@hamsa-ai/voice-agents-sdk 0.3.0 → 4.0.0-beta.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hamsa-ai/voice-agents-sdk",
-  "version": "0.3.0",
+  "version": "4.0.0-beta.1",
   "description": "Hamsa AI - Voice Agents JavaScript SDK",
   "main": "dist/index.cjs.js",
   "module": "dist/index.esm.js",
@@ -24,16 +24,23 @@
     "build:types": "tsc",
     "build:webpack": "webpack --config webpack.config.js",
     "build": "npm run clean && npm run build:types && npm run build:webpack",
+    "test": "jest",
+    "test:watch": "jest --watch",
     "prepare": "npm run build"
   },
   "author": "Hamsa AI Inc.",
   "license": "MIT",
   "devDependencies": {
+    "@babel/core": "^7.23.0",
+    "@babel/preset-env": "^7.23.0",
     "audio-worklet-loader": "^1.1.0",
+    "babel-jest": "^29.7.0",
     "babel-loader": "^9.1.3",
     "buffer": "^6.0.3",
     "css-loader": "^7.1.2",
     "file-loader": "^6.2.0",
+    "jest": "^29.7.0",
+    "jest-environment-jsdom": "^29.7.0",
     "rimraf": "^6.0.1",
     "style-loader": "^4.0.0",
     "typescript": "^5.5.4",
@@ -42,7 +49,8 @@
     "webpack-node-externals": "^3.0.0"
   },
   "dependencies": {
-    "events": "^3.3.0"
+    "events": "^3.3.0",
+    "livekit-client": "^2.15.4"
   },
   "repository": {
     "type": "git",

package/types/classes/livekit_manager.d.ts

@@ -0,0 +1,52 @@
+/**
+ * LiveKitManager encapsulates all direct interactions with the livekit-client SDK
+ * and replaces the WebSocketManager functionality.
+ */
+export default class LiveKitManager extends EventEmitter<[never]> {
+    /**
+     * Creates a new LiveKitManager instance.
+     * @param {string} lkUrl - The LiveKit WebSocket URL
+     * @param {string} accessToken - The LiveKit access token
+     * @param {Array} tools - Array of tools/functions to be used
+     */
+    constructor(lkUrl: string, accessToken: string, tools?: any[]);
+    lkUrl: string;
+    accessToken: string;
+    tools: any[];
+    room: Room;
+    isConnected: boolean;
+    volume: number;
+    audioElements: Set<any>;
+    isPaused: boolean;
+    /**
+     * Connects to the LiveKit room.
+     * @returns {Promise<void>}
+     */
+    connect(): Promise<void>;
+    /**
+     * Disconnects from the LiveKit room.
+     * @returns {Promise<void>}
+     */
+    disconnect(): Promise<void>;
+    /**
+     * Pauses the call by muting local microphone and pausing remote audio.
+     */
+    pause(): void;
+    /**
+     * Resumes the call by unmuting local microphone and resuming remote audio.
+     */
+    resume(): void;
+    /**
+     * Sets the volume for all audio elements.
+     * @param {number} volume - Volume level between 0.0 and 1.0
+     */
+    setVolume(volume: number): void;
+    /**
+     * Registers client-side tools as RPC methods.
+     * @param {Array} tools - Array of tool objects
+     */
+    registerTools(tools: any[]): void;
+    #private;
+}
+import { EventEmitter } from 'events';
+import { Room } from 'livekit-client';

