicePhase = "preparing"; this.voiceElapsed = 0;
+      try {
+        const settings = await this.runtime.settings.voice(true);
+        if (this.sessionId !== target) return;
+        const continuous = (selectedMode || settings.config.mobileMode || "continuous") === "continuous";
+        if (settings.config.backend === "realtime" && continuous) {
+          const conversation = createRealtimeConversation({ readAloud: settings.config.speak, idleTimeoutMs: settings.config.idleTimeoutMs, recoveryTimeoutMs: settings.config.recoveryTimeoutMs,
+            request: (path, options) => this.runtime.voiceRequest(path, options), socket: (ticket) => this.runtime.voiceSocket(ticket),
+            device: captureSource("mobile", settings.config.device), playback: this.player,
+            onCaption: (text) => { this.voiceCaption = `${this.voiceCaption}${text}`.slice(-300); },
+          });
+          await this.attachVoiceConversation(conversation, target, true);
+          return;
+        }
+        this.reader.stop(); this.player?.stop(); this.player?.unlock();
+        let conversation;
+        const reader = createSentencePlayer({
+          send: (value) => conversation.speak(value),
+          play: (sentence) => sentence.bluetooth === "played" ? Promise.resolve() : this.player.play(wavBlob(sentence.audio)),
+          halt: () => this.player.stop(), silence: () => {},
+          onState: (playing) => conversation?.playbackState(playing),
+          fail: (error) => { this.showSpeakProblem(error); void conversation.controller.disconnect("voice_playback_failed"); },
+        });
+        conversation = createVoiceConversation({
+          request: (path, options) => this.runtime.voiceRequest(path, options),
+          transcribe: (blob, signal) => this.runtime.transcribe(blob, this.voiceLanguage, signal),
+          device: captureSource("mobile", settings.config.device),
+          readAloud: continuous && settings.config.speak,
+          preset: settings.config.preset, endpointSilenceMs: settings.config.endpointSilenceMs, idleTimeoutMs: settings.config.idleTimeoutMs, recoveryTimeoutMs: settings.config.recoveryTimeoutMs,
+          onText: (text) => { this.voiceCaption = text; if (this.voiceCall) this.voiceCall = { ...this.voiceCall, caption: text }; },
+          onTranscript: continuous && settings.config.preset !== "dictation" ? null : ({ text }) => {
+            if (this.sessionId !== target) return;
+            this.draft = [this.draft, text].filter(Boolean).join(" ");
+            this.draftBySession = { ...this.draftBySession, [target || "@pending"]: this.draft };
+          },
+          mode: () => resolveSubmitMode(this.running, false, true, this.busyEnter),
+          playback: { stop: () => { reader.stop(); this.player.stop(); } },
+          onReply: (reply) => { const value = speakable(reply, settings.config.streaming); if (value) reader.follow(value); },
+        });
+        await this.attachVoiceConversation(conversation, target, continuous);
+      } catch (error) { this.voicePhase = "error"; this.voiceError = error.message; }
+    },
+    async attachVoiceConversation(conversation, target, continuous) {
+      this.voiceCaption = ""; this.voiceHistory = null;
+      this.voiceConversation = conversation;
+      this.voiceUnsub?.();
+      this.voiceUnsub = conversation.controller.subscribe(() => {
+        const state = conversation.controller.getSnapshot();
+        if (state.connection === "open" && !target && !this.sessionId && !this.voiceAdoptingSession && state.sessionId) {
+          this.voiceAdoptingSession = state.sessionId;
+          void Promise.resolve(this.runtime.openSession(state.sessionId)).catch((error) => { this.voiceError = error.message; this.voiceAdoptingSession = null; });
+        }
+        this.voiceCall = state.connection === "closed" ? null : { ...state, caption: this.voiceCaption, history: this.voiceHistory };
+        this.voicePhase = state.error ? "error" : state.capture;
+        this.voiceError = state.error || "";
       });
-      if (captureSource("mobile") !== "device") return;
-      if (this.voicePhase === "idle" || this.voicePhase === "error") void this.voiceCap?.start();
-      this.speakGen += 1;
-      this.player?.stop();
-      this.player?.unlock();
-      void this.runtime.interruptSpeech?.();
+      await conversation.start({ sessionId: target || null, continuous });
