@lokutor/sdk 1.1.9 → 1.1.10

package/dist/index.d.mts

@@ -119,13 +119,11 @@ declare class VoiceAgentClient {
     private messages;
     private visemeListeners;
     private wantVisemes;
-    private serverUrl;
     constructor(config: LokutorConfig & {
         prompt: string;
         voice?: VoiceStyle;
         language?: Language;
         visemes?: boolean;
-        serverUrl?: string;
         onVisemes?: (visemes: Viseme[]) => void;
     });
     /**
@@ -328,7 +326,7 @@ declare class BrowserAudioManager {
     private mediaStream;
     private nextPlaybackTime;
     private activeSources;
-    private audioClockOffset;
+    private playbackQueue;
     private inputSampleRate;
     private outputSampleRate;
     private autoGainControl;
@@ -338,7 +336,6 @@ declare class BrowserAudioManager {
     private onInputError?;
     private isMuted;
     private isListening;
-    private resampler;
     constructor(config?: BrowserAudioConfig);
     /**
      * Initialize the AudioContext and analyser
@@ -348,77 +345,60 @@ declare class BrowserAudioManager {
      * Start capturing audio from the microphone
      */
     startMicrophone(onAudioInput: (pcm16Data: Uint8Array) => void): Promise<void>;
+    /**
+     * Internal method to process microphone audio data
+     */
     private _processAudioInput;
+    /**
+     * Stop capturing microphone input
+     */
     stopMicrophone(): void;
     /**
      * Play back audio received from the server
+     * @param pcm16Data Int16 PCM audio data at SPEAKER_SAMPLE_RATE
      */
     playAudio(pcm16Data: Uint8Array): void;
-    private _schedulePlayback;
     /**
-     * Get the current high-precision audio clock offset for viseme synchronization.
-     * Total stream time (in ms) = (audioContext.currentTime - audioClockOffset) * 1000
+     * Internal method to schedule and play audio with sample-accurate timing
      */
-    getAudioClockOffset(): number | null;
+    private _schedulePlayback;
     /**
-     * Reset the audio clock offset (call when a response is interrupted or finished)
+     * Stop all currently playing audio and clear the queue
      */
-    resetAudioClock(): void;
     stopPlayback(): void;
+    /**
+     * Toggle mute state
+     */
     setMuted(muted: boolean): void;
-    isMicMuted(): boolean;
-    getAmplitude(): number;
-    getFrequencyData(): Uint8Array;
-    getWaveformData(): Uint8Array;
-    cleanup(): void;
-    getAudioContext(): AudioContext | null;
-}
-
-/**
- * High-level AI Voice Agent for browser-based conversations.
- *
- * This class orchestrates microphone input, AI processing, and
- * speaker output, providing a simple interface for building
- * voice assistants with lip-sync support.
- */
-declare class VoiceAgent {
-    private client;
-    private audioManager;
-    private options;
-    private isConnected;
-    private visemeQueue;
-    constructor(options: VoiceAgentOptions & {
-        apiKey: string;
-    });
     /**
-     * Initialize hardware and connect to the AI server.
-     * This must be called in response to a user guesture (like a click)
-     * to satisfy browser AudioContext requirements.
+     * Get current mute state
      */
-    connect(): Promise<boolean>;
+    isMicMuted(): boolean;
     /**
-     * Get the current amplitude/volume of the microphone or output audio.
-     * Useful for voice activity visualization.
-     * @returns value between 0 and 1
+     * Get current amplitude from analyser (for visualization)
+     * Returns value between 0 and 1
      */
     getAmplitude(): number;
     /**
-     * Mute or unmute the microphone.
+     * Get frequency data from analyser for visualization
      */
-    toggleMute(): boolean;
+    getFrequencyData(): Uint8Array;
     /**
-     * High-precision method to get visemes that should be active
-     * at the current playback frame. Use this in a requestAnimationFrame loop.
+     * Get time-domain data from analyser for waveform visualization
      */
-    getFrameVisemes(): Viseme[];
+    getWaveformData(): Uint8Array;
     /**
-     * Change the system prompt mid-conversation.
+     * Cleanup and close AudioContext
      */
-    updatePrompt(newPrompt: string): void;
+    cleanup(): void;
     /**
-     * Disconnect and release audio resources.
