@lokutor/sdk 1.0.0 → 1.1.0

package/dist/index.d.mts

@@ -47,7 +47,7 @@ interface LokutorConfig {
     serverUrl?: string;
     onTranscription?: (text: string) => void;
     onResponse?: (text: string) => void;
-    onAudio?: (data: Buffer) => void;
+    onAudio?: (data: Uint8Array) => void;
     onStatus?: (status: string) => void;
     onError?: (error: any) => void;
 }
@@ -98,7 +98,7 @@ declare class VoiceAgentClient {
      * Send raw PCM audio data to the server
      * @param audioData Int16 PCM audio buffer
      */
-    sendAudio(audioData: Buffer | Uint8Array): void;
+    sendAudio(audioData: Uint8Array): void;
     /**
      * Handle incoming binary data (audio response)
      */
@@ -109,7 +109,7 @@ declare class VoiceAgentClient {
     private handleTextMessage;
     private audioListeners;
     private emit;
-    onAudio(callback: (data: Buffer) => void): void;
+    onAudio(callback: (data: Uint8Array) => void): void;
     /**
      * Disconnect from the server
      */
@@ -121,9 +121,6 @@ declare class VoiceAgentClient {
 declare class TTSClient {
     private apiKey;
     private serverUrl;
-    private onAudioCallback?;
-    private onVisemesCallback?;
-    private onErrorCallback?;
     constructor(config: {
         apiKey: string;
         serverUrl?: string;
@@ -141,7 +138,7 @@ declare class TTSClient {
         speed?: number;
         steps?: number;
         visemes?: boolean;
-        onAudio?: (data: Buffer) => void;
+        onAudio?: (data: Uint8Array) => void;
         onVisemes?: (visemes: any[]) => void;
         onError?: (error: any) => void;
     }): Promise<void>;
@@ -157,7 +154,7 @@ declare function simpleConversation(config: LokutorConfig & {
  */
 declare function simpleTTS(options: SynthesizeOptions & {
     apiKey: string;
-    onAudio: (buf: Buffer) => void;
+    onAudio: (buf: Uint8Array) => void;
 }): Promise<void>;
 
 export { AUDIO_CONFIG, DEFAULT_URLS, Language, type LokutorConfig, type SynthesizeOptions, TTSClient, VoiceAgentClient, VoiceStyle, simpleConversation, simpleTTS };

package/dist/index.d.ts

@@ -47,7 +47,7 @@ interface LokutorConfig {
     serverUrl?: string;
     onTranscription?: (text: string) => void;
     onResponse?: (text: string) => void;
-    onAudio?: (data: Buffer) => void;
+    onAudio?: (data: Uint8Array) => void;
     onStatus?: (status: string) => void;
     onError?: (error: any) => void;
 }
@@ -98,7 +98,7 @@ declare class VoiceAgentClient {
      * Send raw PCM audio data to the server
      * @param audioData Int16 PCM audio buffer
      */
-    sendAudio(audioData: Buffer | Uint8Array): void;
+    sendAudio(audioData: Uint8Array): void;
     /**
      * Handle incoming binary data (audio response)
      */
@@ -109,7 +109,7 @@ declare class VoiceAgentClient {
     private handleTextMessage;
     private audioListeners;
     private emit;
-    onAudio(callback: (data: Buffer) => void): void;
+    onAudio(callback: (data: Uint8Array) => void): void;
     /**
      * Disconnect from the server
      */
@@ -121,9 +121,6 @@ declare class VoiceAgentClient {
 declare class TTSClient {
     private apiKey;
     private serverUrl;
-    private onAudioCallback?;
-    private onVisemesCallback?;
-    private onErrorCallback?;
     constructor(config: {
         apiKey: string;
         serverUrl?: string;
@@ -141,7 +138,7 @@ declare class TTSClient {
         speed?: number;
         steps?: number;
         visemes?: boolean;
-        onAudio?: (data: Buffer) => void;
+        onAudio?: (data: Uint8Array) => void;
         onVisemes?: (visemes: any[]) => void;
         onError?: (error: any) => void;
     }): Promise<void>;
@@ -157,7 +154,7 @@ declare function simpleConversation(config: LokutorConfig & {
  */
 declare function simpleTTS(options: SynthesizeOptions & {
     apiKey: string;
-    onAudio: (buf: Buffer) => void;
+    onAudio: (buf: Uint8Array) => void;
 }): Promise<void>;
 
 export { AUDIO_CONFIG, DEFAULT_URLS, Language, type LokutorConfig, type SynthesizeOptions, TTSClient, VoiceAgentClient, VoiceStyle, simpleConversation, simpleTTS };

package/dist/index.js

@@ -67,7 +67,14 @@ var DEFAULT_URLS = {
 };
 