+      if (this.voiceReleased && !continuous) conversation.controller.stopRecording();
     },
-    onHoldEnd(detail) {
-      void this.runtime.reportLog?.("voice", "hold-end", {
-        via: detail?.via || "",
-        phase: this.voicePhase,
-      });
-      this.voiceCap?.stop(false);
+    onHoldEnd() {
+      this.voiceReleased = true;
+      if (!this.voiceConversation?.controller.getSnapshot().continuous) this.voiceConversation?.controller.stopRecording();
+    },
+    async voiceAction(action) {
+      try {
+        if (action === "mute") await this.voiceConversation?.setReadAloud(this.voiceCall.outputMuted);
+        if (action === "history") {
+          this.voiceHistory = (await this.voiceConversation?.records()).records;
+          if (this.voiceCall) this.voiceCall = { ...this.voiceCall, history: this.voiceHistory };
+        }
+        if (action === "clear-history") { await this.voiceConversation?.clearRecords(); this.voiceHistory = null; }
+        if (action === "input-pause") await this.voiceConversation?.controller.pauseInput();
+        if (action === "input-resume") this.voiceConversation?.controller.resumeInput();
+        if (action === "stop") await this.voiceConversation?.stop();
+        if (action === "playback") await this.voiceConversation?.controller.stopPlayback();
+        if (action === "cancel") await this.voiceConversation?.controller.cancelTask();
+        if (action === "resume") await this.voiceConversation?.resume();
+      } catch (error) { this.voiceError = error.message; }
     },
     showSpeakProblem(err) {
       this.speakNotice = speakProblemText(this.t, err);
@@ -1203,13 +1289,14 @@ export default {
       this.voiceSpeak = next;
       this.speakBusy = true;
       if (!next) {
-        this.speakGen += 1;
+        this.reader.stop();
         this.player?.stop();
         void this.runtime.interruptSpeech?.();
       }
       try {
         const voice = await this.runtime.settings.saveVoice({ speak: next });
         this.voiceSpeak = voice?.config ? voice.config.speak === true : next;
+        this.voiceStreaming = voice?.config?.streaming === true;
         const problem = this.voiceSpeak ? speakToggleNotice(voice?.ttsModels) : "";
         this.speakNotice = problem ? this.t(`error.${problem}`) : "";
       } catch (err) {
@@ -1219,22 +1306,6 @@ export default {
         this.speakBusy = false;
       }
     },
-    async speakReply(text, gen) {
-      if (!this.voiceSpeak) return;
-      try {
-        const result = await this.runtime.speak?.(text);
-        if (gen !== this.speakGen) return;
-        if (!result?.audio) {
-          this.showSpeakProblem(Object.assign(new Error("voice_speak_failed"), { code: "voice_speak_failed" }));
-          return;
-        }
-        if (!browserPlays("mobile", result.bluetooth)) return;
-        await this.player?.play(wavBlob(result.audio));
-      } catch (err) {
-        if (gen !== this.speakGen) return;
-        this.showSpeakProblem(err);
-      }
-    },
     toggleHistory() {
       if (this.panel === "history") {
         this.panel = "";
@@ -1379,27 +1450,6 @@ export default {
       if (!sessionId) return;
       return this.runtime.archiveSession(sessionId);
     },
-    async sendTranscript(text) {
-      const spoken = String(text ?? "").trim();
-      if (!spoken || !this.sessionId || this.modelBlocked || this.historyLoading || this.upload?.pending) return;
-      const mode = resolveSubmitMode(this.running, false, true, this.busyEnter);
-      this.sending = true;
-      this.pendingUser = this.running ? "" : spoken;
-      this.runtime.pinToBottom();
-      this.pinLog();
-      try {
-        const sent = await this.runtime.send([{ type: "text", text: spoken }], mode);
-        if (!sent.ok) {
-          this.pendingUser = "";
-          this.failed = sent.error?.message || sent.error?.code || "send";
-        }
-      } catch (err) {
-        this.pendingUser = "";
-        this.failed = err instanceof Error ? err.message : "send";
-      } finally {
-        this.sending = false;
-      }
-    },
     async send(mode = "queue") {
       const text = this.draft.trim();
       if (!this.canSend) return;

package/src/chat/Composer.vue

@@ -82,6 +82,10 @@
           </button>
         </figure>
       </div>
