@glydeunity/voice-sdk 1.6.28 → 1.6.32

package/dist/index.d.ts

@@ -736,6 +736,7 @@ export declare class GlydeText extends BaseGlydeClient<GlydeTextConfig, TextChat
     private contextSwitchManager;
     private currentContextType;
     private agentMetadata;
+    private initPromise;
     /**
      * Create a new GlydeText instance
      * @param config - Configuration options
@@ -758,6 +759,10 @@ export declare class GlydeText extends BaseGlydeClient<GlydeTextConfig, TextChat
      * @returns Array with the greeting message
      */
     initialize(): Promise<ChatMessage[]>;
+    /**
+     * Internal initialization logic (called exactly once)
+     */
+    private doInitialize;
     /**
      * Get a context-appropriate greeting for non-screening contexts
      * @returns Greeting message based on context type
@@ -904,6 +909,7 @@ export declare class GlydeVoice extends BaseGlydeClient<GlydeVoiceConfig, VoiceE
     private microphoneStarted;
     private readonly outputSampleRate;
     private readonly inputSampleRate;
+    private readonly playbackSpeed;
     private isAgentSpeaking;
     private agentAudioDoneReceived;
     private contextSwitchManager;
@@ -962,9 +968,12 @@ export declare class GlydeVoice extends BaseGlydeClient<GlydeVoiceConfig, VoiceE
      */
     private handleAudioBuffer;
     /**
-     * Resample audio from 24kHz to 48kHz using linear interpolation
+     * Resample audio from 24kHz to the AudioContext rate (48kHz), adjusted for
+     * playback speed. At speed 1.0 the ratio is 2x (standard 24k->48k upsample).
+     * Higher speeds produce fewer output samples per input sample, causing the
+     * worklet to drain the buffer faster and the speech to sound faster.
      */
-    private resample24kTo48k;
+    private resampleToPlaybackRate;
     /**
      * Clear the playback buffer (for interruption handling)
      */
@@ -1119,6 +1128,14 @@ export declare interface GlydeVoiceConfig extends BaseGlydeConfig {
     limitToLast?: number;
     /** Override Deepgram configuration (e.g. model, tags). System prompt and greeting remain server-controlled. */
     deepgramConfig?: DeepgramAgentConfig;
+    /**
+     * Playback speed multiplier for agent speech audio.
+     * 1.0 = normal speed, 1.2 = 20% faster, 0.8 = 20% slower.
+     * Adjusts the resampling ratio rather than using Web Audio playbackRate,
+     * so it introduces slight pitch shift proportional to the speed change.
+     * Recommended range: 0.8 - 1.5. Default: 1.2
+     */
+    playbackSpeed?: number;
     /**
      * Additional tool/function definitions to merge with server-provided functions.
      * Server-defined functions are always included; this allows adding extra tools only.