@estuary-ai/sdk 0.1.7 → 0.1.9

package/dist/index.mjs

@@ -3871,11 +3871,30 @@ var AudioPlayer = class {
   currentSource = null;
   currentMessageId = null;
   isPlaying = false;
+  _isCleared = false;
+  _interruptedMessageId = null;
   constructor(sampleRate, onEvent) {
     this.sampleRate = sampleRate;
     this.onEvent = onEvent;
   }
+  /** Whether audio is currently playing */
+  get playing() {
+    return this.isPlaying;
+  }
+  /** The messageId of the currently playing audio, or null */
+  get playingMessageId() {
+    return this.currentMessageId;
+  }
+  /** Mark a messageId as interrupted so late-arriving chunks are dropped */
+  setInterruptedMessageId(id) {
+    this._interruptedMessageId = id;
+  }
   enqueue(voice) {
+    if (voice.messageId === this._interruptedMessageId) return;
+    if (this._interruptedMessageId && voice.messageId !== this._interruptedMessageId) {
+      this._interruptedMessageId = null;
+    }
+    this._isCleared = false;
     const ctx = this.getAudioContext();
     if (!ctx) return;
     const pcm16 = base64ToInt16Array(voice.audio);
@@ -3888,6 +3907,7 @@ var AudioPlayer = class {
     }
   }
   clear() {
+    this._isCleared = true;
     this.queue.length = 0;
     if (this.currentSource) {
       try {
@@ -3897,6 +3917,7 @@ var AudioPlayer = class {
       }
       this.currentSource = null;
     }
+    if (this.audioElement) this.audioElement.muted = true;
     this.isPlaying = false;
     this.currentMessageId = null;
   }
@@ -3940,6 +3961,7 @@ var AudioPlayer = class {
     return ctx;
   }
   playNext() {
+    if (this._isCleared) return;
     const ctx = this.getAudioContext();
     if (!ctx || this.queue.length === 0) {
       if (this.isPlaying && this.currentMessageId) {
@@ -3963,9 +3985,11 @@ var AudioPlayer = class {
     source.connect(this.mediaStreamDest ?? ctx.destination);
     this.currentSource = source;
     source.onended = () => {
+      if (this._isCleared) return;
       this.currentSource = null;
       this.playNext();
     };
+    if (this.audioElement) this.audioElement.muted = false;
     ctx.resume().catch(() => {
     });
     source.start();
@@ -4071,6 +4095,7 @@ var EstuaryClient = class extends TypedEventEmitter {
   _memory;
   _sessionInfo = null;
   actionParsers = /* @__PURE__ */ new Map();
+  _hasAutoInterrupted = false;
   constructor(config) {
     super();
     this.config = config;
@@ -4121,6 +4146,7 @@ var EstuaryClient = class extends TypedEventEmitter {
   interrupt(messageId) {
     this.ensureConnected();
     this.socketManager.emitEvent("client_interrupt", { message_id: messageId });
+    this.audioPlayer?.setInterruptedMessageId(messageId ?? this.audioPlayer.playingMessageId);
     this.audioPlayer?.clear();
     if (this.config.suppressMicDuringPlayback) {
       this.voiceManager?.setSuppressed?.(false);
@@ -4162,6 +4188,7 @@ var EstuaryClient = class extends TypedEventEmitter {
     if (!this.audioPlayer && typeof AudioContext !== "undefined") {
       this.audioPlayer = new AudioPlayer(sampleRate, (event) => {
         if (event.type === "started") {
+          this._hasAutoInterrupted = false;
           this.emit("audioPlaybackStarted", event.messageId);
           if (this.config.suppressMicDuringPlayback) {
             this.voiceManager?.setSuppressed?.(true);
@@ -4222,10 +4249,15 @@ var EstuaryClient = class extends TypedEventEmitter {
     this.socketManager.on("connectionStateChanged", (state) => this.emit("connectionStateChanged", state));
     this.socketManager.on("botResponse", (response) => this.handleBotResponse(response));
     this.socketManager.on("botVoice", (voice) => this.handleBotVoice(voice));
-    this.socketManager.on("sttResponse", (response) => this.emit("sttResponse", response));
+    this.socketManager.on("sttResponse", (response) => {
+      this.maybeAutoInterrupt(response);
+      this.emit("sttResponse", response);
+    });
     this.socketManager.on("interrupt", (data) => {
+      this.audioPlayer?.setInterruptedMessageId(data.messageId ?? null);
       this.audioPlayer?.clear();
       this.actionParsers.clear();
+      this._hasAutoInterrupted = false;
       if (this.config.suppressMicDuringPlayback) {
         this.voiceManager?.setSuppressed?.(false);
       }
@@ -4265,6 +4297,15 @@ var EstuaryClient = class extends TypedEventEmitter {
     this.emit("botVoice", voice);
     this.audioPlayer?.enqueue(voice);
   }
+  maybeAutoInterrupt(stt) {
+    if ((this.config.autoInterruptOnSpeech ?? true) === false) return;
+    if (this.config.suppressMicDuringPlayback) return;
+    if (stt.isFinal) return;
+    if (!this.audioPlayer?.playing) return;
+    if (this._hasAutoInterrupted) return;
+    this._hasAutoInterrupted = true;
+    this.interrupt();
+  }
 };
 
 export { ConnectionState, EstuaryClient, parseActions };