+      <select v-if="inputMode === 'voice' && !voiceCall" :value="voiceSelection || voiceDefaultMode" @change="voiceSelection = $event.target.value" class="lares-composer__voice-mode" :aria-label="t('settings.mode')">
+        <option value="single">{{ t('settings.mode.single') }}</option>
+        <option value="continuous">{{ t('settings.mode.continuous') }}</option>
+      </select>
       <div class="lares-composer__line" :data-expanded="expanded ? 'true' : 'false'">
         <button
           type="button"
@@ -117,8 +121,7 @@
           :data-phase="voicePhase"
           :disabled="voiceBlocked"
           :aria-label="voiceLabel"
-          @pointerdown="onVoiceDown"
-          @touchstart.prevent="onVoiceTouchStart"
+          @click="onVoiceClick"
           @contextmenu.prevent
         >
           <span v-if="voicePhase === 'preparing' || voicePhase === 'transcribing'" class="lares-composer__talk-status">
@@ -181,6 +184,25 @@
         </div>
       </div>
     </div>
+    <div v-if="voiceCall" class="lares-composer__hint" role="status" aria-live="polite">
+      <button v-if="voiceCall.continuous" type="button" @click="$emit('voice-action', 'mute')">{{ t(voiceCall.outputMuted ? 'call.unmute' : 'call.mute') }}</button>
+      <button v-if="voiceCall.continuous" type="button" @click="$emit('voice-action', 'history')">{{ t('call.history') }}</button>
+      <details v-if="voiceCall.history" open>
+        <summary>{{ t('call.history') }}</summary>
+        <p v-for="row in voiceCall.history" :key="row.id">{{ row.role }}: {{ row.text }}</p>
+        <button type="button" @click="$emit('voice-action', 'clear-history')">{{ t('call.clearHistory') }}</button>
+      </details>
+      <svg v-if="voiceCall.capture === 'recording' && voiceCall.inputLevel != null" width="80" height="24" aria-hidden="true">
+        <line v-for="(level, index) in levels" :key="index" :x1="index * 5" :x2="index * 5" :y1="12 - Math.min(11, level * 800)" :y2="12 + Math.min(11, level * 800)" stroke="currentColor" stroke-width="2" />
+      </svg>
+      <span v-if="voiceCall.caption">{{ voiceCall.caption }}</span>
+      <span>{{ voiceCall.capture === 'transcribing' ? t('mic.transcribing') : voiceCall.capture === 'recording' ? t('call.listening') : voiceCall.inputPaused ? t('call.inputPaused') : voiceCall.playback !== 'idle' ? t('call.halfDuplex') : t('call.waiting') }}</span>
+      <span v-if="voiceCall.playback !== 'idle'">{{ t('call.speaking') }}</span>
+      <button v-if="!voiceCall.continuous && voiceCall.capture === 'recording'" type="button" @click="$emit('hold-end')">{{ t('mic.finish') }}</button>
+      <button v-if="voiceCall.continuous && voiceCall.connection === 'open'" type="button" @click="$emit('voice-action', voiceCall.inputPaused ? 'input-resume' : 'input-pause')">{{ t(voiceCall.inputPaused ? 'call.resumeInput' : 'call.pauseInput') }}</button>
+      <button type="button" @click="$emit('voice-action', voiceCall.connection === 'paused' ? 'resume' : 'stop')">{{ t(voiceCall.connection === 'paused' ? 'call.resume' : voiceCall.continuous ? 'call.end' : 'mic.cancel') }}</button>
+      <button v-if="voiceCall.playback !== 'idle'" type="button" @click="$emit('voice-action', 'playback')">{{ t('call.stopPlayback') }}</button>
+    </div>
     <p v-if="voiceError && voicePhase === 'error'" class="lares-composer__hint" data-status="error">{{ voiceError }}</p>
     <p v-for="item in failures" :key="item.id" class="lares-composer__hint" data-status="error">
       {{ t("upload.failed", { name: item.name, reason: t(`error.${item.code}`) }) }}
@@ -189,7 +211,7 @@
 </template>
 
 <script>
-import { formatElapsed } from "./voice.js";
+import { formatElapsed } from "@olares/lares-core/voice/recorder";
 import {
   REFERENCE_DEBOUNCE_MS,
   activeReference,
@@ -208,6 +230,7 @@ export default {
     canSend: { type: Boolean, default: false },
     attachPending: { type: Boolean, default: false },
     attachDisabled: { type: Boolean, default: false },
+    voiceCall: { type: Object, default: null },
     voicePhase: { type: String, default: "idle" },
     voiceElapsed: { type: Number, default: 0 },
     voiceError: { type: String, default: "" },
@@ -219,11 +242,12 @@ export default {
     commands: { type: Array, default: null },
     commandsError: { type: String, default: "" },
     hero: { type: Boolean, default: false },
+    voiceDefaultMode: { type: String, default: "continuous" },