+     * Get current audio context state
      */
-    disconnect(): void;
+    getState(): 'running' | 'suspended' | 'closed' | 'interrupted' | null;
+    /**
+     * Check if microphone is currently listening
+     */
+    isRecording(): boolean;
 }
 
-export { AUDIO_CONFIG, type AnalyserConfig, type BrowserAudioConfig, BrowserAudioManager, type BrowserAudioOptions, DEFAULT_URLS, Language, type LokutorConfig, StreamResampler, type SynthesizeOptions, TTSClient, type Viseme, VoiceAgent, VoiceAgentClient, type VoiceAgentOptions, VoiceStyle, applyLowPassFilter, bytesToPcm16, calculateRMS, float32ToPcm16, normalizeAudio, pcm16ToBytes, pcm16ToFloat32, resample, resampleWithAntiAliasing, simpleConversation, simpleTTS };
+export { AUDIO_CONFIG, type AnalyserConfig, type BrowserAudioConfig, BrowserAudioManager, type BrowserAudioOptions, DEFAULT_URLS, Language, type LokutorConfig, StreamResampler, type SynthesizeOptions, TTSClient, type Viseme, VoiceAgentClient, type VoiceAgentOptions, VoiceStyle, applyLowPassFilter, bytesToPcm16, calculateRMS, float32ToPcm16, normalizeAudio, pcm16ToBytes, pcm16ToFloat32, resample, resampleWithAntiAliasing, simpleConversation, simpleTTS };

package/dist/index.d.ts

@@ -119,13 +119,11 @@ declare class VoiceAgentClient {
     private messages;
     private visemeListeners;
     private wantVisemes;
-    private serverUrl;
     constructor(config: LokutorConfig & {
         prompt: string;
         voice?: VoiceStyle;
         language?: Language;
         visemes?: boolean;
-        serverUrl?: string;
         onVisemes?: (visemes: Viseme[]) => void;
     });
     /**
@@ -328,7 +326,7 @@ declare class BrowserAudioManager {
     private mediaStream;
     private nextPlaybackTime;
     private activeSources;
-    private audioClockOffset;
+    private playbackQueue;
     private inputSampleRate;
     private outputSampleRate;
     private autoGainControl;
@@ -338,7 +336,6 @@ declare class BrowserAudioManager {
     private onInputError?;
     private isMuted;
     private isListening;
-    private resampler;
     constructor(config?: BrowserAudioConfig);
     /**
      * Initialize the AudioContext and analyser
@@ -348,77 +345,60 @@ declare class BrowserAudioManager {
      * Start capturing audio from the microphone
      */
     startMicrophone(onAudioInput: (pcm16Data: Uint8Array) => void): Promise<void>;
+    /**
+     * Internal method to process microphone audio data
+     */
     private _processAudioInput;
+    /**
+     * Stop capturing microphone input
+     */
     stopMicrophone(): void;
     /**
      * Play back audio received from the server
+     * @param pcm16Data Int16 PCM audio data at SPEAKER_SAMPLE_RATE
      */
     playAudio(pcm16Data: Uint8Array): void;
-    private _schedulePlayback;
     /**
-     * Get the current high-precision audio clock offset for viseme synchronization.
-     * Total stream time (in ms) = (audioContext.currentTime - audioClockOffset) * 1000
+     * Internal method to schedule and play audio with sample-accurate timing
      */
-    getAudioClockOffset(): number | null;
+    private _schedulePlayback;
     /**
-     * Reset the audio clock offset (call when a response is interrupted or finished)
+     * Stop all currently playing audio and clear the queue
      */
-    resetAudioClock(): void;
     stopPlayback(): void;
+    /**
+     * Toggle mute state
+     */
     setMuted(muted: boolean): void;
-    isMicMuted(): boolean;
-    getAmplitude(): number;
-    getFrequencyData(): Uint8Array;
-    getWaveformData(): Uint8Array;
-    cleanup(): void;
-    getAudioContext(): AudioContext | null;
-}
-
-/**
- * High-level AI Voice Agent for browser-based conversations.
- *
- * This class orchestrates microphone input, AI processing, and
- * speaker output, providing a simple interface for building
- * voice assistants with lip-sync support.
- */
-declare class VoiceAgent {
-    private client;
-    private audioManager;
-    private options;
-    private isConnected;
-    private visemeQueue;
-    constructor(options: VoiceAgentOptions & {
-        apiKey: string;
-    });
     /**
-     * Initialize hardware and connect to the AI server.
-     * This must be called in response to a user guesture (like a click)
-     * to satisfy browser AudioContext requirements.
+     * Get current mute state
      */
-    connect(): Promise<boolean>;
+    isMicMuted(): boolean;
     /**
-     * Get the current amplitude/volume of the microphone or output audio.
-     * Useful for voice activity visualization.
-     * @returns value between 0 and 1
+     * Get current amplitude from analyser (for visualization)
+     * Returns value between 0 and 1
      */
     getAmplitude(): number;
     /**
-     * Mute or unmute the microphone.
+     * Get frequency data from analyser for visualization
      */
-    toggleMute(): boolean;
+    getFrequencyData(): Uint8Array;
     /**
-     * High-precision method to get visemes that should be active
-     * at the current playback frame. Use this in a requestAnimationFrame loop.
+     * Get time-domain data from analyser for waveform visualization
      */
-    getFrameVisemes(): Viseme[];
+    getWaveformData(): Uint8Array;
     /**
-     * Change the system prompt mid-conversation.