package/types/main.d.ts

@@ -6,17 +6,14 @@ export class HamsaVoiceAgent extends EventEmitter<[never]> {
      * @param {string} apiKey - API key.
      * @param {object} [config] - Optional config.
      * @param {string} [config.API_URL="https://api.tryhamsa.com"] - API URL.
-     * @param {string} [config.WS_URL="wss://bots.tryhamsa.com/stream"] - WebSocket URL.
      */
-    constructor(apiKey: string, { API_URL, WS_URL, }?: {
+    constructor(apiKey: string, { API_URL, }?: {
         API_URL?: string;
-        WS_URL?: string;
     });
-    webSocketManager: WebSocketManager;
+    liveKitManager: LiveKitManager;
     apiKey: string;
     API_URL: string;
-    WS_URL: string;
-    jobId: string;
+    jobId: any;
     wakeLockManager: ScreenWakeLock;
     /**
      * Sets the volume for the audio playback.
@@ -27,7 +24,7 @@ export class HamsaVoiceAgent extends EventEmitter<[never]> {
      * Starts a new voice agent call.
      * @param {object} options - Configuration options for the call.
      */
-    start({ agentId, params, voiceEnablement, tools }: object): Promise<void>;
+    start({ agentId, params, voiceEnablement, tools, }: object): Promise<void>;
     /**
      * Ends the current voice agent call.
      */
@@ -51,6 +48,6 @@ export class HamsaVoiceAgent extends EventEmitter<[never]> {
     getJobDetails(maxRetries?: number, initialRetryInterval?: number, backoffFactor?: number): Promise<any>;
     #private;
 }
-import { EventEmitter } from 'events';
-import WebSocketManager from './classes/websocket_manager';
+import { EventEmitter } from "events";
+import LiveKitManager from "./classes/livekit_manager";
 import ScreenWakeLock from "./classes/screen_wake_lock";

package/dist/index.cjs.js.LICENSE.txt

@@ -1,8 +0,0 @@
-/*!
- * The buffer module from node.js, for the browser.
- *
- * @author   Feross Aboukhadijeh <https://feross.org>
- * @license  MIT
- */
-
-/*! ieee754. BSD-3-Clause License. Feross Aboukhadijeh <https://feross.org/opensource> */

package/dist/index.esm.js.LICENSE.txt

@@ -1,8 +0,0 @@
-/*!
- * The buffer module from node.js, for the browser.
- *
- * @author   Feross Aboukhadijeh <https://feross.org>
- * @license  MIT
- */
-
-/*! ieee754. BSD-3-Clause License. Feross Aboukhadijeh <https://feross.org/opensource> */

package/dist/index.umd.js.LICENSE.txt

@@ -1,8 +0,0 @@
-/*!
- * The buffer module from node.js, for the browser.
- *
- * @author   Feross Aboukhadijeh <https://feross.org>
- * @license  MIT
- */
-
-/*! ieee754. BSD-3-Clause License. Feross Aboukhadijeh <https://feross.org/opensource> */

package/types/classes/audio-player-processor.worklet.d.ts

@@ -1,8 +0,0 @@
-declare class AudioPlayerProcessor {
-    audioData: any[];
-    isPaused: boolean;
-    marks: any[];
-    isDone: boolean;
-    clearAllData(): void;
-    process(inputs: any, outputs: any): boolean;
-}

package/types/classes/audio-processor.worklet.d.ts

@@ -1,4 +0,0 @@
-declare class AudioProcessor {
-    encodeBase64(bytes: any): string;
-    process(inputs: any, outputs: any, parameters: any): boolean;
-}