 // src/client.ts
-var import_ws = require("ws");
+function base64ToUint8Array(base64) {
+  const binaryString = atob(base64);
+  const bytes = new Uint8Array(binaryString.length);
+  for (let i = 0; i < binaryString.length; i++) {
+    bytes[i] = binaryString.charCodeAt(i);
+  }
+  return bytes;
+}
 var VoiceAgentClient = class {
   ws = null;
   apiKey;
@@ -100,36 +107,36 @@ var VoiceAgentClient = class {
   async connect() {
     return new Promise((resolve, reject) => {
       try {
-        console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
-        const headers = {};
+        let url = this.serverUrl;
         if (this.apiKey) {
-          headers["X-API-Key"] = this.apiKey;
+          const separator = url.includes("?") ? "&" : "?";
+          url += `${separator}api_key=${this.apiKey}`;
         }
-        this.ws = new import_ws.WebSocket(this.serverUrl, {
-          headers
-        });
-        this.ws.on("open", () => {
+        console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
+        this.ws = new WebSocket(url);
+        this.ws.binaryType = "arraybuffer";
+        this.ws.onopen = () => {
           this.isConnected = true;
           console.log("\u2705 Connected to voice agent!");
           this.sendConfig();
           resolve(true);
-        });
-        this.ws.on("message", (data, isBinary) => {
-          if (isBinary) {
-            this.handleBinaryMessage(data);
+        };
+        this.ws.onmessage = async (event) => {
+          if (event.data instanceof ArrayBuffer) {
+            this.handleBinaryMessage(new Uint8Array(event.data));
           } else {
-            this.handleTextMessage(data.toString());
+            this.handleTextMessage(event.data.toString());
           }
-        });
-        this.ws.on("error", (err) => {
+        };
+        this.ws.onerror = (err) => {
           console.error("\u274C WebSocket error:", err);
           if (this.onError) this.onError(err);
           if (!this.isConnected) reject(err);
-        });
-        this.ws.on("close", () => {
+        };
+        this.ws.onclose = () => {
           this.isConnected = false;
           console.log("Disconnected");
-        });
+        };
       } catch (err) {
         if (this.onError) this.onError(err);
         reject(err);
@@ -152,7 +159,7 @@ var VoiceAgentClient = class {
    */
   sendAudio(audioData) {
     if (this.ws && this.isConnected) {
-      this.ws.send(audioData, { binary: true });
+      this.ws.send(audioData);
     }
   }
   /**
@@ -170,7 +177,7 @@ var VoiceAgentClient = class {
       switch (msg.type) {
         case "audio":
           if (msg.data) {
-            const buffer = Buffer.from(msg.data, "base64");
+            const buffer = base64ToUint8Array(msg.data);
             this.handleBinaryMessage(buffer);
           }
           break;
@@ -224,9 +231,6 @@ var VoiceAgentClient = class {
 var TTSClient = class {
   apiKey;
   serverUrl;
-  onAudioCallback;
-  onVisemesCallback;
-  onErrorCallback;
   constructor(config) {
     this.apiKey = config.apiKey;
     this.serverUrl = config.serverUrl || DEFAULT_URLS.TTS;
@@ -240,12 +244,14 @@ var TTSClient = class {
   synthesize(options) {
     return new Promise((resolve, reject) => {
       try {
-        const headers = {};
+        let url = this.serverUrl;
         if (this.apiKey) {
-          headers["X-API-Key"] = this.apiKey;
+          const separator = url.includes("?") ? "&" : "?";
+          url += `${separator}api_key=${this.apiKey}`;
         }
-        const ws = new import_ws.WebSocket(this.serverUrl, { headers });
-        ws.on("open", () => {
+        const ws = new WebSocket(url);
+        ws.binaryType = "arraybuffer";
+        ws.onopen = () => {
           const req = {
             text: options.text,
             voice: options.voice || "F1" /* F1 */,
@@ -255,27 +261,27 @@ var TTSClient = class {
             visemes: options.visemes || false
           };
           ws.send(JSON.stringify(req));
-        });
-        ws.on("message", (data, isBinary) => {
-          if (isBinary) {
-            if (options.onAudio) options.onAudio(data);
+        };
+        ws.onmessage = async (event) => {
+          if (event.data instanceof ArrayBuffer) {
+            if (options.onAudio) options.onAudio(new Uint8Array(event.data));
           } else {
             try {
-              const msg = JSON.parse(data.toString());
+              const msg = JSON.parse(event.data.toString());
               if (Array.isArray(msg) && options.onVisemes) {
                 options.onVisemes(msg);
               }
             } catch (e) {
             }
           }
-        });
-        ws.on("error", (err) => {
+        };
+        ws.onerror = (err) => {
           if (options.onError) options.onError(err);
           reject(err);
-        });
-        ws.on("close", () => {
+        };
+        ws.onclose = () => {
           resolve();
-        });
+        };
       } catch (err) {
         if (options.onError) options.onError(err);
         reject(err);

package/dist/index.mjs

@@ -34,7 +34,14 @@ var DEFAULT_URLS = {
 };
 