+     * Cleanup and close AudioContext
      */
-    updatePrompt(newPrompt: string): void;
+    cleanup(): void;
     /**
-     * Disconnect and release audio resources.
+     * Get current audio context state
      */
-    disconnect(): void;
+    getState(): 'running' | 'suspended' | 'closed' | 'interrupted' | null;
+    /**
+     * Check if microphone is currently listening
+     */
+    isRecording(): boolean;
 }
 
-export { AUDIO_CONFIG, type AnalyserConfig, type BrowserAudioConfig, BrowserAudioManager, type BrowserAudioOptions, DEFAULT_URLS, Language, type LokutorConfig, StreamResampler, type SynthesizeOptions, TTSClient, type Viseme, VoiceAgent, VoiceAgentClient, type VoiceAgentOptions, VoiceStyle, applyLowPassFilter, bytesToPcm16, calculateRMS, float32ToPcm16, normalizeAudio, pcm16ToBytes, pcm16ToFloat32, resample, resampleWithAntiAliasing, simpleConversation, simpleTTS };
+export { AUDIO_CONFIG, type AnalyserConfig, type BrowserAudioConfig, BrowserAudioManager, type BrowserAudioOptions, DEFAULT_URLS, Language, type LokutorConfig, StreamResampler, type SynthesizeOptions, TTSClient, type Viseme, VoiceAgentClient, type VoiceAgentOptions, VoiceStyle, applyLowPassFilter, bytesToPcm16, calculateRMS, float32ToPcm16, normalizeAudio, pcm16ToBytes, pcm16ToFloat32, resample, resampleWithAntiAliasing, simpleConversation, simpleTTS };

package/dist/index.js

@@ -26,7 +26,6 @@ __export(index_exports, {
   Language: () => Language,
   StreamResampler: () => StreamResampler,
   TTSClient: () => TTSClient,
-  VoiceAgent: () => VoiceAgent,
   VoiceAgentClient: () => VoiceAgentClient,
   VoiceStyle: () => VoiceStyle,
   applyLowPassFilter: () => applyLowPassFilter,
@@ -105,13 +104,11 @@ var VoiceAgentClient = class {
   messages = [];
   visemeListeners = [];
   wantVisemes = false;
-  serverUrl = null;
   constructor(config) {
     this.apiKey = config.apiKey;
     this.prompt = config.prompt;
     this.voice = config.voice || "F1" /* F1 */;
     this.language = config.language || "en" /* ENGLISH */;
-    this.serverUrl = config.serverUrl || null;
     this.onTranscription = config.onTranscription;
     this.onResponse = config.onResponse;
     this.onAudioCallback = config.onAudio;
@@ -126,12 +123,12 @@ var VoiceAgentClient = class {
   async connect() {
     return new Promise((resolve, reject) => {
       try {
-        let url = this.serverUrl || DEFAULT_URLS.VOICE_AGENT;
+        let url = DEFAULT_URLS.VOICE_AGENT;
         if (this.apiKey) {
           const separator = url.includes("?") ? "&" : "?";
           url += `${separator}api_key=${this.apiKey}`;
         }
-        console.log(`\u{1F517} Connecting to ${url}...`);
+        console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
         this.ws = new WebSocket(url);
         this.ws.binaryType = "arraybuffer";
         this.ws.onopen = () => {
@@ -504,8 +501,7 @@ var BrowserAudioManager = class {
   // Playback scheduling
   nextPlaybackTime = 0;
   activeSources = [];
-  // High-precision clock anchor for viseme sync
-  audioClockOffset = null;
+  playbackQueue = [];
   // Configuration
   inputSampleRate;
   outputSampleRate;
@@ -518,7 +514,6 @@ var BrowserAudioManager = class {
   // Audio processing state
   isMuted = false;
   isListening = false;
-  resampler = null;
   constructor(config = {}) {
     this.inputSampleRate = config.inputSampleRate ?? AUDIO_CONFIG.SAMPLE_RATE;
     this.outputSampleRate = config.outputSampleRate ?? AUDIO_CONFIG.SPEAKER_SAMPLE_RATE;
@@ -542,6 +537,7 @@ var BrowserAudioManager = class {
     }
     if (this.audioContext.state === "suspended") {
       await this.audioContext.resume();
+      console.log("\u{1F442} AudioContext resumed");
     }
     if (analyserConfig?.enabled !== false) {
       this.analyserNode = this.audioContext.createAnalyser();
@@ -555,7 +551,6 @@ var BrowserAudioManager = class {
     if (!this.audioContext) {
       await this.init();
     }
