@hivegpt/hiveai-angular 0.0.581 → 0.0.583

package/fesm2015/hivegpt-hiveai-angular.js

@@ -5,15 +5,14 @@ import { HttpClient, HttpHeaders } from '@angular/common/http';
 import * as i0 from '@angular/core';
 import { Injectable, InjectionToken, Inject, PLATFORM_ID, Optional, NgZone, EventEmitter, Component, Injector, Output, Input, ElementRef, ChangeDetectionStrategy, ChangeDetectorRef, Renderer2, ViewContainerRef, ViewChild, ViewChildren, NgModule, Pipe } from '@angular/core';
 import { DomSanitizer } from '@angular/platform-browser';
-import { BehaviorSubject, of, throwError, Subject, Subscription, combineLatest } from 'rxjs';
-import { switchMap, catchError, filter, take, map, takeUntil, tap } from 'rxjs/operators';
+import { BehaviorSubject, of, throwError, Subject, Subscription, merge, concat, timer, combineLatest } from 'rxjs';
+import { switchMap, catchError, filter, take, map, takeUntil, distinctUntilChanged, startWith, tap } from 'rxjs/operators';
 import { isPlatformBrowser, CommonModule, DOCUMENT } from '@angular/common';
 import { Socket } from 'ngx-socket-io';
 import { Validators, FormBuilder, FormsModule, ReactiveFormsModule } from '@angular/forms';
 import * as SpeechSDK from 'microsoft-cognitiveservices-speech-sdk';
 import * as marked from 'marked';
 import { __awaiter } from 'tslib';
-import Daily from '@daily-co/daily-js';
 import { MatIconModule } from '@angular/material/icon';
 import { MatSidenavModule } from '@angular/material/sidenav';
 import { QuillModule } from 'ngx-quill';
@@ -684,8 +683,8 @@ BotsService.ctorParameters = () => [
 ];
 
 /**
- * Audio analyzer for waveform visualization only.
- * Do NOT use isUserSpeaking$ for call state; speaking state must come from Daily.js.
+ * Audio analyzer for waveform visualization and local (mic) speaking detection.
+ * VoiceAgentService may combine this with WebSocket server events for call state.
  */
 class AudioAnalyzerService {
     constructor() {
@@ -806,70 +805,77 @@ AudioAnalyzerService.decorators = [
 ];
 
 /**
- * WebSocket-only client for voice agent signaling.
+ * Native WebSocket client for voice session (signaling, transcripts, speaking hints).
  * CRITICAL: Uses native WebSocket only. NO Socket.IO, NO ngx-socket-io.
  *
- * Responsibilities:
- * - Connect to ws_url (from POST /ai/ask-voice response)
- * - Parse JSON messages (room_created, user_transcript, bot_transcript)
- * - Emit roomCreated$, userTranscript$, botTranscript$
- * - NO audio logic, NO mic logic. Audio is handled by Daily.js (WebRTC).
+ * Connects to `ws_url` from `POST {baseUrl}/ai/ask-voice-socket`.
+ * Parses JSON messages for transcripts and optional assistant/user speaking flags.
  */
 class WebSocketVoiceClientService {
-    constructor() {
+    constructor(ngZone) {
+        this.ngZone = ngZone;
         this.ws = null;
-        this.roomCreatedSubject = new Subject();
+        /** True when {@link disconnect} initiated the close (not counted as remote close). */
+        this.closeInitiatedByClient = false;
+        this.openedSubject = new Subject();
+        this.remoteCloseSubject = new Subject();
         this.userTranscriptSubject = new Subject();
         this.botTranscriptSubject = new Subject();
-        /** Emits room_url when backend sends room_created. */
-        this.roomCreated$ = this.roomCreatedSubject.asObservable();
-        /** Emits user transcript updates. */
+        this.assistantSpeakingSubject = new Subject();
+        this.serverUserSpeakingSubject = new Subject();
+        this.audioChunkSubject = new Subject();
+        /** Fires once each time the WebSocket reaches OPEN. */
+        this.opened$ = this.openedSubject.asObservable();
+        /** Fires when the socket closes without a client-initiated {@link disconnect}. */
+        this.remoteClose$ = this.remoteCloseSubject.asObservable();
         this.userTranscript$ = this.userTranscriptSubject.asObservable();
-        /** Emits bot transcript updates. */
         this.botTranscript$ = this.botTranscriptSubject.asObservable();
+        /** Assistant/bot speaking, when the server sends explicit events (see {@link handleJsonMessage}). */
+        this.assistantSpeaking$ = this.assistantSpeakingSubject.asObservable();
+        /** User speaking from server-side VAD, if provided. */
+        this.serverUserSpeaking$ = this.serverUserSpeakingSubject.asObservable();
+        /** Binary audio frames from server (when backend streams bot audio over WS). */
+        this.audioChunk$ = this.audioChunkSubject.asObservable();
     }
-    /** Connect to signaling WebSocket. No audio over this connection. */
     connect(wsUrl) {
         var _a;
         if (((_a = this.ws) === null || _a === void 0 ? void 0 : _a.readyState) === WebSocket.OPEN) {
             return;
         }
         if (this.ws) {
+            this.closeInitiatedByClient = true;
             this.ws.close();
-            this.ws = null;
         }
         try {
-            this.ws = new WebSocket(wsUrl);
-            this.ws.onmessage = (event) => {
-                var _a;
-                try {
-                    const msg = JSON.parse(event.data);
-                    if ((msg === null || msg === void 0 ? void 0 : msg.type) === 'room_created') {
-                        const roomUrl = ((_a = msg.room_url) !== null && _a !== void 0 ? _a : msg.roomUrl);
-                        if (typeof roomUrl === 'string') {
-                            this.roomCreatedSubject.next(roomUrl);
-                        }
-                    }
-                    else if ((msg === null || msg === void 0 ? void 0 : msg.type) === 'user_transcript' && typeof msg.text === 'string') {
-                        this.userTranscriptSubject.next({
-                            text: msg.text,
-                            final: msg.final === true,
-                        });
