@layercode/js-sdk 2.8.1 → 2.8.3

package/dist/types/index.d.ts

@@ -84,6 +84,13 @@ interface LayercodeClientOptions {
     audioInput?: boolean;
     /** Whether audio output is enabled. I.e. do we play the sound in the browser client */
     audioOutput?: boolean;
+    /**
+     * When true, defers actual audio hardware initialization (AudioContext, mic permissions)
+     * until setAudioInput(true) or setAudioOutput(true) is called.
+     * The server will still be told this is a voice session, but no browser audio APIs are touched until needed.
+     * This avoids Chrome's autoplay policy blocking AudioContext before user gesture.
+     */
+    deferAudioInit?: boolean;
     /** Fired when audio input flag changes */
     audioInputChanged?: (audioInput: boolean) => void;
     /** Fired when audio output flag changes */
@@ -143,9 +150,11 @@ declare class LayercodeClient implements ILayercodeClient {
     private canInterrupt;
     private userIsSpeaking;
     private agentIsSpeaking;
+    private agentIsPlayingAudio;
     private recorderStarted;
     private readySent;
     private currentTurnId;
+    private sentReplayFinishedForDisabledOutput;
     private audioBuffer;
     private vadConfig;
     private deviceId;
@@ -220,6 +229,11 @@ declare class LayercodeClient implements ILayercodeClient {
     audioInputDisconnect(): Promise<void>;
     setAudioInput(state: boolean): Promise<void>;
     setAudioOutput(state: boolean): Promise<void>;
+    /**
+     * Syncs the reported agentSpeaking state with the actual audio playback state.
+     * Called when audioOutput is enabled to ensure proper state synchronization.
+     */
+    private _syncAgentSpeakingState;
     /** Emitters for audio flags */
     private _emitAudioInput;
     private _emitAudioOutput;

package/dist/types/interfaces.d.ts

@@ -1,4 +1,4 @@
-export type LayercodeMessageType = 'client.audio' | 'trigger.turn.start' | 'trigger.turn.end' | 'trigger.response.audio.replay_finished' | 'vad_events' | 'client.ready' | 'client.response.text' | 'client.response.data' | 'turn.start' | 'response.audio' | 'response.text' | 'response.data' | 'user.transcript.interim_delta' | 'user.transcript.delta' | 'user.transcript';
+export type LayercodeMessageType = 'client.audio' | 'trigger.turn.start' | 'trigger.turn.end' | 'trigger.response.audio.replay_finished' | 'vad_events' | 'client.ready' | 'client.response.text' | 'client.response.data' | 'turn.start' | 'response.audio' | 'response.text' | 'response.data' | 'response.end' | 'user.transcript.interim_delta' | 'user.transcript.delta' | 'user.transcript';
 export interface BaseLayercodeMessage {
     type: LayercodeMessageType;
     event_id?: string;
@@ -68,6 +68,10 @@ export interface ServerResponseUserTranscript extends BaseLayercodeMessage {
     content: string;
     turn_id: string;
 }
-export type ServerMessage = ServerTurnMessage | ServerResponseAudioMessage | ServerResponseTextMessage | ServerResponseDataMessage | ServerResponseUserTranscriptInterimDelta | ServerResponseUserTranscriptDelta | ServerResponseUserTranscript;
+export interface ServerResponseEndMessage extends BaseLayercodeMessage {
+    type: 'response.end';
+    turn_id: string;
+}
+export type ServerMessage = ServerTurnMessage | ServerResponseAudioMessage | ServerResponseTextMessage | ServerResponseDataMessage | ServerResponseEndMessage | ServerResponseUserTranscriptInterimDelta | ServerResponseUserTranscriptDelta | ServerResponseUserTranscript;
 export type ClientMessage = ClientAudioMessage | ClientTriggerTurnMessage | ClientTriggerResponseAudioReplayFinishedMessage | ClientVadEventsMessage | ClientReadyMessage | ClientResponseTextMessage | ClientResponseDataMessage;
 export type LayercodeMessage = ClientMessage | ServerMessage;

package/dist/types/wavtools/lib/analysis/audio_analysis.d.ts

@@ -31,7 +31,7 @@ export class AudioAnalysis {
     constructor(audioElement: HTMLAudioElement, audioBuffer?: AudioBuffer | null);
     fftResults: any[];
     audio: HTMLAudioElement;
-    context: OfflineAudioContext | AudioContext;
+    context: AudioContext | OfflineAudioContext;
     analyser: AnalyserNode;
     sampleRate: number;
     audioBuffer: AudioBuffer | null;

package/package.json

@@ -2,7 +2,7 @@
   "author": "Layercode",
   "license": "MIT",
   "name": "@layercode/js-sdk",
-  "version": "2.8.1",
+  "version": "2.8.3",
   "description": "Layercode JavaScript SDK for browser usage",
   "type": "module",
   "main": "dist/layercode-js-sdk.esm.js",