package/types/classes/audio_player.d.ts

@@ -1,67 +0,0 @@
-export default class AudioPlayer {
-    /**
-     * Constructs a new AudioPlayer instance.
-     * @param {WebSocket} ws - The WebSocket connection.
-     * @param {function} onSpeaking - Callback when speaking starts.
-     * @param {function} onListening - Callback when listening starts.
-     * @param {function} onStreamReady - Callback when MediaStream is ready.
-     */
-    constructor(ws: WebSocket, onSpeaking: Function, onListening: Function, onStreamReady: Function);
-    audioContext: any;
-    ws: WebSocket;
-    isPaused: boolean;
-    onSpeakingCB: Function;
-    onListeningCB: Function;
-    isPlaying: boolean;
-    gainNode: any;
-    mediaStreamDestination: any;
-    onStreamReady: Function;
-    /**
-     * Initializes the AudioWorklet and sets up the processor.
-     */
-    initAudioWorklet(): Promise<void>;
-    processor: AudioWorkletNode;
-    /**
-     * Enqueues audio data to be played.
-     * @param {string} base64Data - Base64 encoded PCM16 audio data.
-     */
-    enqueueAudio(base64Data: string): void;
-    /**
-     * Pauses audio playback.
-     */
-    pause(): void;
-    /**
-     * Resumes audio playback.
-     */
-    resume(): void;
-    /**
-     * Stops audio playback and clears the buffer.
-     */
-    stopAndClear(): void;
-    /**
-     * Adds a mark to the audio stream.
-     * @param {string} markName - Name of the mark.
-     */
-    addMark(markName: string): void;
-    /**
-     * Converts PCM16 data to Float32.
-     * @param {Uint8Array} pcm16Array - PCM16 audio data.
-     * @returns {Float32Array} Float32 audio samples.
-     */
-    pcm16ToFloat32(pcm16Array: Uint8Array): Float32Array;
-    /**
-     * Updates the playing state and triggers callbacks.
-     * @param {boolean} isPlaying - Indicates whether audio is playing.
-     */
-    updatePlayingState(isPlaying: boolean): void;
-    /**
-     * Sets the volume of the audio playback.
-     * @param {number} volume - Volume level between 0.0 and 1.0.
-     */
-    setVolume(volume: number): void;
-    /**
-     * Returns the MediaStream capturing the audio being played.
-     * @returns {MediaStream}
-     */
-    getMediaStream(): MediaStream;
-}

package/types/classes/audio_recorder.d.ts

@@ -1,37 +0,0 @@
-export default class AudioRecorder {
-    /**
-     * Constructs a new AudioRecorder instance.
-     * @param {function} onStreamReady - Callback invoked when MediaStream is ready.
-     */
-    constructor(onStreamReady: Function);
-    audioContext: any;
-    mediaStreamSource: any;
-    audioWorkletNode: AudioWorkletNode;
-    mediaStream: MediaStream;
-    isPaused: boolean;
-    mediaStreamDestination: any;
-    onStreamReady: Function;
-    /**
-     * Starts streaming audio by capturing from the microphone and sending it over WebSocket.
-     * @param {WebSocket} ws - The WebSocket connection.
-     * @returns {Promise<void>}
-     */
-    startStreaming(ws: WebSocket): Promise<void>;
-    /**
-     * Pauses audio streaming.
-     */
-    pause(): void;
-    /**
-     * Resumes audio streaming.
-     */
-    resume(): void;
-    /**
-     * Stops audio streaming and releases resources.
-     */
-    stop(): void;
-    /**
-     * Returns the MediaStream capturing the local audio being recorded.
-     * @returns {MediaStream | null}
-     */
-    getMediaStream(): MediaStream | null;
-}