-                    }
-                    else if ((msg === null || msg === void 0 ? void 0 : msg.type) === 'bot_transcript' && typeof msg.text === 'string') {
-                        this.botTranscriptSubject.next(msg.text);
+            const socket = new WebSocket(wsUrl);
+            this.ws = socket;
+            socket.onopen = () => {
+                if (this.ws !== socket)
+                    return;
+                this.ngZone.run(() => this.openedSubject.next());
+            };
+            socket.onmessage = (event) => {
+                if (this.ws !== socket)
+                    return;
+                void this.handleIncomingMessage(event.data);
+            };
+            socket.onerror = () => {
+                this.ngZone.run(() => {
+                    if (this.ws === socket && socket.readyState !== WebSocket.CLOSED) {
+                        socket.close();
                     }
+                });
+            };
+            socket.onclose = () => {
+                if (this.ws === socket) {
+                    this.ws = null;
                 }
-                catch (_b) {
-                    // Ignore non-JSON or unknown messages
+                const client = this.closeInitiatedByClient;
+                this.closeInitiatedByClient = false;
+                if (!client) {
+                    this.ngZone.run(() => this.remoteCloseSubject.next());
                 }
             };
-            this.ws.onerror = () => {
-                this.disconnect();
-            };
-            this.ws.onclose = () => {
-                this.ws = null;
-            };
         }
         catch (err) {
             console.error('WebSocketVoiceClient: connect failed', err);
@@ -877,344 +883,140 @@ class WebSocketVoiceClientService {
             throw err;
         }
     }
-    /** Disconnect and cleanup. */
-    disconnect() {
-        if (this.ws) {
-            this.ws.close();
-            this.ws = null;
-        }
-    }
-    /** Whether the WebSocket is open. */
-    get isConnected() {
-        var _a;
-        return ((_a = this.ws) === null || _a === void 0 ? void 0 : _a.readyState) === WebSocket.OPEN;
-    }
-}
-WebSocketVoiceClientService.ɵprov = i0.ɵɵdefineInjectable({ factory: function WebSocketVoiceClientService_Factory() { return new WebSocketVoiceClientService(); }, token: WebSocketVoiceClientService, providedIn: "root" });
-WebSocketVoiceClientService.decorators = [
-    { type: Injectable, args: [{
-                providedIn: 'root',
-            },] }
-];
-
-/**
- * Daily.js WebRTC client for voice agent audio.
- * Responsibilities:
- * - Create and manage Daily CallObject
- * - Join Daily room using room_url
- * - Handle mic capture + speaker playback
- * - Bot speaking detection via AnalyserNode on remote track (instant)
- * - User speaking detection via active-speaker-change
- * - Expose speaking$ (bot speaking), userSpeaking$ (user speaking), micMuted$
- * - Expose localStream$ for waveform visualization (AudioAnalyzerService)
- */
-class DailyVoiceClientService {
-    constructor(ngZone) {
-        this.ngZone = ngZone;
-        this.callObject = null;
-        this.localStream = null;
-        this.localSessionId = null;
-        /** Explicit playback of remote (bot) audio; required in some browsers. */
-        this.remoteAudioElement = null;
-        /** AnalyserNode-based remote audio monitor for instant bot speaking detection. */
-        this.remoteAudioContext = null;
-        this.remoteSpeakingRAF = null;
-        this.speakingSubject = new BehaviorSubject(false);
-        this.userSpeakingSubject = new BehaviorSubject(false);
-        this.micMutedSubject = new BehaviorSubject(false);
-        this.localStreamSubject = new BehaviorSubject(null);
-        /** True when bot (remote participant) is the active speaker. */
-        this.speaking$ = this.speakingSubject.asObservable();
-        /** True when user (local participant) is the active speaker. */
-        this.userSpeaking$ = this.userSpeakingSubject.asObservable();
-        /** True when mic is muted. */
-        this.micMuted$ = this.micMutedSubject.asObservable();
-        /** Emits local mic stream for waveform visualization. */
-        this.localStream$ = this.localStreamSubject.asObservable();
-    }
-    /**
-     * Connect to Daily room. Acquires mic first for waveform, then joins with audio.
-     * @param roomUrl Daily room URL (from room_created)
-     * @param token Optional meeting token
-     */
-    connect(roomUrl, token) {
+    handleIncomingMessage(payload) {
         return __awaiter(this, void 0, void 0, function* () {
-            if (this.callObject) {
-                yield this.disconnect();
+            if (typeof payload === 'string') {
+                this.handleJsonString(payload);
+                return;
             }
-            try {
-                // Get mic stream for both Daily and waveform (single capture)
-                const stream = yield navigator.mediaDevices.getUserMedia({ audio: true });
-                const audioTrack = stream.getAudioTracks()[0];
-                if (!audioTrack) {
-                    stream.getTracks().forEach((t) => t.stop());
-                    throw new Error('No audio track');
-                }
-                this.localStream = stream;
-                this.localStreamSubject.next(stream);
-                // Create audio-only call object
-                // videoSource: false = no camera, audioSource = our mic track
-                const callObject = Daily.createCallObject({
-                    videoSource: false,
-                    audioSource: audioTrack,
-                });
-                this.callObject = callObject;
-                this.setupEventHandlers(callObject);
-                // Join room; Daily handles playback of remote (bot) audio automatically.
