@hivegpt/hiveai-angular 0.0.458 → 0.0.460

package/bundles/hivegpt-hiveai-angular.umd.js

@@ -1264,13 +1264,10 @@
      * - Create and manage Daily CallObject
      * - Join Daily room using room_url
      * - Handle mic capture + speaker playback
-     * - Provide real-time speaking detection via active-speaker-change (primary)
-     *   and track-started/track-stopped (fallback for immediate feedback)
+     * - 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)
-     *
-     * Speaking state flips immediately when agent audio starts playing.
-     * If user speaks while bot is talking, state switches to listening.
      */
     var DailyVoiceClientService = /** @class */ (function () {
         function DailyVoiceClientService(ngZone) {
@@ -1280,6 +1277,9 @@
             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 rxjs.BehaviorSubject(false);
             this.userSpeakingSubject = new rxjs.BehaviorSubject(false);
             this.micMutedSubject = new rxjs.BehaviorSubject(false);
@@ -1353,24 +1353,21 @@
         };
         DailyVoiceClientService.prototype.setupEventHandlers = function (call) {
             var _this = this;
-            // active-speaker-change: primary source for real-time speaking detection.
-            // Emits when the loudest participant changes; bot speaking = remote is active.
+            // 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', function (event) {
                 _this.ngZone.run(function () {
                     var _a;
                     var peerId = (_a = event === null || event === void 0 ? void 0 : event.activeSpeaker) === null || _a === void 0 ? void 0 : _a.peerId;
                     if (!peerId || !_this.localSessionId) {
-                        _this.speakingSubject.next(false);
                         _this.userSpeakingSubject.next(false);
                         return;
                     }
                     var isLocal = peerId === _this.localSessionId;
                     _this.userSpeakingSubject.next(isLocal);
-                    _this.speakingSubject.next(!isLocal);
                 });
             });
-            // track-started / track-stopped: set up remote audio playback only.
-            // Speaking state is driven by active-speaker-change (actual audio energy), NOT track lifecycle.
+            // track-started / track-stopped: set up remote audio playback + AnalyserNode monitor.
             call.on('track-started', function (event) {
                 _this.ngZone.run(function () {
                     var _a, _b, _c, _d;
@@ -1382,6 +1379,7 @@
                         var 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);
                         }
                     }
                 });
@@ -1392,6 +1390,7 @@
                     var p = event === null || event === void 0 ? void 0 : event.participant;
                     var 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();
                     }
                 });
@@ -1446,6 +1445,69 @@
                 console.warn('DailyVoiceClient: failed to create remote audio element', err);
             }
         };
+        /**
+         * Monitor remote audio track energy via AnalyserNode.
+         * Polls at ~60fps and flips speakingSubject based on actual audio energy.
+         */
+        DailyVoiceClientService.prototype.monitorRemoteAudio = function (track) {
+            var _this = this;
+            this.stopRemoteAudioMonitor();
+            try {
+                var ctx = new AudioContext();
+                var source = ctx.createMediaStreamSource(new MediaStream([track]));
+                var analyser_1 = ctx.createAnalyser();
+                analyser_1.fftSize = 256;
+                source.connect(analyser_1);
+                this.remoteAudioContext = ctx;
+                var dataArray_1 = new Uint8Array(analyser_1.frequencyBinCount);
+                var THRESHOLD_1 = 5;
+                var SILENCE_MS_1 = 1500;
+                var lastSoundTime_1 = 0;
+                var isSpeaking_1 = false;
+                var poll_1 = function () {
+                    if (!_this.remoteAudioContext)
+                        return;
+                    analyser_1.getByteFrequencyData(dataArray_1);
+                    var sum = 0;
+                    for (var i = 0; i < dataArray_1.length; i++) {
+                        sum += dataArray_1[i];
+                    }
+                    var avg = sum / dataArray_1.length;
+                    var now = Date.now();
+                    if (avg > THRESHOLD_1) {
+                        lastSoundTime_1 = now;
+                        if (!isSpeaking_1) {
+                            isSpeaking_1 = true;
+                            console.log("[VoiceDebug] Bot audio energy detected (speaking=true) \u2014 avg=" + avg.toFixed(1) + " \u2014 " + new Date().toISOString());
+                            _this.ngZone.run(function () {
+                                _this.userSpeakingSubject.next(false);
+                                _this.speakingSubject.next(true);
+                            });
+                        }
+                    }
+                    else if (isSpeaking_1 && now - lastSoundTime_1 > SILENCE_MS_1) {
+                        isSpeaking_1 = false;
+                        console.log("[VoiceDebug] Bot audio silence detected (speaking=false) \u2014 " + new Date().toISOString());
+                        _this.ngZone.run(function () { return _this.speakingSubject.next(false); });
+                    }
+                    _this.remoteSpeakingRAF = requestAnimationFrame(poll_1);
+                };
+                this.remoteSpeakingRAF = requestAnimationFrame(poll_1);
+            }
+            catch (err) {
+                console.warn('DailyVoiceClient: failed to create remote audio monitor', err);
+            }
+        };
+        DailyVoiceClientService.prototype.stopRemoteAudioMonitor = function () {
+            if (this.remoteSpeakingRAF) {
+                cancelAnimationFrame(this.remoteSpeakingRAF);
+                this.remoteSpeakingRAF = null;
+            }
+            if (this.remoteAudioContext) {
+                this.remoteAudioContext.close().catch(function () { });
+                this.remoteAudioContext = null;
+            }
+        };
         DailyVoiceClientService.prototype.stopRemoteAudio = function () {
             if (this.remoteAudioElement) {
                 try {
@@ -1492,6 +1554,7 @@
             });
         };
         DailyVoiceClientService.prototype.cleanup = function () {
+            this.stopRemoteAudioMonitor();
             this.stopRemoteAudio();
             if (this.callObject) {
                 this.callObject.destroy().catch(function () { });