-    this.resampler = new StreamResampler(this.audioContext.sampleRate, this.inputSampleRate);
     try {
       this.onAudioInput = onAudioInput;
       this.isListening = true;
@@ -571,7 +566,9 @@ var BrowserAudioManager = class {
       this.scriptProcessor = this.audioContext.createScriptProcessor(
         bufferSize,
         1,
+        // input channels
         1
+        // output channels
       );
       this.mediaStreamAudioSourceNode.connect(this.scriptProcessor);
       this.scriptProcessor.connect(this.audioContext.destination);
@@ -588,19 +585,40 @@ var BrowserAudioManager = class {
       throw err;
     }
   }
+  /**
+   * Internal method to process microphone audio data
+   */
   _processAudioInput(event) {
-    if (!this.onAudioInput || !this.audioContext || !this.isListening || this.isMuted) return;
-    const inputData = event.inputBuffer.getChannelData(0);
-    event.outputBuffer.getChannelData(0).fill(0);
-    const resampled = this.resampler ? this.resampler.process(inputData) : inputData;
-    if (resampled && resampled.length > 0) {
-      const int16Data = float32ToPcm16(resampled);
-      this.onAudioInput(new Uint8Array(int16Data.buffer, int16Data.byteOffset, int16Data.byteLength));
+    if (!this.onAudioInput || !this.audioContext || !this.isListening) return;
+    if (this.isMuted) return;
+    const inputBuffer = event.inputBuffer;
+    const inputData = inputBuffer.getChannelData(0);
+    const outputBuffer = event.outputBuffer;
+    for (let i = 0; i < outputBuffer.getChannelData(0).length; i++) {
+      outputBuffer.getChannelData(0)[i] = 0;
+    }
+    const hardwareRate = this.audioContext.sampleRate;
+    let processedData = new Float32Array(inputData);
+    if (hardwareRate !== this.inputSampleRate) {
+      processedData = resampleWithAntiAliasing(
+        processedData,
+        hardwareRate,
+        this.inputSampleRate
+      );
     }
+    const int16Data = float32ToPcm16(processedData);
+    const uint8Data = new Uint8Array(
+      int16Data.buffer,
+      int16Data.byteOffset,
+      int16Data.byteLength
+    );
+    this.onAudioInput(uint8Data);
   }
+  /**
+   * Stop capturing microphone input
+   */
   stopMicrophone() {
     this.isListening = false;
-    this.resampler = null;
     if (this.mediaStream) {
       this.mediaStream.getTracks().forEach((track) => track.stop());
       this.mediaStream = null;
@@ -613,12 +631,17 @@ var BrowserAudioManager = class {
       this.mediaStreamAudioSourceNode.disconnect();
       this.mediaStreamAudioSourceNode = null;
     }
+    console.log("\u{1F3A4} Microphone stopped");
   }
   /**
    * Play back audio received from the server
+   * @param pcm16Data Int16 PCM audio data at SPEAKER_SAMPLE_RATE
    */
   playAudio(pcm16Data) {
-    if (!this.audioContext) return;
+    if (!this.audioContext) {
+      console.warn("AudioContext not initialized");
+      return;
+    }
     const int16Array = new Int16Array(
       pcm16Data.buffer,
       pcm16Data.byteOffset,
@@ -633,17 +656,18 @@ var BrowserAudioManager = class {
     audioBuffer.getChannelData(0).set(float32Data);
     this._schedulePlayback(audioBuffer);
   }
+  /**
+   * Internal method to schedule and play audio with sample-accurate timing
+   */
   _schedulePlayback(audioBuffer) {
     if (!this.audioContext) return;
     const currentTime = this.audioContext.currentTime;
     const duration = audioBuffer.length / this.outputSampleRate;
     const startTime = Math.max(
       currentTime + 0.01,
+      // Minimum 10ms delay
       this.nextPlaybackTime
     );
-    if (this.audioClockOffset === null) {
-      this.audioClockOffset = startTime;
-    }
     this.nextPlaybackTime = startTime + duration;
     const source = this.audioContext.createBufferSource();
     source.buffer = audioBuffer;
@@ -661,18 +685,8 @@ var BrowserAudioManager = class {
     };
   }
   /**
-   * Get the current high-precision audio clock offset for viseme synchronization.