-                // Only pass token when it's a non-empty string (Daily rejects undefined/non-string).
-                const joinOptions = { url: roomUrl };
-                if (typeof token === 'string' && token.trim() !== '') {
-                    joinOptions.token = token;
-                }
-                yield callObject.join(joinOptions);
-                console.log(`[VoiceDebug] Room connected (Daily join complete) — ${new Date().toISOString()}`);
-                const participants = callObject.participants();
-                if (participants === null || participants === void 0 ? void 0 : participants.local) {
-                    this.localSessionId = participants.local.session_id;
-                }
-                // Initial mute state: Daily starts with audio on
-                this.micMutedSubject.next(!callObject.localAudio());
+            if (payload instanceof ArrayBuffer) {
+                this.handleBinaryMessage(payload);
+                return;
             }
-            catch (err) {
-                this.cleanup();
-                throw err;
+            if (payload instanceof Blob) {
+                const ab = yield payload.arrayBuffer();
+                this.handleBinaryMessage(ab);
             }
         });
     }
-    setupEventHandlers(call) {
-        // active-speaker-change: used ONLY for user speaking detection.
-        // Bot speaking is detected by our own AnalyserNode (instant, no debounce).
-        call.on('active-speaker-change', (event) => {
-            this.ngZone.run(() => {
-                var _a;
-                const peerId = (_a = event === null || event === void 0 ? void 0 : event.activeSpeaker) === null || _a === void 0 ? void 0 : _a.peerId;
-                if (!peerId || !this.localSessionId) {
-                    this.userSpeakingSubject.next(false);
-                    return;
-                }
-                const isLocal = peerId === this.localSessionId;
-                this.userSpeakingSubject.next(isLocal);
-            });
-        });
-        // track-started / track-stopped: set up remote audio playback + AnalyserNode monitor.
-        call.on('track-started', (event) => {
-            this.ngZone.run(() => {
-                var _a, _b, _c, _d;
-                const p = event === null || event === void 0 ? void 0 : event.participant;
-                const type = (_a = event === null || event === void 0 ? void 0 : event.type) !== null && _a !== void 0 ? _a : (_b = event === null || event === void 0 ? void 0 : event.track) === null || _b === void 0 ? void 0 : _b.kind;
-                const track = event === null || event === void 0 ? void 0 : event.track;
-                if (p && !p.local && type === 'audio') {
-                    console.log(`[VoiceDebug] Got audio track from backend (track-started) — readyState=${track === null || track === void 0 ? void 0 : track.readyState}, muted=${track === null || track === void 0 ? void 0 : track.muted} — ${new Date().toISOString()}`);
-                    const audioTrack = track !== null && track !== void 0 ? track : (_d = (_c = p.tracks) === null || _c === void 0 ? void 0 : _c.audio) === null || _d === void 0 ? void 0 : _d.track;
-                    if (audioTrack && typeof audioTrack === 'object') {
-                        this.playRemoteTrack(audioTrack);
-                        this.monitorRemoteAudio(audioTrack);
-                    }
-                }
-            });
-        });
-        call.on('track-stopped', (event) => {
-            this.ngZone.run(() => {
-                var _a, _b;
-                const p = event === null || event === void 0 ? void 0 : event.participant;
-                const type = (_a = event === null || event === void 0 ? void 0 : event.type) !== null && _a !== void 0 ? _a : (_b = event === null || event === void 0 ? void 0 : event.track) === null || _b === void 0 ? void 0 : _b.kind;
-                if (p && !p.local && type === 'audio') {
-                    this.stopRemoteAudioMonitor();
-                    this.stopRemoteAudio();
-                }
-            });
-        });
-        call.on('left-meeting', () => {
-            this.ngZone.run(() => this.cleanup());
-        });
-        call.on('error', (event) => {
-            this.ngZone.run(() => {
-                var _a;
-                console.error('DailyVoiceClient: Daily error', (_a = event === null || event === void 0 ? void 0 : event.errorMsg) !== null && _a !== void 0 ? _a : event);
-                this.cleanup();
-            });
-        });
-    }
-    /**
-     * Play remote (bot) audio track via a dedicated audio element.
-     * Required in many browsers where Daily's internal playback does not output to speakers.
-     */
-    playRemoteTrack(track) {
-        this.stopRemoteAudio();
+    handleJsonString(jsonText) {
         try {
-            console.log(`[VoiceDebug] playRemoteTrack called — track.readyState=${track.readyState}, track.muted=${track.muted} — ${new Date().toISOString()}`);
-            track.onunmute = () => {
-                console.log(`[VoiceDebug] Remote audio track UNMUTED (audio data arriving) — ${new Date().toISOString()}`);
-            };
-            const stream = new MediaStream([track]);
-            const audio = new Audio();
-            audio.autoplay = true;
-            audio.srcObject = stream;
-            this.remoteAudioElement = audio;
-            audio.onplaying = () => {
-                console.log(`[VoiceDebug] Audio element PLAYING (browser started playback) — ${new Date().toISOString()}`);
-            };
-            let firstTimeUpdate = true;
-            audio.ontimeupdate = () => {
-                if (firstTimeUpdate) {
-                    firstTimeUpdate = false;
-                    console.log(`[VoiceDebug] Audio element first TIMEUPDATE (actual audio output) — ${new Date().toISOString()}`);
-                }
-            };
-            const p = audio.play();
-            if (p && typeof p.then === 'function') {
-                p.then(() => {
-                    console.log(`[VoiceDebug] audio.play() resolved — ${new Date().toISOString()}`);
-                }).catch((err) => {
-                    console.warn('DailyVoiceClient: remote audio play failed (may need user gesture)', err);
-                });
-            }
+            const msg = JSON.parse(jsonText);
+            this.ngZone.run(() => this.handleJsonMessage(msg));
         }
-        catch (err) {
-            console.warn('DailyVoiceClient: failed to create remote audio element', err);
+        catch (_a) {
+            // Ignore non-JSON
         }
     }
-    /**
-     * Monitor remote audio track energy via AnalyserNode.