     inputMode: { type: String, default: "text" },
     t: { type: Function, required: true },
   },
   emits: [
-    "update:draft", "update:inputMode", "send", "files", "voice", "hold-start", "hold-end",
+    "update:draft", "update:inputMode", "send", "files", "voice", "voice-action", "hold-start", "hold-end",
     "open-commands", "open-references", "pick-command", "remove-image",
   ],
   data() {
@@ -232,6 +256,8 @@ export default {
       commandsOpen: false,
       commandTyped: false,
       commandIndex: 0,
+      levels: [],
+      voiceSelection: "",
       referenceOpen: false,
       referenceToken: null,
       referenceIndex: 0,
@@ -244,10 +270,10 @@ export default {
     },
     talkLabel() {
       if (this.voicePhase === "preparing") return this.t("composer.opening");
-      if (this.voicePhase === "recording") return `${this.t("composer.hold")} · ${this.elapsed}`;
+      if (this.voicePhase === "recording") return `${this.t("call.listening")} · ${this.elapsed}`;
       if (this.voicePhase === "transcribing") return this.t("mic.transcribing");
       if (this.holding) return this.t("composer.hold");
-      return this.t("composer.holdToTalk");
+      return this.t("call.start");
     },
     voiceBusy() {
       return this.voicePhase === "preparing"
@@ -264,9 +290,9 @@ export default {
     voiceLabel() {
       if (this.voicePhase === "error" && this.voiceError) return this.voiceError;
       if (this.voicePhase === "preparing") return this.t("composer.opening");
-      if (this.voicePhase === "recording") return this.t("composer.hold");
+      if (this.voicePhase === "recording") return this.t("call.listening");
       if (this.voicePhase === "transcribing") return this.t("mic.transcribing");
-      return this.t("composer.voice");
+      return this.t("call.start");
     },
     voiceBlocked() {
       return this.voicePhase === "transcribing" || this.sending;
@@ -285,6 +311,7 @@ export default {
     },
   },
   watch: {
+    "voiceCall.inputLevel"(level) { if (level != null) this.levels = [...this.levels.slice(-15), level]; },
     draft(value) {
       this.autosize();
       if (value && this.inputMode === "voice" && !this.voiceBusy) this.setMode("text");
@@ -447,55 +474,15 @@ export default {
         event.preventDefault();
       }
     },
-    onVoiceDown(event) {
-      if (event.pointerType === "touch") return;
-      if (event.button > 0 || this.voiceBlocked || this.holding) return;
-      event.preventDefault();
-      event.currentTarget.setPointerCapture?.(event.pointerId);
-      this.beginHold("mouse");
-    },
-    onVoiceTouchStart() {
-      if (this.voiceBlocked || this.holdKind === "touch") return;
-      this.beginHold("touch");
-    },
-    beginHold(kind) {
-      const starting = !this.holding;
-      this.detachHold();
-      this.holding = true;
-      this.holdKind = kind;
-      if (starting) this.$refs.input?.blur();
-      const end = (event) => {
-        if (kind === "touch" && event.type.startsWith("pointer")) return;
-        if (kind === "mouse" && event.type.startsWith("touch")) return;
-        this.detachHold();
-        if (!this.holding) return;
-        this.holding = false;
-        this.holdKind = "";
-        this.$emit("hold-end", { via: event.type });
-      };
-      this.holdEnd = end;
-      if (kind === "touch") {
-        window.addEventListener("touchend", end);
-        window.addEventListener("touchcancel", end);
-      } else {
-        window.addEventListener("pointerup", end, true);
-        window.addEventListener("pointercancel", end, true);
-      }
-      if (starting) this.$emit("hold-start", { via: kind });
-    },
-    detachHold() {
-      if (!this.holdEnd) return;
-      window.removeEventListener("touchend", this.holdEnd);
-      window.removeEventListener("touchcancel", this.holdEnd);
-      window.removeEventListener("pointerup", this.holdEnd, true);
-      window.removeEventListener("pointercancel", this.holdEnd, true);
-      this.holdEnd = null;
+    onVoiceClick() {
+      if (this.voiceBusy) return;
+      this.$emit("hold-start", { mode: this.voiceSelection || undefined });
     },
     clearHold() {
-      this.detachHold();
       this.holding = false;
       this.holdKind = "";
     },
+    detachHold() {},
     pickFiles() {
       this.$refs.picker?.click();
     },
@@ -915,3 +902,8 @@ export default {
   }
 }
 </style>
+
+<style scoped>