-   * Total stream time (in ms) = (audioContext.currentTime - audioClockOffset) * 1000
-   */
-  getAudioClockOffset() {
-    return this.audioClockOffset;
-  }
-  /**
-   * Reset the audio clock offset (call when a response is interrupted or finished)
+   * Stop all currently playing audio and clear the queue
    */
-  resetAudioClock() {
-    this.audioClockOffset = null;
-  }
   stopPlayback() {
     this.activeSources.forEach((source) => {
       try {
@@ -681,15 +695,26 @@ var BrowserAudioManager = class {
       }
     });
     this.activeSources = [];
-    this.nextPlaybackTime = 0;
-    this.resetAudioClock();
+    this.playbackQueue = [];
+    this.nextPlaybackTime = this.audioContext?.currentTime ?? 0;
+    console.log("\u{1F507} Playback stopped");
   }
+  /**
+   * Toggle mute state
+   */
   setMuted(muted) {
     this.isMuted = muted;
   }
+  /**
+   * Get current mute state
+   */
   isMicMuted() {
     return this.isMuted;
   }
+  /**
+   * Get current amplitude from analyser (for visualization)
+   * Returns value between 0 and 1
+   */
   getAmplitude() {
     if (!this.analyserNode) return 0;
     const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
@@ -697,18 +722,31 @@ var BrowserAudioManager = class {
     const rms = calculateRMS(dataArray);
     return Math.min(rms * 10, 1);
   }
+  /**
+   * Get frequency data from analyser for visualization
+   */
   getFrequencyData() {
-    if (!this.analyserNode) return new Uint8Array(0);
+    if (!this.analyserNode) {
+      return new Uint8Array(0);
+    }
     const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
     this.analyserNode.getByteFrequencyData(dataArray);
     return dataArray;
   }
+  /**
+   * Get time-domain data from analyser for waveform visualization
+   */
   getWaveformData() {
-    if (!this.analyserNode) return new Uint8Array(0);
+    if (!this.analyserNode) {
+      return new Uint8Array(0);
+    }
     const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
     this.analyserNode.getByteTimeDomainData(dataArray);
     return dataArray;
   }
+  /**
+   * Cleanup and close AudioContext
+   */
   cleanup() {
     this.stopMicrophone();
     this.stopPlayback();
@@ -717,124 +755,17 @@ var BrowserAudioManager = class {
       this.analyserNode = null;
     }
   }
-  getAudioContext() {
-    return this.audioContext;
-  }
-};
-
-// src/voice-agent.ts
-var VoiceAgent = class {
-  client;
-  audioManager;
-  options;
-  isConnected = false;
-  visemeQueue = [];
-  constructor(options) {
-    this.options = options;
-    this.client = new VoiceAgentClient({
-      apiKey: options.apiKey,
-      prompt: options.prompt || "You are a helpful and friendly AI assistant.",
-      voice: options.voice || "F1" /* F1 */,
-      language: options.language || "en" /* ENGLISH */,
-      visemes: options.visemes ?? true,
-      serverUrl: options.serverUrl,
-      onTranscription: (text) => {
-        if (options.onTranscription) options.onTranscription(text, true);
-      },
-      onResponse: (text) => {
-        if (options.onTranscription) options.onTranscription(text, false);
-      },
-      onAudio: (data) => {
-        this.audioManager.playAudio(data);
-      },
-      onVisemes: (visemes) => {
-        this.visemeQueue.push(...visemes);
-        if (options.onVisemes) options.onVisemes(visemes);
-      },
-      onStatus: (status) => {
-        if (options.onStatusChange) options.onStatusChange(status);
-        if (status === "interrupted" || status === "thinking") {
-          this.audioManager.stopPlayback();
-          this.visemeQueue = [];
-        }
-      },
-      onError: (err) => {
-        if (options.onError) options.onError(err);
-      }
-    });
-    this.audioManager = new BrowserAudioManager({
-      autoGainControl: true,
-      echoCancellation: true,
-      noiseSuppression: true
-    });
-  }
   /**
-   * Initialize hardware and connect to the AI server.
-   * This must be called in response to a user guesture (like a click)
-   * to satisfy browser AudioContext requirements.
+   * Get current audio context state
    */
-  async connect() {
-    try {
-      await this.audioManager.init();
-      const connected = await this.client.connect();
-      if (!connected) return false;
-      this.isConnected = true;
-      await this.audioManager.startMicrophone((pcm16Data) => {
-        if (this.isConnected) {
-          this.client.sendAudio(pcm16Data);
-        }
-      });
-      return true;
-    } catch (err) {
-      if (this.options.onError) this.options.onError(err);
-      return false;
-    }
+  getState() {
+    return this.audioContext?.state ?? null;
   }
   /**
-   * Get the current amplitude/volume of the microphone or output audio.
-   * Useful for voice activity visualization.
-   * @returns value between 0 and 1
+   * Check if microphone is currently listening
    */
-  getAmplitude() {
-    return this.audioManager.getAmplitude();
-  }
-  /**
-   * Mute or unmute the microphone.
-   */
-  toggleMute() {
-    const currentState = this.audioManager.isMicMuted();
-    this.audioManager.setMuted(!currentState);
-    return !currentState;
-  }
-  /**
-   * High-precision method to get visemes that should be active 
-   * at the current playback frame. Use this in a requestAnimationFrame loop.
-   */
-  getFrameVisemes() {
-    const offset = this.audioManager.getAudioClockOffset();
-    const audioCtx = this.audioManager.getAudioContext();
-    if (offset === null || !audioCtx) return [];
-    const streamTime = (audioCtx.currentTime - offset) * 1e3;
-    const currentBatch = [];
-    while (this.visemeQueue.length > 0 && this.visemeQueue[0].t * 1e3 <= streamTime) {
-      currentBatch.push(this.visemeQueue.shift());
-    }
-    return currentBatch;
-  }
-  /**
-   * Change the system prompt mid-conversation.