-     * Polls at ~60fps and flips speakingSubject based on actual audio energy.
-     */
-    monitorRemoteAudio(track) {
-        this.stopRemoteAudioMonitor();
+    handleBinaryMessage(buffer) {
+        // Some backends wrap JSON events inside binary WS frames.
+        const maybeText = this.tryDecodeUtf8(buffer);
+        if (maybeText !== null) {
+            this.handleJsonString(maybeText);
+            return;
+        }
+        // Otherwise treat binary as streamed assistant audio.
+        this.ngZone.run(() => this.audioChunkSubject.next(buffer));
+    }
+    tryDecodeUtf8(buffer) {
         try {
-            const ctx = new AudioContext();
-            const source = ctx.createMediaStreamSource(new MediaStream([track]));
-            const analyser = ctx.createAnalyser();
-            analyser.fftSize = 256;
-            source.connect(analyser);
-            this.remoteAudioContext = ctx;
-            const dataArray = new Uint8Array(analyser.frequencyBinCount);
-            const THRESHOLD = 5;
-            const SILENCE_MS = 1500;
-            let lastSoundTime = 0;
-            let isSpeaking = false;
-            const poll = () => {
-                if (!this.remoteAudioContext)
-                    return;
-                analyser.getByteFrequencyData(dataArray);
-                let sum = 0;
-                for (let i = 0; i < dataArray.length; i++) {
-                    sum += dataArray[i];
-                }
-                const avg = sum / dataArray.length;
-                const now = Date.now();
-                if (avg > THRESHOLD) {
-                    lastSoundTime = now;
-                    if (!isSpeaking) {
-                        isSpeaking = true;
-                        console.log(`[VoiceDebug] Bot audio energy detected (speaking=true) — avg=${avg.toFixed(1)} — ${new Date().toISOString()}`);
-                        this.ngZone.run(() => {
-                            this.userSpeakingSubject.next(false);
-                            this.speakingSubject.next(true);
-                        });
-                    }
-                }
-                else if (isSpeaking && now - lastSoundTime > SILENCE_MS) {
-                    isSpeaking = false;
-                    console.log(`[VoiceDebug] Bot audio silence detected (speaking=false) — ${new Date().toISOString()}`);
-                    this.ngZone.run(() => this.speakingSubject.next(false));
-                }
-                this.remoteSpeakingRAF = requestAnimationFrame(poll);
-            };
-            this.remoteSpeakingRAF = requestAnimationFrame(poll);
+            const text = new TextDecoder('utf-8', { fatal: true }).decode(buffer);
+            const trimmed = text.trim();
+            if (!trimmed || (trimmed[0] !== '{' && trimmed[0] !== '[')) {
+                return null;
+            }
+            return trimmed;
         }
-        catch (err) {
-            console.warn('DailyVoiceClient: failed to create remote audio monitor', err);
+        catch (_a) {
+            return null;
         }
     }
-    stopRemoteAudioMonitor() {
-        if (this.remoteSpeakingRAF) {
-            cancelAnimationFrame(this.remoteSpeakingRAF);
-            this.remoteSpeakingRAF = null;
+    handleJsonMessage(msg) {
+        const type = msg.type;
+        const typeStr = typeof type === 'string' ? type : '';
+        if (typeStr === 'session_ready' || typeStr === 'connected' || typeStr === 'voice_session_started') {
+            return;
         }
-        if (this.remoteAudioContext) {
-            this.remoteAudioContext.close().catch(() => { });
-            this.remoteAudioContext = null;
+        if (typeStr === 'assistant_speaking' ||
+            typeStr === 'bot_speaking') {
+            if (msg.active === true || msg.speaking === true) {
+                this.assistantSpeakingSubject.next(true);
+            }
+            else if (msg.active === false || msg.speaking === false) {
+                this.assistantSpeakingSubject.next(false);
+            }
+            return;
         }
-    }
-    stopRemoteAudio() {
-        if (this.remoteAudioElement) {
-            try {
-                this.remoteAudioElement.pause();
-                this.remoteAudioElement.srcObject = null;
+        if (typeStr === 'user_speaking') {
+            if (msg.active === true || msg.speaking === true) {
+                this.serverUserSpeakingSubject.next(true);
+            }
+            else if (msg.active === false || msg.speaking === false) {
+                this.serverUserSpeakingSubject.next(false);
             }
-            catch (_) { }
-            this.remoteAudioElement = null;
+            return;
         }
-    }
-    /** Set mic muted state. */
-    setMuted(muted) {
-        if (!this.callObject)
+        if (typeStr === 'input_audio_buffer.speech_started') {
+            this.serverUserSpeakingSubject.next(true);
+            return;
+        }
+        if (typeStr === 'input_audio_buffer.speech_stopped') {
+            this.serverUserSpeakingSubject.next(false);
+            return;
+        }
+        if (typeStr === 'response.audio.delta') {
+            this.assistantSpeakingSubject.next(true);
+            return;
+        }
+        if (typeStr === 'response.audio.done' ||
+            typeStr === 'response.output_audio.done') {
+            this.assistantSpeakingSubject.next(false);
+            return;
+        }
+        if (typeStr === 'user_transcript' && typeof msg.text === 'string') {
+            this.userTranscriptSubject.next({
+                text: msg.text,
+                final: msg.final === true,
+            });
             return;
-        this.callObject.setLocalAudio(!muted);
-        this.micMutedSubject.next(muted);
+        }
+        if (typeStr === 'bot_transcript' && typeof msg.text === 'string') {
+            this.botTranscriptSubject.next(msg.text);
+        }
     }
-    /** Disconnect and cleanup. */
     disconnect() {