+.lares-composer__voice-mode { margin: 8px 12px; border: 0; border-radius: 8px; padding: 6px; background: transparent; color: inherit; font: inherit; }
+.lares-composer__hint button { min-height: 44px; border: 0; border-radius: 12px; padding: 8px 12px; background: var(--lares-bg-secondary, #f0f0f0); color: inherit; font: inherit; }
+</style>

package/src/chat/playback.js

@@ -8,10 +8,15 @@ const SILENT_WAV = "data:audio/wav;base64,UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAE
 export function createAnswerPlayer() {
   const audio = typeof Audio === "function" ? new Audio() : null;
   let url = "";
-  const release = () => {
-    if (!url) return;
-    URL.revokeObjectURL(url);
-    url = "";
+  let settle = null;
+  const finish = (played) => {
+    const done = settle;
+    settle = null;
+    if (url) {
+      URL.revokeObjectURL(url);
+      url = "";
+    }
+    done?.(played);
   };
   return {
     unlock() {
@@ -19,23 +24,27 @@ export function createAnswerPlayer() {
       audio.src = SILENT_WAV;
       void audio.play().catch(() => {});
     },
+    /** Resolves when the clip ends (true), or is stopped or fails (false). */
     async play(blob) {
       if (!audio || !blob) return false;
-      release();
+      finish(false);
       url = URL.createObjectURL(blob);
       audio.src = url;
-      audio.onended = release;
+      const ended = new Promise((resolve) => {
+        settle = resolve;
+      });
+      audio.onended = () => finish(true);
+      audio.onerror = () => finish(false);
       try {
         await audio.play();
-        return true;
       } catch {
-        release();
-        return false;
+        finish(false);
       }
+      return ended;
     },
     stop() {
       audio?.pause();
-      release();
+      finish(false);
     },
   };
 }

package/src/host.js

@@ -12,7 +12,7 @@ import {
   rawFileUrl,
 } from "@olares/lares-core/files/preview-workspace";
 import { FILES_UPLOAD_PATH, uploadFile } from "@olares/lares-core/files/upload-client";
-import { API as VOICE_API, postTranscribe } from "@olares/lares-core/voice/client";
+import { API as VOICE_API, postTranscribe, voiceHeaders } from "@olares/lares-core/voice/client";
 import { postReport, REPORT_PATH } from "@olares/lares-core/log/report";
 
 function pageOrigin() {
@@ -270,6 +270,11 @@ export function createHostClient(ports = {}) {
         request: direct,
       });
     },
+    voiceSocket(ticket) {
+      const auth = socketProtocols(ports);
+      return new WebSocket(muxWsUrl(urlFor(`${VOICE_API}/realtime`), pageOrigin()), ["lares-voice", `lares-ticket.${ticket}`, ...(Array.isArray(auth) ? auth : auth ? [auth] : [])]);
+    },
+    voiceRequest(path, options) { return request(path, options); },
     transcribe(blob, language, signal) {
       return postTranscribe(blob, language, signal, {
         baseUrl: urlFor(VOICE_API),
@@ -278,13 +283,19 @@ export function createHostClient(ports = {}) {
     },
     reportLog(scope, event, fields) {
       return postReport({ scope, event, ...(fields ?? {}) }, {
-        request,
+        request: direct,
         url: urlFor(REPORT_PATH),
       });
     },
-    async speak(text) {
-      const res = await request(`${VOICE_API}/speak`, { method: "POST", body: { text } });
+    /** Hands over a streaming answer's text; with `part`, also returns that sentence's audio. */
+    async speak({ key, text, done, part }) {
+      const res = await request(`${VOICE_API}/speak`, {
+        method: "POST",
+        headers: voiceHeaders("device"),
+        body: Number.isInteger(part) ? { key, text, done, part } : { key, text, done },
+      });
       const body = res?.body;
+      if (res?.ok && body && !Number.isInteger(part)) return { sentences: Number(body.sentences) || 0 };
       if (res?.ok && body && typeof body.audio === "string") {
         return {
           bluetooth: body.bluetooth === "played" ? "played" : "skipped",
@@ -297,7 +308,7 @@ export function createHostClient(ports = {}) {
       throw Object.assign(new Error(message), { code });
     },
     async interruptSpeech() {
-      const res = await request(`${VOICE_API}/playback/stop`, { method: "POST" });
+      const res = await request(`${VOICE_API}/playback/stop`, { method: "POST", headers: voiceHeaders("device") });
       return Boolean(res?.ok);
     },
     async openMux(signal) {

