@estuary-ai/sdk 0.1.7 → 0.1.8

package/dist/index.d.mts

@@ -23,6 +23,8 @@ interface EstuaryConfig {
     realtimeMemory?: boolean;
     /** Suppress mic during TTS playback (software AEC fallback, disables barge-in). Default: false */
     suppressMicDuringPlayback?: boolean;
+    /** Proactively interrupt bot audio when user starts speaking (default: true) */
+    autoInterruptOnSpeech?: boolean;
 }
 type VoiceTransport = 'websocket' | 'livekit' | 'auto';
 declare enum ConnectionState {
@@ -243,6 +245,7 @@ declare class EstuaryClient extends TypedEventEmitter<EstuaryEventMap> {
     private _memory;
     private _sessionInfo;
     private actionParsers;
+    private _hasAutoInterrupted;
     constructor(config: EstuaryConfig);
     /** Memory API client for querying memories, graphs, and facts */
     get memory(): MemoryClient;
@@ -282,6 +285,7 @@ declare class EstuaryClient extends TypedEventEmitter<EstuaryEventMap> {
     private forwardSocketEvents;
     private handleBotResponse;
     private handleBotVoice;
+    private maybeAutoInterrupt;
 }
 
 declare enum ErrorCode {

package/dist/index.js

@@ -9203,11 +9203,30 @@ 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);
@@ -9220,6 +9239,7 @@ var AudioPlayer = class {
     }
   }
   clear() {
+    this._isCleared = true;
     this.queue.length = 0;
     if (this.currentSource) {
       try {
@@ -9272,6 +9292,7 @@ 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) {
@@ -9295,6 +9316,7 @@ var AudioPlayer = class {
     source.connect(this.mediaStreamDest ?? ctx.destination);
     this.currentSource = source;
     source.onended = () => {
+      if (this._isCleared) return;
       this.currentSource = null;
       this.playNext();
     };
@@ -9406,6 +9428,7 @@ var EstuaryClient = class extends TypedEventEmitter {
   _memory;
   _sessionInfo = null;
   actionParsers = /* @__PURE__ */ new Map();
+  _hasAutoInterrupted = false;
   constructor(config) {
     super();
     this.config = config;
@@ -9456,6 +9479,7 @@ 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);
@@ -9497,6 +9521,7 @@ 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);
@@ -9557,10 +9582,15 @@ 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);
       }
@@ -9600,6 +9630,15 @@ 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();
+  }
 };
 
 // src/index.ts