-        return __awaiter(this, void 0, void 0, function* () {
-            if (!this.callObject) {
-                this.cleanup();
-                return;
-            }
-            try {
-                yield this.callObject.leave();
-            }
-            catch (e) {
-                // ignore
-            }
-            this.cleanup();
-        });
-    }
-    cleanup() {
-        this.stopRemoteAudioMonitor();
-        this.stopRemoteAudio();
-        if (this.callObject) {
-            this.callObject.destroy().catch(() => { });
-            this.callObject = null;
-        }
-        if (this.localStream) {
-            this.localStream.getTracks().forEach((t) => t.stop());
-            this.localStream = null;
+        if (!this.ws) {
+            return;
         }
-        this.localSessionId = null;
-        this.speakingSubject.next(false);
-        this.userSpeakingSubject.next(false);
-        this.localStreamSubject.next(null);
-        // Keep last micMuted state; will reset on next connect
+        this.closeInitiatedByClient = true;
+        this.ws.close();
+    }
+    get isConnected() {
+        var _a;
+        return ((_a = this.ws) === null || _a === void 0 ? void 0 : _a.readyState) === WebSocket.OPEN;
     }
 }
-DailyVoiceClientService.ɵprov = i0.ɵɵdefineInjectable({ factory: function DailyVoiceClientService_Factory() { return new DailyVoiceClientService(i0.ɵɵinject(i0.NgZone)); }, token: DailyVoiceClientService, providedIn: "root" });
-DailyVoiceClientService.decorators = [
+WebSocketVoiceClientService.ɵprov = i0.ɵɵdefineInjectable({ factory: function WebSocketVoiceClientService_Factory() { return new WebSocketVoiceClientService(i0.ɵɵinject(i0.NgZone)); }, token: WebSocketVoiceClientService, providedIn: "root" });
+WebSocketVoiceClientService.decorators = [
     { type: Injectable, args: [{
                 providedIn: 'root',
             },] }
 ];
-DailyVoiceClientService.ctorParameters = () => [
+WebSocketVoiceClientService.ctorParameters = () => [
     { type: NgZone }
 ];
 
 /**
- * Voice agent orchestrator. Coordinates WebSocket (signaling) and Daily.js (WebRTC audio).
- *
- * CRITICAL: This service must NEVER use Socket.IO or ngx-socket-io. Voice flow uses only:
- * - Native WebSocket (WebSocketVoiceClientService) for signaling (room_created, transcripts)
- * - Daily.js (DailyVoiceClientService) for WebRTC audio. Audio does NOT flow over WebSocket.
- *
- * - Maintains callState, statusText, duration, isMicMuted, isUserSpeaking, audioLevels
- * - Uses WebSocket for room_created and transcripts only (no audio)
- * - Uses Daily.js for all audio, mic, and real-time speaking detection
+ * Voice agent orchestrator: single WebSocket (`ws_url` from POST /ai/ask-voice-socket)
+ * for session events, transcripts, and optional speaking hints; local mic for capture
+ * and waveform only (no Daily/WebRTC room).
  */
 class VoiceAgentService {
-    constructor(audioAnalyzer, wsClient, dailyClient, platformTokenRefresh, 
+    constructor(audioAnalyzer, wsClient, platformTokenRefresh, 
     /** `Object` not `object` — ngc metadata collection rejects the `object` type in DI params. */
     platformId) {
         this.audioAnalyzer = audioAnalyzer;
         this.wsClient = wsClient;
-        this.dailyClient = dailyClient;
         this.platformTokenRefresh = platformTokenRefresh;
         this.platformId = platformId;
         this.callStateSubject = new BehaviorSubject('idle');
@@ -1227,6 +1029,11 @@ class VoiceAgentService {
         this.botTranscriptSubject = new Subject();
         this.callStartTime = 0;
         this.durationInterval = null;
+        this.localMicStream = null;
+        this.remoteAudioContext = null;
+        this.pendingRemoteAudio = [];
+        this.remoteAudioPlaying = false;
+        this.endCall$ = new Subject();
         this.subscriptions = new Subscription();
         this.destroy$ = new Subject();
         this.callState$ = this.callStateSubject.asObservable();
@@ -1237,8 +1044,13 @@ class VoiceAgentService {
         this.audioLevels$ = this.audioLevelsSubject.asObservable();
         this.userTranscript$ = this.userTranscriptSubject.asObservable();
         this.botTranscript$ = this.botTranscriptSubject.asObservable();
-        // Waveform visualization only - do NOT use for speaking state
         this.subscriptions.add(this.audioAnalyzer.audioLevels$.subscribe((levels) => this.audioLevelsSubject.next(levels)));
+        this.subscriptions.add(this.wsClient.remoteClose$
+            .pipe(takeUntil(this.destroy$))
+            .subscribe(() => void this.handleRemoteClose()));
+        this.subscriptions.add(this.wsClient.audioChunk$
+            .pipe(takeUntil(this.destroy$))
+            .subscribe((chunk) => this.enqueueRemoteAudio(chunk)));
     }
     ngOnDestroy() {
         this.destroy$.next();
@@ -1249,11 +1061,13 @@ class VoiceAgentService {
     resetToIdle() {
         if (this.callStateSubject.value === 'idle')
             return;
+        this.endCall$.next();
         this.stopDurationTimer();
+        this.callStartTime = 0;
         this.audioAnalyzer.stop();
+        this.stopLocalMic();
+        this.resetRemoteAudioPlayback();
         this.wsClient.disconnect();
-        // Fire-and-forget: Daily disconnect is async; connect() will await if needed
-        void this.dailyClient.disconnect();
         this.callStateSubject.next('idle');
         this.statusTextSubject.next('');
         this.durationSubject.next('0:00');
@@ -1268,9 +1082,6 @@ class VoiceAgentService {
                 this.callStateSubject.next('connecting');
                 this.statusTextSubject.next('Connecting...');
                 let accessToken = token;
-                // Align with chat drawer token handling: always delegate to
-                // PlatformTokenRefreshService when we have a usersApiUrl, so it can
-                // fall back to stored tokens even if the caller passed an empty token.