-   */
-  updatePrompt(newPrompt) {
-    this.client.updatePrompt(newPrompt);
-  }
-  /**
-   * Disconnect and release audio resources.
-   */
-  disconnect() {
-    this.isConnected = false;
-    this.client.disconnect();
-    this.audioManager.cleanup();
-    this.visemeQueue = [];
+  isRecording() {
+    return this.isListening;
   }
 };
 // Annotate the CommonJS export names for ESM import in node:
@@ -845,7 +776,6 @@ var VoiceAgent = class {
   Language,
   StreamResampler,
   TTSClient,
-  VoiceAgent,
   VoiceAgentClient,
   VoiceStyle,
   applyLowPassFilter,

package/dist/index.mjs

@@ -60,13 +60,11 @@ var VoiceAgentClient = class {
   messages = [];
   visemeListeners = [];
   wantVisemes = false;
-  serverUrl = null;
   constructor(config) {
     this.apiKey = config.apiKey;
     this.prompt = config.prompt;
     this.voice = config.voice || "F1" /* F1 */;
     this.language = config.language || "en" /* ENGLISH */;
-    this.serverUrl = config.serverUrl || null;
     this.onTranscription = config.onTranscription;
     this.onResponse = config.onResponse;
     this.onAudioCallback = config.onAudio;
@@ -81,12 +79,12 @@ var VoiceAgentClient = class {
   async connect() {
     return new Promise((resolve, reject) => {
       try {
-        let url = this.serverUrl || DEFAULT_URLS.VOICE_AGENT;
+        let url = DEFAULT_URLS.VOICE_AGENT;
         if (this.apiKey) {
           const separator = url.includes("?") ? "&" : "?";
           url += `${separator}api_key=${this.apiKey}`;
         }
-        console.log(`\u{1F517} Connecting to ${url}...`);
+        console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
         this.ws = new WebSocket(url);
         this.ws.binaryType = "arraybuffer";
         this.ws.onopen = () => {
@@ -459,8 +457,7 @@ var BrowserAudioManager = class {
   // Playback scheduling
   nextPlaybackTime = 0;
   activeSources = [];
-  // High-precision clock anchor for viseme sync
-  audioClockOffset = null;
+  playbackQueue = [];
   // Configuration
   inputSampleRate;
   outputSampleRate;
@@ -473,7 +470,6 @@ var BrowserAudioManager = class {
   // Audio processing state
   isMuted = false;
   isListening = false;
-  resampler = null;
   constructor(config = {}) {
     this.inputSampleRate = config.inputSampleRate ?? AUDIO_CONFIG.SAMPLE_RATE;
     this.outputSampleRate = config.outputSampleRate ?? AUDIO_CONFIG.SPEAKER_SAMPLE_RATE;
@@ -497,6 +493,7 @@ var BrowserAudioManager = class {
     }
     if (this.audioContext.state === "suspended") {
       await this.audioContext.resume();
+      console.log("\u{1F442} AudioContext resumed");
     }
     if (analyserConfig?.enabled !== false) {
       this.analyserNode = this.audioContext.createAnalyser();
@@ -510,7 +507,6 @@ var BrowserAudioManager = class {
     if (!this.audioContext) {
       await this.init();
     }
-    this.resampler = new StreamResampler(this.audioContext.sampleRate, this.inputSampleRate);
     try {
       this.onAudioInput = onAudioInput;
       this.isListening = true;
@@ -526,7 +522,9 @@ var BrowserAudioManager = class {
       this.scriptProcessor = this.audioContext.createScriptProcessor(
         bufferSize,
         1,
+        // input channels
         1
+        // output channels
       );
       this.mediaStreamAudioSourceNode.connect(this.scriptProcessor);
       this.scriptProcessor.connect(this.audioContext.destination);
@@ -543,19 +541,40 @@ var BrowserAudioManager = class {
       throw err;
     }
   }
+  /**
+   * Internal method to process microphone audio data
+   */
   _processAudioInput(event) {
-    if (!this.onAudioInput || !this.audioContext || !this.isListening || this.isMuted) return;
-    const inputData = event.inputBuffer.getChannelData(0);
-    event.outputBuffer.getChannelData(0).fill(0);
-    const resampled = this.resampler ? this.resampler.process(inputData) : inputData;
-    if (resampled && resampled.length > 0) {
-      const int16Data = float32ToPcm16(resampled);
-      this.onAudioInput(new Uint8Array(int16Data.buffer, int16Data.byteOffset, int16Data.byteLength));
+    if (!this.onAudioInput || !this.audioContext || !this.isListening) return;
+    if (this.isMuted) return;
+    const inputBuffer = event.inputBuffer;
+    const inputData = inputBuffer.getChannelData(0);
+    const outputBuffer = event.outputBuffer;
+    for (let i = 0; i < outputBuffer.getChannelData(0).length; i++) {
+      outputBuffer.getChannelData(0)[i] = 0;
+    }
