@hivegpt/hiveai-angular 0.0.582 → 0.0.583

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

@@ -1357,6 +1357,7 @@
             this.botTranscriptSubject = new rxjs.Subject();
             this.assistantSpeakingSubject = new rxjs.Subject();
             this.serverUserSpeakingSubject = new rxjs.Subject();
+            this.audioChunkSubject = new rxjs.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}. */
@@ -1367,6 +1368,8 @@
             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();
         }
         WebSocketVoiceClientService.prototype.connect = function (wsUrl) {
             var _this = this;
@@ -1389,16 +1392,7 @@
                 socket_1.onmessage = function (event) {
                     if (_this.ws !== socket_1)
                         return;
-                    if (typeof event.data !== 'string') {
-                        return;
-                    }
-                    try {
-                        var msg_1 = JSON.parse(event.data);
-                        _this.ngZone.run(function () { return _this.handleJsonMessage(msg_1); });
-                    }
-                    catch (_a) {
-                        // Ignore non-JSON
-                    }
+                    void _this.handleIncomingMessage(event.data);
                 };
                 socket_1.onerror = function () {
                     _this.ngZone.run(function () {
@@ -1424,6 +1418,65 @@
                 throw err;
             }
         };
+        WebSocketVoiceClientService.prototype.handleIncomingMessage = function (payload) {
+            return __awaiter(this, void 0, void 0, function () {
+                var ab;
+                return __generator(this, function (_b) {
+                    switch (_b.label) {
+                        case 0:
+                            if (typeof payload === 'string') {
+                                this.handleJsonString(payload);
+                                return [2 /*return*/];
+                            }
+                            if (payload instanceof ArrayBuffer) {
+                                this.handleBinaryMessage(payload);
+                                return [2 /*return*/];
+                            }
+                            if (!(payload instanceof Blob)) return [3 /*break*/, 2];
+                            return [4 /*yield*/, payload.arrayBuffer()];
+                        case 1:
+                            ab = _b.sent();
+                            this.handleBinaryMessage(ab);
+                            _b.label = 2;
+                        case 2: return [2 /*return*/];
+                    }
+                });
+            });
+        };
+        WebSocketVoiceClientService.prototype.handleJsonString = function (jsonText) {
+            var _this = this;
+            try {
+                var msg_1 = JSON.parse(jsonText);
+                this.ngZone.run(function () { return _this.handleJsonMessage(msg_1); });
+            }
+            catch (_a) {
+                // Ignore non-JSON
+            }
+        };
+        WebSocketVoiceClientService.prototype.handleBinaryMessage = function (buffer) {
+            var _this = this;
+            // Some backends wrap JSON events inside binary WS frames.
+            var maybeText = this.tryDecodeUtf8(buffer);
+            if (maybeText !== null) {
+                this.handleJsonString(maybeText);
+                return;
+            }
+            // Otherwise treat binary as streamed assistant audio.
+            this.ngZone.run(function () { return _this.audioChunkSubject.next(buffer); });
+        };
+        WebSocketVoiceClientService.prototype.tryDecodeUtf8 = function (buffer) {
+            try {
+                var text = new TextDecoder('utf-8', { fatal: true }).decode(buffer);
+                var trimmed = text.trim();
+                if (!trimmed || (trimmed[0] !== '{' && trimmed[0] !== '[')) {
+                    return null;
+                }
+                return trimmed;
+            }
+            catch (_a) {
+                return null;
+            }
+        };
         WebSocketVoiceClientService.prototype.handleJsonMessage = function (msg) {
             var type = msg.type;
             var typeStr = typeof type === 'string' ? type : '';
@@ -1529,6 +1582,9 @@
             this.callStartTime = 0;
             this.durationInterval = null;
             this.localMicStream = null;
+            this.remoteAudioContext = null;
+            this.pendingRemoteAudio = [];
+            this.remoteAudioPlaying = false;
             this.endCall$ = new rxjs.Subject();
             this.subscriptions = new rxjs.Subscription();
             this.destroy$ = new rxjs.Subject();
@@ -1544,6 +1600,9 @@
             this.subscriptions.add(this.wsClient.remoteClose$
                 .pipe(operators.takeUntil(this.destroy$))
                 .subscribe(function () { return void _this.handleRemoteClose(); }));
+            this.subscriptions.add(this.wsClient.audioChunk$
+                .pipe(operators.takeUntil(this.destroy$))
+                .subscribe(function (chunk) { return _this.enqueueRemoteAudio(chunk); }));
         }
         VoiceAgentService.prototype.ngOnDestroy = function () {
             this.destroy$.next();
@@ -1559,6 +1618,7 @@
             this.callStartTime = 0;
             this.audioAnalyzer.stop();