                 if (usersApiUrl && isPlatformBrowser(this.platformId)) {
                     try {
                         const ensured = yield this.platformTokenRefresh
@@ -1286,7 +1097,7 @@ class VoiceAgentService {
                     }
                 }
                 const baseUrl = apiUrl.replace(/\/$/, '');
-                const postUrl = `${baseUrl}/ai/ask-voice`;
+                const postUrl = `${baseUrl}/ai/ask-voice-socket`;
                 const headers = {
                     'Content-Type': 'application/json',
                     Authorization: `Bearer ${accessToken}`,
@@ -1298,7 +1109,6 @@ class VoiceAgentService {
                     eventToken,
                     'ngrok-skip-browser-warning': 'true',
                 };
-                // POST to get ws_url for signaling
                 const res = yield fetch(postUrl, {
                     method: 'POST',
                     headers,
@@ -1312,33 +1122,21 @@ class VoiceAgentService {
                     throw new Error(`HTTP ${res.status}`);
                 }
                 const json = yield res.json();
-                const wsUrl = json === null || json === void 0 ? void 0 : json.rn_ws_url;
-                if (!wsUrl || typeof wsUrl !== 'string') {
+                const wsUrl = (typeof (json === null || json === void 0 ? void 0 : json.ws_url) === 'string' && json.ws_url) ||
+                    (typeof (json === null || json === void 0 ? void 0 : json.rn_ws_url) === 'string' && json.rn_ws_url);
+                if (!wsUrl) {
                     throw new Error('No ws_url in response');
                 }
-                // Subscribe to room_created BEFORE connecting to avoid race
-                this.wsClient.roomCreated$
-                    .pipe(take(1), takeUntil(this.destroy$))
-                    .subscribe((roomUrl) => __awaiter(this, void 0, void 0, function* () {
-                    try {
-                        yield this.onRoomCreated(roomUrl);
-                    }
-                    catch (err) {
-                        console.error('Daily join failed:', err);
-                        this.callStateSubject.next('ended');
-                        this.statusTextSubject.next('Connection failed');
-                        yield this.disconnect();
-                        throw err;
-                    }
-                }));
-                // Forward transcripts from WebSocket
+                const untilCallEnds$ = merge(this.destroy$, this.endCall$);
                 this.subscriptions.add(this.wsClient.userTranscript$
-                    .pipe(takeUntil(this.destroy$))
+                    .pipe(takeUntil(untilCallEnds$))
                     .subscribe((t) => this.userTranscriptSubject.next(t)));
                 this.subscriptions.add(this.wsClient.botTranscript$
-                    .pipe(takeUntil(this.destroy$))
+                    .pipe(takeUntil(untilCallEnds$))
                     .subscribe((t) => this.botTranscriptSubject.next(t)));
-                // Connect signaling WebSocket (no audio over WS)
+                this.subscriptions.add(this.wsClient.opened$
+                    .pipe(takeUntil(untilCallEnds$), take(1))
+                    .subscribe(() => void this.onWebsocketOpened()));
                 this.wsClient.connect(wsUrl);
             }
             catch (error) {
@@ -1350,59 +1148,185 @@ class VoiceAgentService {
             }
         });
     }
-    onRoomCreated(roomUrl) {
+    onWebsocketOpened() {
         return __awaiter(this, void 0, void 0, function* () {
-            // Connect Daily.js for WebRTC audio
-            yield this.dailyClient.connect(roomUrl);
-            // Waveform: use local mic stream from Daily client
-            this.dailyClient.localStream$
-                .pipe(filter((s) => s != null), take(1))
-                .subscribe((stream) => {
-                this.audioAnalyzer.start(stream);
-            });
-            this.subscriptions.add(this.dailyClient.userSpeaking$.subscribe((s) => this.isUserSpeakingSubject.next(s)));
-            this.subscriptions.add(combineLatest([
-                this.dailyClient.speaking$,
-                this.dailyClient.userSpeaking$,
-            ]).subscribe(([bot, user]) => {
-                const current = this.callStateSubject.value;
-                if (current === 'connecting' && !bot) {
-                    return;
-                }
-                if (current === 'connecting' && bot) {
+            if (this.callStateSubject.value !== 'connecting') {
+                return;
+            }
+            try {
+                yield this.startLocalMic();
+                this.statusTextSubject.next('Connected');
+                this.callStateSubject.next('connected');
+                this.wireSpeakingState();
+            }
+            catch (err) {
+                console.error('[HiveGpt Voice] Mic or session setup failed', err);
+                this.callStateSubject.next('ended');
+                this.statusTextSubject.next('Microphone unavailable');
+                yield this.disconnect();
+            }
+        });
+    }
+    wireSpeakingState() {
+        const untilCallEnds$ = merge(this.destroy$, this.endCall$);