package/src/mobile/MobileShell.vue

@@ -72,6 +72,8 @@
             :can-send="state.canSend"
             :attach-pending="state.upload.pending > 0"
             :attach-disabled="!state.sessionId || state.historyLoading || state.starting"
+            :voice-call="state.voiceCall"
+            :voice-default-mode="state.voiceDefaultMode"
             :voice-phase="state.voicePhase"
             :voice-elapsed="state.voiceElapsed"
             :voice-error="state.voiceErrorText"
@@ -91,6 +93,7 @@
             @voice="$emit('voice')"
             @hold-start="$emit('hold-start', $event)"
             @hold-end="$emit('hold-end', $event)"
+            @voice-action="$emit('voice-action', $event)"
             @open-commands="$emit('open-commands')"
             @open-references="$emit('open-references', $event)"
             @pick-command="$emit('pick-command', $event)"
@@ -133,6 +136,8 @@
             :can-send="state.canSend"
             :attach-pending="state.upload.pending > 0"
             :attach-disabled="!state.sessionId || state.historyLoading || state.starting"
+            :voice-call="state.voiceCall"
+            :voice-default-mode="state.voiceDefaultMode"
             :voice-phase="state.voicePhase"
             :voice-elapsed="state.voiceElapsed"
             :voice-error="state.voiceErrorText"
@@ -152,6 +157,7 @@
             @voice="$emit('voice')"
             @hold-start="$emit('hold-start', $event)"
             @hold-end="$emit('hold-end', $event)"
+            @voice-action="$emit('voice-action', $event)"
             @open-commands="$emit('open-commands')"
             @open-references="$emit('open-references', $event)"
             @pick-command="$emit('pick-command', $event)"
@@ -249,7 +255,7 @@ export default {
   emits: [
     "history", "create", "toggle-speak", "close-history", "pick-session", "retry", "open-file",
     "answer", "update-draft", "send", "queue-action", "open-model", "open-effort", "files",
-    "voice", "hold-start", "hold-end", "close-model", "choose-model",
+    "voice", "voice-action", "hold-start", "hold-end", "close-model", "choose-model",
     "close-effort", "choose-effort", "close-preview",
     "open-commands", "open-references", "pick-command", "remove-image",
     "refresh-workspaces", "browse-workspaces", "browse-directory",

package/src/settings/AgentSettings.vue

@@ -41,7 +41,6 @@
       <p class="lares-agent__hint">{{ t("settings.input.busyEnter.hint") }}</p>
       <p v-if="conversationError" class="lares-agent__hint" data-status="error">{{ conversationError }}</p>
     </section>
-    <!-- Voice input parked.
     <section class="lares-agent__group">
       <h2>{{ tVoice("settings.title") }}</h2>
       <div class="lares-agent__list">
@@ -62,9 +61,43 @@
           @click="openSheet('voiceLang')"
         />
       </div>
+      <label class="lares-agent__hint">{{ tVoice('settings.preset') }}
+        <select :value="voice?.config?.preset || 'assistant'" @change="patchVoice('preset', $event.target.value)">
+          <option v-for="preset in ['dictation', 'chat', 'assistant', 'direct']" :key="preset" :value="preset">{{ tVoice(`settings.preset.${preset}`) }}</option>
+        </select>
+      </label>
+      <label class="lares-agent__hint">{{ tVoice('settings.chatModel') }}
+        <select :value="voice?.config?.chatModel || 'default-chat'" @change="patchVoice('chatModel', $event.target.value)">
+          <option value="default-chat">default-chat</option>
+          <option v-for="model in voice?.chatModels || []" :key="model" :value="model">{{ model }}</option>
+        </select>
+      </label>
+      <label class="lares-agent__hint">{{ tVoice('settings.backend') }}
+        <select :value="voice?.config?.backend || 'pipeline'" @change="patchVoice('backend', $event.target.value)">
+          <option value="pipeline">{{ tVoice('settings.backend.pipeline') }}</option>
+          <option v-if="voice?.realtimeReady?.includes(voice?.config?.realtimeModel)" value="realtime">Realtime</option>
+        </select>
+      </label>
+      <label v-if="voice?.realtimeModels?.length" class="lares-agent__hint">{{ tVoice('settings.realtime.model') }}
+        <select :value="voice?.config?.realtimeModel || ''" @change="patchVoice('realtimeModel', $event.target.value)">
+          <option value="" disabled>—</option>
+          <option v-for="model in voice.realtimeModels" :key="model" :value="model">{{ model }}</option>