+    const hardwareRate = this.audioContext.sampleRate;
+    let processedData = new Float32Array(inputData);
+    if (hardwareRate !== this.inputSampleRate) {
+      processedData = resampleWithAntiAliasing(
+        processedData,
+        hardwareRate,
+        this.inputSampleRate
+      );
     }
+    const int16Data = float32ToPcm16(processedData);
+    const uint8Data = new Uint8Array(
+      int16Data.buffer,
+      int16Data.byteOffset,
+      int16Data.byteLength
+    );
+    this.onAudioInput(uint8Data);
   }
+  /**
+   * Stop capturing microphone input
+   */
   stopMicrophone() {
     this.isListening = false;
-    this.resampler = null;
     if (this.mediaStream) {
       this.mediaStream.getTracks().forEach((track) => track.stop());
       this.mediaStream = null;
@@ -568,12 +587,17 @@ var BrowserAudioManager = class {
       this.mediaStreamAudioSourceNode.disconnect();
       this.mediaStreamAudioSourceNode = null;
     }
+    console.log("\u{1F3A4} Microphone stopped");
   }
   /**
    * Play back audio received from the server
+   * @param pcm16Data Int16 PCM audio data at SPEAKER_SAMPLE_RATE
    */
   playAudio(pcm16Data) {
-    if (!this.audioContext) return;
+    if (!this.audioContext) {
+      console.warn("AudioContext not initialized");
+      return;
+    }
     const int16Array = new Int16Array(
       pcm16Data.buffer,
       pcm16Data.byteOffset,
@@ -588,17 +612,18 @@ var BrowserAudioManager = class {
     audioBuffer.getChannelData(0).set(float32Data);
     this._schedulePlayback(audioBuffer);
   }
+  /**
+   * Internal method to schedule and play audio with sample-accurate timing
+   */
   _schedulePlayback(audioBuffer) {
     if (!this.audioContext) return;
     const currentTime = this.audioContext.currentTime;
     const duration = audioBuffer.length / this.outputSampleRate;
     const startTime = Math.max(
       currentTime + 0.01,
+      // Minimum 10ms delay
       this.nextPlaybackTime
     );
-    if (this.audioClockOffset === null) {
-      this.audioClockOffset = startTime;
-    }
     this.nextPlaybackTime = startTime + duration;
     const source = this.audioContext.createBufferSource();
     source.buffer = audioBuffer;
@@ -616,18 +641,8 @@ var BrowserAudioManager = class {
     };
   }
   /**
-   * Get the current high-precision audio clock offset for viseme synchronization.
-   * Total stream time (in ms) = (audioContext.currentTime - audioClockOffset) * 1000
-   */
-  getAudioClockOffset() {
-    return this.audioClockOffset;
-  }
-  /**
-   * Reset the audio clock offset (call when a response is interrupted or finished)
+   * Stop all currently playing audio and clear the queue
    */
-  resetAudioClock() {
-    this.audioClockOffset = null;
-  }
   stopPlayback() {
     this.activeSources.forEach((source) => {
       try {
@@ -636,15 +651,26 @@ var BrowserAudioManager = class {
       }
     });
     this.activeSources = [];
-    this.nextPlaybackTime = 0;
-    this.resetAudioClock();
+    this.playbackQueue = [];
+    this.nextPlaybackTime = this.audioContext?.currentTime ?? 0;
+    console.log("\u{1F507} Playback stopped");
   }
+  /**
+   * Toggle mute state
+   */
   setMuted(muted) {
     this.isMuted = muted;
   }
+  /**
+   * Get current mute state
+   */
   isMicMuted() {
     return this.isMuted;
   }
+  /**
+   * Get current amplitude from analyser (for visualization)
+   * Returns value between 0 and 1
+   */
   getAmplitude() {
     if (!this.analyserNode) return 0;
     const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
@@ -652,18 +678,31 @@ var BrowserAudioManager = class {
     const rms = calculateRMS(dataArray);
     return Math.min(rms * 10, 1);
   }
+  /**
+   * Get frequency data from analyser for visualization
+   */
   getFrequencyData() {
-    if (!this.analyserNode) return new Uint8Array(0);
+    if (!this.analyserNode) {
+      return new Uint8Array(0);
+    }
     const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
     this.analyserNode.getByteFrequencyData(dataArray);
     return dataArray;
   }
+  /**
+   * Get time-domain data from analyser for waveform visualization
+   */
   getWaveformData() {
-    if (!this.analyserNode) return new Uint8Array(0);
+    if (!this.analyserNode) {
+      return new Uint8Array(0);
+    }
     const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
     this.analyserNode.getByteTimeDomainData(dataArray);
     return dataArray;
   }
+  /**
+   * Cleanup and close AudioContext