+        const transcriptDrivenAssistant$ = this.wsClient.botTranscript$.pipe(switchMap(() => concat(of(true), timer(800).pipe(map(() => false)))), distinctUntilChanged());
+        const assistantTalking$ = merge(this.wsClient.assistantSpeaking$, transcriptDrivenAssistant$).pipe(distinctUntilChanged(), startWith(false));
+        const userTalking$ = combineLatest([
+            this.audioAnalyzer.isUserSpeaking$,
+            this.wsClient.serverUserSpeaking$.pipe(startWith(false)),
+        ]).pipe(map(([local, server]) => local || server), distinctUntilChanged(), startWith(false));
+        this.subscriptions.add(combineLatest([assistantTalking$, userTalking$])
+            .pipe(takeUntil(untilCallEnds$))
+            .subscribe(([bot, user]) => {
+            const current = this.callStateSubject.value;
+            if (user) {
+                this.isUserSpeakingSubject.next(true);
+                this.callStateSubject.next('listening');
+            }
+            else {
+                this.isUserSpeakingSubject.next(false);
+            }
+            if (user) {
+                return;
+            }
+            if (bot) {
+                if (this.callStartTime === 0) {
                     this.callStartTime = Date.now();
                     this.startDurationTimer();
-                    this.callStateSubject.next('talking');
-                    return;
-                }
-                if (user) {
-                    this.callStateSubject.next('listening');
                 }
-                else if (bot) {
-                    this.callStateSubject.next('talking');
+                this.callStateSubject.next('talking');
+            }
+            else if (current === 'talking' || current === 'listening') {
+                this.callStateSubject.next('connected');
+            }
+        }));
+    }
+    startLocalMic() {
+        return __awaiter(this, void 0, void 0, function* () {
+            this.stopLocalMic();
+            const stream = yield navigator.mediaDevices.getUserMedia({ audio: true });
+            const track = stream.getAudioTracks()[0];
+            if (!track) {
+                stream.getTracks().forEach((t) => t.stop());
+                throw new Error('No audio track');
+            }
+            this.localMicStream = stream;
+            this.isMicMutedSubject.next(!track.enabled);
+            this.audioAnalyzer.start(stream);
+        });
+    }
+    stopLocalMic() {
+        if (this.localMicStream) {
+            this.localMicStream.getTracks().forEach((t) => t.stop());
+            this.localMicStream = null;
+        }
+    }
+    enqueueRemoteAudio(chunk) {
+        this.pendingRemoteAudio.push(chunk.slice(0));
+        if (!this.remoteAudioPlaying) {
+            void this.playRemoteAudioQueue();
+        }
+    }
+    playRemoteAudioQueue() {
+        return __awaiter(this, void 0, void 0, function* () {
+            this.remoteAudioPlaying = true;
+            const context = this.getOrCreateRemoteAudioContext();
+            while (this.pendingRemoteAudio.length > 0) {
+                const chunk = this.pendingRemoteAudio.shift();
+                if (!chunk)
+                    continue;
+                try {
+                    const decoded = yield this.decodeAudioChunk(context, chunk);
+                    this.assistantAudioStarted();
+                    yield this.playDecodedBuffer(context, decoded);
                 }
-                else if (current === 'talking' || current === 'listening') {
-                    this.callStateSubject.next('connected');
+                catch (_a) {
+                    // Ignore undecodable chunks; server may mix non-audio binary events.
                 }
-            }));
-            this.subscriptions.add(this.dailyClient.micMuted$.subscribe((muted) => this.isMicMutedSubject.next(muted)));
-            this.statusTextSubject.next('Connecting...');
+            }
+            this.remoteAudioPlaying = false;
+            this.assistantAudioStopped();
+        });
+    }
+    getOrCreateRemoteAudioContext() {
+        if (!this.remoteAudioContext || this.remoteAudioContext.state === 'closed') {
+            this.remoteAudioContext = new AudioContext();
+        }
+        if (this.remoteAudioContext.state === 'suspended') {
+            void this.remoteAudioContext.resume();
+        }
+        return this.remoteAudioContext;
+    }
+    decodeAudioChunk(context, chunk) {
+        return new Promise((resolve, reject) => {
+            context.decodeAudioData(chunk.slice(0), resolve, reject);
+        });
+    }
+    playDecodedBuffer(context, buffer) {
+        return new Promise((resolve) => {
+            const source = context.createBufferSource();
+            source.buffer = buffer;
+            source.connect(context.destination);
+            source.onended = () => resolve();
+            source.start();
+        });
+    }
+    assistantAudioStarted() {
+        if (this.callStartTime === 0) {
+            this.callStartTime = Date.now();
+            this.startDurationTimer();
+        }
+        this.callStateSubject.next('talking');
+    }
+    assistantAudioStopped() {
+        if (this.callStateSubject.value === 'talking') {