+        </select>
+      </label>
+      <p class="lares-agent__hint">{{ tVoice('settings.realtime.hint') }}</p>
+      <label class="lares-agent__hint">{{ tVoice('settings.mobileMode') }}
+        <select :value="voice?.config?.mobileMode || 'continuous'" @change="patchVoice('mobileMode', $event.target.value)">
+          <option value="single">{{ tVoice('settings.mode.single') }}</option>
+          <option value="continuous">{{ tVoice('settings.mode.continuous') }}</option>
+        </select>
+      </label>
+      <label class="lares-agent__hint">{{ tVoice('settings.device') }}
+        <select :value="voice?.config?.device || 'auto'" @change="patchVoice('device', $event.target.value)">
+          <option v-for="device in ['auto', 'device', 'headset']" :key="device" :value="device">{{ tVoice(`settings.device.${device}`) }}</option>
+        </select>
+      </label>
       <p v-if="voiceError" class="lares-agent__hint" data-status="error">{{ voiceError }}</p>
     </section>
-    -->
     <section class="lares-agent__group">
       <h2>{{ tSearch("settings.title") }}</h2>
       <LaresSelect

package/src/chat/voice.js

@@ -1,269 +0,0 @@
-import {
-  formatElapsed,
-  isRecordingTooShort,
-  pickRecordingMime,
-} from "@olares/lares-core/voice/recorder";
-
-export { formatElapsed };
-
-/** Phone mic. Bare booleans are exact and iOS rejects the headset profile. */
-const DEVICE_AUDIO = {
-  channelCount: { ideal: 1 },
-  echoCancellation: { ideal: true },
-  noiseSuppression: { ideal: true },
-  autoGainControl: { ideal: true },
-};
-
-const CHUNK_MS = 250;
-
-function micFailure(err) {
-  const name = err?.name || "";
-  if (name === "NotAllowedError" || name === "PermissionDeniedError") return "voice_permission_denied";
-  if (name === "SecurityError") return "voice_unsupported";
-  return "voice_failed";
-}
-
-async function openDeviceMic(mediaDevices) {
-  try {
-    return await mediaDevices.getUserMedia({ audio: DEVICE_AUDIO });
-  } catch (err) {
-    if (micFailure(err) !== "voice_failed") throw err;
-    return mediaDevices.getUserMedia({ audio: true });
-  }
-}
-
-export function createVoiceCapture({
-  transcribe,
-  submit,
-  onPhase,
-  onTrace,
-  now = Date.now,
-  media,
-}) {
-  const mediaDevices = media?.mediaDevices ?? globalThis.navigator?.mediaDevices;
-  const Recorder = media?.MediaRecorder ?? globalThis.MediaRecorder;
-  let recorder = null;
-  let stream = null;
-  let chunks = [];
-  let started = 0;
-  let timer = 0;
-  let errorTimer = 0;
-  let finishTimer = 0;
-  let cancel = false;
-  let released = false;
-  let finished = false;
-  let dead = false;
-  let abort = null;
-  let phase = "idle";
-
-  const emit = (next, extra = {}) => {
-    phase = next;
-    onPhase?.(next, extra);
-  };
-
-  const trace = (event, fields) => {
-    try {
-      onTrace?.(event, fields);
-    } catch {
-      /* a trace must not change the take */
-    }
-  };
-
-  const stopTimer = () => {
-    if (timer) {
-      clearInterval(timer);
-      timer = 0;
-    }
-  };
-
-  const release = () => {
-    stream?.getTracks().forEach((track) => track.stop());
-    stream = null;
-    recorder = null;
-  };
-
-  const fail = (code) => {
-    if (dead) return;
-    if (errorTimer) clearTimeout(errorTimer);
-    emit("error", { error: code, elapsed: 0 });
-    errorTimer = setTimeout(() => {
-      errorTimer = 0;
-      if (!dead && phase === "error") emit("idle", { error: "", elapsed: 0 });
-    }, 3200);
-  };
-
-  const finish = async () => {
-    if (finished || dead) return;
-    finished = true;
-    stopTimer();
-    const took = now() - started;
-    const type = recorder?.mimeType || chunks[0]?.type || "audio/webm";
-    const blobChunks = chunks;
-    chunks = [];
-    release();
-    if (cancel) {
-      emit("idle", { error: "", elapsed: 0 });
-      return;
-    }
-    const blob = new Blob(blobChunks, { type });
-    trace("blob", { bytes: blob.size, type, ms: took, chunks: blobChunks.length });
-    if (isRecordingTooShort(took, blob.size)) {
-      trace("too-short", { bytes: blob.size, ms: took });
-      fail("voice_too_short");
-      return;
-    }
-    emit("transcribing", { elapsed: took, error: "" });
-    abort = new AbortController();
-    try {
-      const text = String(await transcribe(blob, abort.signal) ?? "").trim();