+   */
   cleanup() {
     this.stopMicrophone();
     this.stopPlayback();
@@ -672,124 +711,17 @@ var BrowserAudioManager = class {
       this.analyserNode = null;
     }
   }
-  getAudioContext() {
-    return this.audioContext;
-  }
-};
-
-// src/voice-agent.ts
-var VoiceAgent = class {
-  client;
-  audioManager;
-  options;
-  isConnected = false;
-  visemeQueue = [];
-  constructor(options) {
-    this.options = options;
-    this.client = new VoiceAgentClient({
-      apiKey: options.apiKey,
-      prompt: options.prompt || "You are a helpful and friendly AI assistant.",
-      voice: options.voice || "F1" /* F1 */,
-      language: options.language || "en" /* ENGLISH */,
-      visemes: options.visemes ?? true,
-      serverUrl: options.serverUrl,
-      onTranscription: (text) => {
-        if (options.onTranscription) options.onTranscription(text, true);
-      },
-      onResponse: (text) => {
-        if (options.onTranscription) options.onTranscription(text, false);
-      },
-      onAudio: (data) => {
-        this.audioManager.playAudio(data);
-      },
-      onVisemes: (visemes) => {
-        this.visemeQueue.push(...visemes);
-        if (options.onVisemes) options.onVisemes(visemes);
-      },
-      onStatus: (status) => {
-        if (options.onStatusChange) options.onStatusChange(status);
-        if (status === "interrupted" || status === "thinking") {
-          this.audioManager.stopPlayback();
-          this.visemeQueue = [];
-        }
-      },
-      onError: (err) => {
-        if (options.onError) options.onError(err);
-      }
-    });
-    this.audioManager = new BrowserAudioManager({
-      autoGainControl: true,
-      echoCancellation: true,
-      noiseSuppression: true
-    });
-  }
   /**
-   * Initialize hardware and connect to the AI server.
-   * This must be called in response to a user guesture (like a click)
-   * to satisfy browser AudioContext requirements.
+   * Get current audio context state
    */
-  async connect() {
-    try {
-      await this.audioManager.init();
-      const connected = await this.client.connect();
-      if (!connected) return false;
-      this.isConnected = true;
-      await this.audioManager.startMicrophone((pcm16Data) => {
-        if (this.isConnected) {
-          this.client.sendAudio(pcm16Data);
-        }
-      });
-      return true;
-    } catch (err) {
-      if (this.options.onError) this.options.onError(err);
-      return false;
-    }
+  getState() {
+    return this.audioContext?.state ?? null;
   }
   /**
-   * Get the current amplitude/volume of the microphone or output audio.
-   * Useful for voice activity visualization.
-   * @returns value between 0 and 1
+   * Check if microphone is currently listening
    */
-  getAmplitude() {
-    return this.audioManager.getAmplitude();
-  }
-  /**
-   * Mute or unmute the microphone.
-   */
-  toggleMute() {
-    const currentState = this.audioManager.isMicMuted();
-    this.audioManager.setMuted(!currentState);
-    return !currentState;
-  }
-  /**
-   * High-precision method to get visemes that should be active 
-   * at the current playback frame. Use this in a requestAnimationFrame loop.
-   */
-  getFrameVisemes() {
-    const offset = this.audioManager.getAudioClockOffset();
-    const audioCtx = this.audioManager.getAudioContext();
-    if (offset === null || !audioCtx) return [];
-    const streamTime = (audioCtx.currentTime - offset) * 1e3;
-    const currentBatch = [];
-    while (this.visemeQueue.length > 0 && this.visemeQueue[0].t * 1e3 <= streamTime) {
-      currentBatch.push(this.visemeQueue.shift());
-    }
-    return currentBatch;
-  }
-  /**
-   * Change the system prompt mid-conversation.
-   */
-  updatePrompt(newPrompt) {
-    this.client.updatePrompt(newPrompt);
-  }
-  /**
-   * Disconnect and release audio resources.
-   */
-  disconnect() {
-    this.isConnected = false;
-    this.client.disconnect();
-    this.audioManager.cleanup();
-    this.visemeQueue = [];
+  isRecording() {
+    return this.isListening;
   }
 };
 export {
@@ -799,7 +731,6 @@ export {
   Language,
   StreamResampler,
   TTSClient,
-  VoiceAgent,
   VoiceAgentClient,
   VoiceStyle,
   applyLowPassFilter,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lokutor/sdk",
-  "version": "1.1.9",
+  "version": "1.1.10",
   "description": "JavaScript/TypeScript SDK for Lokutor Real-time Voice AI",
   "main": "./dist/index.js",
   "module": "./dist/index.mjs",