             this.stopLocalMic();
+            this.resetRemoteAudioPlayback();
             this.wsClient.disconnect();
             this.callStateSubject.next('idle');
             this.statusTextSubject.next('');
@@ -1762,6 +1822,92 @@
                 this.localMicStream = null;
             }
         };
+        VoiceAgentService.prototype.enqueueRemoteAudio = function (chunk) {
+            this.pendingRemoteAudio.push(chunk.slice(0));
+            if (!this.remoteAudioPlaying) {
+                void this.playRemoteAudioQueue();
+            }
+        };
+        VoiceAgentService.prototype.playRemoteAudioQueue = function () {
+            return __awaiter(this, void 0, void 0, function () {
+                var context, chunk, decoded, _a_1;
+                return __generator(this, function (_b) {
+                    switch (_b.label) {
+                        case 0:
+                            this.remoteAudioPlaying = true;
+                            context = this.getOrCreateRemoteAudioContext();
+                            _b.label = 1;
+                        case 1:
+                            if (!(this.pendingRemoteAudio.length > 0)) return [3 /*break*/, 7];
+                            chunk = this.pendingRemoteAudio.shift();
+                            if (!chunk)
+                                return [3 /*break*/, 1];
+                            _b.label = 2;
+                        case 2:
+                            _b.trys.push([2, 5, , 6]);
+                            return [4 /*yield*/, this.decodeAudioChunk(context, chunk)];
+                        case 3:
+                            decoded = _b.sent();
+                            this.assistantAudioStarted();
+                            return [4 /*yield*/, this.playDecodedBuffer(context, decoded)];
+                        case 4:
+                            _b.sent();
+                            return [3 /*break*/, 6];
+                        case 5:
+                            _a_1 = _b.sent();
+                            return [3 /*break*/, 6];
+                        case 6: return [3 /*break*/, 1];
+                        case 7:
+                            this.remoteAudioPlaying = false;
+                            this.assistantAudioStopped();
+                            return [2 /*return*/];
+                    }
+                });
+            });
+        };
+        VoiceAgentService.prototype.getOrCreateRemoteAudioContext = function () {
+            if (!this.remoteAudioContext || this.remoteAudioContext.state === 'closed') {
+                this.remoteAudioContext = new AudioContext();
+            }
+            if (this.remoteAudioContext.state === 'suspended') {
+                void this.remoteAudioContext.resume();
+            }
+            return this.remoteAudioContext;
+        };
+        VoiceAgentService.prototype.decodeAudioChunk = function (context, chunk) {
+            return new Promise(function (resolve, reject) {
+                context.decodeAudioData(chunk.slice(0), resolve, reject);
+            });
+        };
+        VoiceAgentService.prototype.playDecodedBuffer = function (context, buffer) {
+            return new Promise(function (resolve) {
+                var source = context.createBufferSource();
+                source.buffer = buffer;
+                source.connect(context.destination);
+                source.onended = function () { return resolve(); };
+                source.start();
+            });
+        };
+        VoiceAgentService.prototype.assistantAudioStarted = function () {
+            if (this.callStartTime === 0) {
+                this.callStartTime = Date.now();
+                this.startDurationTimer();
+            }
+            this.callStateSubject.next('talking');
+        };
+        VoiceAgentService.prototype.assistantAudioStopped = function () {
+            if (this.callStateSubject.value === 'talking') {
+                this.callStateSubject.next('connected');
+            }
+        };
+        VoiceAgentService.prototype.resetRemoteAudioPlayback = function () {
+            this.pendingRemoteAudio = [];
+            this.remoteAudioPlaying = false;
+            if (this.remoteAudioContext && this.remoteAudioContext.state !== 'closed') {
+                this.remoteAudioContext.close().catch(function () { });
+            }
+            this.remoteAudioContext = null;
+        };
         VoiceAgentService.prototype.handleRemoteClose = function () {
             return __awaiter(this, void 0, void 0, function () {
                 var state;
@@ -1774,6 +1920,7 @@
                     this.callStartTime = 0;
                     this.audioAnalyzer.stop();
                     this.stopLocalMic();
+                    this.resetRemoteAudioPlayback();
                     this.callStateSubject.next('ended');
                     this.statusTextSubject.next('Connection lost');
                     return [2 /*return*/];
@@ -1788,6 +1935,7 @@
                     this.callStartTime = 0;
                     this.audioAnalyzer.stop();
                     this.stopLocalMic();
+                    this.resetRemoteAudioPlayback();
                     this.wsClient.disconnect();
                     this.callStateSubject.next('ended');
                     this.statusTextSubject.next('Call Ended');