-      if (dead || cancel) return;
-      if (!text) {
-        trace("no-input", {});
-        fail("voice_no_input");
-        return;
-      }
-      trace("transcript", { chars: text.length, head: text.slice(0, 40) });
-      emit("idle", { error: "", elapsed: 0 });
-      submit?.(text);
-    } catch (err) {
-      if (dead || abort.signal.aborted || cancel) return;
-      const code = err instanceof Error ? err.message : "voice_failed";
-      trace("transcribe-error", { code });
-      fail(code);
-    } finally {
-      abort = null;
-    }
-  };
-
-  const scheduleFinish = () => {
-    if (finishTimer) clearTimeout(finishTimer);
-    finishTimer = setTimeout(() => {
-      finishTimer = 0;
-      void finish();
-    }, 30);
-  };
-
-  const endRecorder = () => {
-    if (!recorder || recorder.state === "inactive") {
-      scheduleFinish();
-      return;
-    }
-    try {
-      recorder.requestData?.();
-    } catch {
-      /* the final dataavailable still arrives from stop() */
-    }
-    try {
-      recorder.stop();
-    } catch {
-      scheduleFinish();
-    }
-  };
-
-  return {
-    get phase() {
-      return phase;
-    },
-    async start() {
-      if (dead || phase === "preparing" || phase === "recording" || phase === "transcribing") return;
-      if (typeof Recorder === "undefined" || !mediaDevices?.getUserMedia) {
-        fail("voice_unsupported");
-        return;
-      }
-      cancel = false;
-      released = false;
-      finished = false;
-      trace("open", {});
-      emit("preparing", { elapsed: 0, error: "" });
-      let next;
-      try {
-        next = await openDeviceMic(mediaDevices);
-      } catch (err) {
-        const code = micFailure(err);
-        trace("mic-fail", { code, name: err?.name || "" });
-        if (!dead) fail(code);
-        return;
-      }
-      if (dead || cancel) {
-        next.getTracks().forEach((track) => track.stop());
-        if (!dead && phase === "preparing") emit("idle", { error: "", elapsed: 0 });
-        return;
-      }
-      stream = next;
-      chunks = [];
-      const mime = pickRecordingMime(Recorder);
-      try {
-        recorder = new Recorder(stream, mime ? { mimeType: mime } : undefined);
-      } catch {
-        release();
-        fail("voice_unsupported");
-        return;
-      }
-      recorder.ondataavailable = (event) => {
-        if (event.data?.size) chunks.push(event.data);
-        if (recorder?.state === "inactive") scheduleFinish();
-      };
-      recorder.onstop = () => scheduleFinish();
-      started = now();
-      try {
-        recorder.start(CHUNK_MS);
-      } catch {
-        trace("record-fail", { mime });
-        release();
-        fail("voice_unsupported");
-        return;
-      }
-      trace("recording", { mime: recorder.mimeType || mime || "" });
-      if (released || cancel) {
-        endRecorder();
-        return;
-      }
-      emit("recording", { elapsed: 0, error: "" });
-      timer = setInterval(() => emit("recording", { elapsed: now() - started, error: "" }), 200);
-    },
-    /**
-     * Finger-up finishes the take. The mic opens asynchronously, so the finger
-     * often lifts while still preparing — that is not a cancel.
-     */
-    stop(aborting = false) {
-      if (phase === "transcribing") {
-        if (aborting) {
-          cancel = true;
-          abort?.abort();
-        }
-        return;
-      }
-      if (phase !== "preparing" && phase !== "recording") {
-        trace("stop-ignored", { aborting: Boolean(aborting), phase });
-        return;
-      }
-      trace("stop", {
-        aborting: Boolean(aborting),
-        phase,
-        recorder: recorder?.state || "none",
-      });
-      if (aborting) {
-        cancel = true;
-        released = false;
-        if (recorder && recorder.state !== "inactive") recorder.stop();
-        return;
-      }
-      released = true;
-      stopTimer();
-      if (recorder && recorder.state !== "inactive") {
-        emit("transcribing", { elapsed: now() - started, error: "" });
-        endRecorder();
-      }
-    },
-    dispose() {
-      dead = true;
-      cancel = true;
-      if (errorTimer) clearTimeout(errorTimer);
-      if (finishTimer) clearTimeout(finishTimer);
-      stopTimer();
-      abort?.abort();
-      if (recorder && recorder.state !== "inactive") recorder.stop();
-      release();
-    },
-  };
-}