+            this.callStateSubject.next('connected');
+        }
+    }
+    resetRemoteAudioPlayback() {
+        this.pendingRemoteAudio = [];
+        this.remoteAudioPlaying = false;
+        if (this.remoteAudioContext && this.remoteAudioContext.state !== 'closed') {
+            this.remoteAudioContext.close().catch(() => { });
+        }
+        this.remoteAudioContext = null;
+    }
+    handleRemoteClose() {
+        return __awaiter(this, void 0, void 0, function* () {
+            const state = this.callStateSubject.value;
+            if (state === 'idle' || state === 'ended')
+                return;
+            this.endCall$.next();
+            this.stopDurationTimer();
+            this.callStartTime = 0;
+            this.audioAnalyzer.stop();
+            this.stopLocalMic();
+            this.resetRemoteAudioPlayback();
+            this.callStateSubject.next('ended');
+            this.statusTextSubject.next('Connection lost');
         });
     }
     disconnect() {
         return __awaiter(this, void 0, void 0, function* () {
+            this.endCall$.next();
             this.stopDurationTimer();
+            this.callStartTime = 0;
             this.audioAnalyzer.stop();
-            // Daily first, then WebSocket
-            yield this.dailyClient.disconnect();
+            this.stopLocalMic();
+            this.resetRemoteAudioPlayback();
             this.wsClient.disconnect();
             this.callStateSubject.next('ended');
             this.statusTextSubject.next('Call Ended');
         });
     }
     toggleMic() {
-        const current = this.isMicMutedSubject.value;
-        this.dailyClient.setMuted(!current);
+        var _a;
+        const nextMuted = !this.isMicMutedSubject.value;
+        const track = (_a = this.localMicStream) === null || _a === void 0 ? void 0 : _a.getAudioTracks()[0];
+        if (track) {
+            track.enabled = !nextMuted;
+        }
+        this.isMicMutedSubject.next(nextMuted);
     }
     startDurationTimer() {
         const updateDuration = () => {
@@ -1423,7 +1347,7 @@ class VoiceAgentService {
         }
     }
 }
-VoiceAgentService.ɵprov = i0.ɵɵdefineInjectable({ factory: function VoiceAgentService_Factory() { return new VoiceAgentService(i0.ɵɵinject(AudioAnalyzerService), i0.ɵɵinject(WebSocketVoiceClientService), i0.ɵɵinject(DailyVoiceClientService), i0.ɵɵinject(PlatformTokenRefreshService), i0.ɵɵinject(i0.PLATFORM_ID)); }, token: VoiceAgentService, providedIn: "root" });
+VoiceAgentService.ɵprov = i0.ɵɵdefineInjectable({ factory: function VoiceAgentService_Factory() { return new VoiceAgentService(i0.ɵɵinject(AudioAnalyzerService), i0.ɵɵinject(WebSocketVoiceClientService), i0.ɵɵinject(PlatformTokenRefreshService), i0.ɵɵinject(i0.PLATFORM_ID)); }, token: VoiceAgentService, providedIn: "root" });
 VoiceAgentService.decorators = [
     { type: Injectable, args: [{
                 providedIn: 'root',
@@ -1432,7 +1356,6 @@ VoiceAgentService.decorators = [
 VoiceAgentService.ctorParameters = () => [
     { type: AudioAnalyzerService },
     { type: WebSocketVoiceClientService },
-    { type: DailyVoiceClientService },
     { type: PlatformTokenRefreshService },
     { type: Object, decorators: [{ type: Inject, args: [PLATFORM_ID,] }] }
 ];
@@ -5472,7 +5395,7 @@ ChatBotComponent.propDecorators = {
 };
 
 /**
- * Voice agent module. Uses native WebSocket + Daily.js only.
+ * Voice agent module. Uses native WebSocket for the voice session.
  * Does NOT use Socket.IO or ngx-socket-io.
  */
 class VoiceAgentModule {
@@ -5488,8 +5411,7 @@ VoiceAgentModule.decorators = [
                 providers: [
                     VoiceAgentService,
                     AudioAnalyzerService,
-                    WebSocketVoiceClientService,
-                    DailyVoiceClientService
+                    WebSocketVoiceClientService
                 ],
                 exports: [
                     VoiceAgentModalComponent
@@ -5760,5 +5682,5 @@ HiveGptModule.decorators = [
  * Generated bundle index. Do not edit.
  */
 
-export { AudioAnalyzerService, ChatBotComponent, ChatDrawerComponent, HIVEGPT_AUTH_STORAGE_KEY, HiveGptModule, PlatformTokenRefreshService, VOICE_MODAL_CLOSE_CALLBACK, VOICE_MODAL_CONFIG, VoiceAgentModalComponent, VoiceAgentModule, VoiceAgentService, eClassificationType, hiveGptAuthStorageKeyFactory, BotsService as ɵa, SocketService as ɵb, ConversationService as ɵc, NotificationSocket as ɵd, TranslationService as ɵe, WebSocketVoiceClientService as ɵf, DailyVoiceClientService as ɵg, VideoPlayerComponent as ɵh, SafeHtmlPipe as ɵi, BotHtmlEditorComponent as ɵj };
+export { AudioAnalyzerService, ChatBotComponent, ChatDrawerComponent, HIVEGPT_AUTH_STORAGE_KEY, HiveGptModule, PlatformTokenRefreshService, VOICE_MODAL_CLOSE_CALLBACK, VOICE_MODAL_CONFIG, VoiceAgentModalComponent, VoiceAgentModule, VoiceAgentService, eClassificationType, hiveGptAuthStorageKeyFactory, BotsService as ɵa, SocketService as ɵb, ConversationService as ɵc, NotificationSocket as ɵd, TranslationService as ɵe, WebSocketVoiceClientService as ɵf, VideoPlayerComponent as ɵg, SafeHtmlPipe as ɵh, BotHtmlEditorComponent as ɵi };
 //# sourceMappingURL=hivegpt-hiveai-angular.js.map