package/types/classes/websocket_manager.d.ts

@@ -1,115 +0,0 @@
-export default class WebSocketManager {
-    /**
-     * Constructs a new WebSocketManager instance.
-     * @param {string} url - The WebSocket URL.
-     * @param {string} conversationId - The conversation ID.
-     * @param {function} onError - Callback for error events.
-     * @param {function} onStart - Callback when the WebSocket starts.
-     * @param {function} onTranscriptionReceived - Callback for received transcriptions.
-     * @param {function} onAnswerReceived - Callback for received answers.
-     * @param {function} onSpeaking - Callback when speaking starts.
-     * @param {function} onListening - Callback when listening starts.
-     * @param {function} onClosed - Callback when the WebSocket is closed.
-     * @param {boolean} voiceEnablement - Flag to enable voice features.
-     * @param {Array} tools - Array of tools/functions to be used.
-     * @param {string} apiKey - API key for authentication.
-     * @param {function} onRemoteStreamAvailable - Callback when remote MediaStream is available.
-     * @param {function} onLocalStreamAvailable - Callback when local MediaStream is available.
-     * @param {function} onInfo - Callback for info events.
-     */
-    constructor(url: string, conversationId: string, onError: Function, onStart: Function, onTranscriptionReceived: Function, onAnswerReceived: Function, onSpeaking: Function, onListening: Function, onClosed: Function, voiceEnablement: boolean, tools: any[], apiKey: string, onRemoteStreamAvailable: Function, onLocalStreamAvailable: Function, onInfo: Function);
-    url: string;
-    ws: WebSocket;
-    isConnected: boolean;
-    audioPlayer: AudioPlayer;
-    audioRecorder: AudioRecorder;
-    last_transcription_date: Date;
-    last_voice_byte_date: Date;
-    is_media: boolean;
-    onErrorCB: Function;
-    onStartCB: Function;
-    onTranscriptionReceivedCB: Function;
-    onAnswerReceivedCB: Function;
-    onSpeakingCB: Function;
-    onListeningCB: Function;
-    onClosedCB: Function;
-    voiceEnablement: boolean;
-    tools: any[];
-    apiKey: string;
-    onRemoteStreamAvailable: Function;
-    onLocalStreamAvailable: Function;
-    onInfoCB: Function;
-    /**
-     * Sets the volume for AudioPlayer.
-     * @param {number} volume - Volume level between 0.0 and 1.0.
-     */
-    setVolume(volume: number): void;
-    /**
-     * Initializes and starts the WebSocket connection, AudioPlayer, and AudioRecorder.
-     */
-    startCall(): void;
-    /**
-     * Handles the WebSocket 'open' event.
-     */
-    onOpen(): void;
-    /**
-     * Handles incoming WebSocket messages.
-     * @param {MessageEvent} event - The message event.
-     */
-    onMessage(event: MessageEvent): void;
-    /**
-     * Handles the WebSocket 'close' event.
-     * @param {CloseEvent} event - The close event.
-     */
-    onClose(event: CloseEvent): void;
-    /**
-     * Handles the WebSocket 'error' event.
-     * @param {Event} error - The error event.
-     */
-    onError(error: Event): void;
-    /**
-     * Ends the WebSocket call, stops AudioPlayer and AudioRecorder.
-     */
-    endCall(): void;
-    /**
-     * Pauses the WebSocket call by pausing AudioPlayer and AudioRecorder.
-     */
-    pauseCall(): void;
-    /**
-     * Resumes the WebSocket call by resuming AudioPlayer and AudioRecorder.
-     */
-    resumeCall(): void;
-    /**
-     * Executes the provided tool functions, supporting both synchronous and asynchronous calls.
-     *
-     * This function iterates over the given array of tool objects. For each tool of type
-     * "function", it attempts to locate the corresponding function by its name in the tools list.
-     * The function arguments are parsed from a JSON string and passed to the function.
-     * If the function executes successfully, its response is captured; otherwise, an error
-     * message is returned. The results are returned in the same order as the input array.
-     *
-     * @param {Array} tools_array - An array of tool objects to execute. Each object should
-     * have the following structure:
-     *   {
-     *     id: <unique identifier>,
-     *     type: "function",
-     *     function: {
-     *       name: <string>,           // The function's name.
-     *       arguments: <string>       // A JSON string representing the function arguments.
-     *     }
-     *   }
-     * @returns {Promise<Array>} A promise that resolves to an array of results. Each result object has
-     * the structure:
-     *   {
-     *     id: <tool id>,
-     *     function: {
-     *       name: <function name>,
-     *       response: <result of the function call or an error message>
-     *     }
-     *   }
-     */
-    run_tools(tools_array: any[]): Promise<any[]>;
-    #private;
-}
-import AudioPlayer from './audio_player';
-import AudioRecorder from './audio_recorder';