 // src/client.ts
-import { WebSocket } from "ws";
+function base64ToUint8Array(base64) {
+  const binaryString = atob(base64);
+  const bytes = new Uint8Array(binaryString.length);
+  for (let i = 0; i < binaryString.length; i++) {
+    bytes[i] = binaryString.charCodeAt(i);
+  }
+  return bytes;
+}
 var VoiceAgentClient = class {
   ws = null;
   apiKey;
@@ -67,36 +74,36 @@ var VoiceAgentClient = class {
   async connect() {
     return new Promise((resolve, reject) => {
       try {
-        console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
-        const headers = {};
+        let url = this.serverUrl;
         if (this.apiKey) {
-          headers["X-API-Key"] = this.apiKey;
+          const separator = url.includes("?") ? "&" : "?";
+          url += `${separator}api_key=${this.apiKey}`;
         }
-        this.ws = new WebSocket(this.serverUrl, {
-          headers
-        });
-        this.ws.on("open", () => {
+        console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
+        this.ws = new WebSocket(url);
+        this.ws.binaryType = "arraybuffer";
+        this.ws.onopen = () => {
           this.isConnected = true;
           console.log("\u2705 Connected to voice agent!");
           this.sendConfig();
           resolve(true);
-        });
-        this.ws.on("message", (data, isBinary) => {
-          if (isBinary) {
-            this.handleBinaryMessage(data);
+        };
+        this.ws.onmessage = async (event) => {
+          if (event.data instanceof ArrayBuffer) {
+            this.handleBinaryMessage(new Uint8Array(event.data));
           } else {
-            this.handleTextMessage(data.toString());
+            this.handleTextMessage(event.data.toString());
           }
-        });
-        this.ws.on("error", (err) => {
+        };
+        this.ws.onerror = (err) => {
           console.error("\u274C WebSocket error:", err);
           if (this.onError) this.onError(err);
           if (!this.isConnected) reject(err);
-        });
-        this.ws.on("close", () => {
+        };
+        this.ws.onclose = () => {
           this.isConnected = false;
           console.log("Disconnected");
-        });
+        };
       } catch (err) {
         if (this.onError) this.onError(err);
         reject(err);
@@ -119,7 +126,7 @@ var VoiceAgentClient = class {
    */
   sendAudio(audioData) {
     if (this.ws && this.isConnected) {
-      this.ws.send(audioData, { binary: true });
+      this.ws.send(audioData);
     }
   }
   /**
@@ -137,7 +144,7 @@ var VoiceAgentClient = class {
       switch (msg.type) {
         case "audio":
           if (msg.data) {
-            const buffer = Buffer.from(msg.data, "base64");
+            const buffer = base64ToUint8Array(msg.data);
             this.handleBinaryMessage(buffer);
           }
           break;
@@ -191,9 +198,6 @@ var VoiceAgentClient = class {
 var TTSClient = class {
   apiKey;
   serverUrl;
-  onAudioCallback;
-  onVisemesCallback;
-  onErrorCallback;
   constructor(config) {
     this.apiKey = config.apiKey;
     this.serverUrl = config.serverUrl || DEFAULT_URLS.TTS;
@@ -207,12 +211,14 @@ var TTSClient = class {
   synthesize(options) {
     return new Promise((resolve, reject) => {
       try {
-        const headers = {};
+        let url = this.serverUrl;
         if (this.apiKey) {
-          headers["X-API-Key"] = this.apiKey;
+          const separator = url.includes("?") ? "&" : "?";
+          url += `${separator}api_key=${this.apiKey}`;
         }
-        const ws = new WebSocket(this.serverUrl, { headers });
-        ws.on("open", () => {
+        const ws = new WebSocket(url);
+        ws.binaryType = "arraybuffer";
+        ws.onopen = () => {
           const req = {
             text: options.text,
             voice: options.voice || "F1" /* F1 */,
@@ -222,27 +228,27 @@ var TTSClient = class {
             visemes: options.visemes || false
           };
           ws.send(JSON.stringify(req));
-        });
-        ws.on("message", (data, isBinary) => {
-          if (isBinary) {
-            if (options.onAudio) options.onAudio(data);
+        };
+        ws.onmessage = async (event) => {
+          if (event.data instanceof ArrayBuffer) {
+            if (options.onAudio) options.onAudio(new Uint8Array(event.data));
           } else {
             try {
-              const msg = JSON.parse(data.toString());
+              const msg = JSON.parse(event.data.toString());
               if (Array.isArray(msg) && options.onVisemes) {
                 options.onVisemes(msg);
               }
             } catch (e) {
             }
           }
-        });
-        ws.on("error", (err) => {
+        };
+        ws.onerror = (err) => {
           if (options.onError) options.onError(err);
           reject(err);
-        });
-        ws.on("close", () => {
+        };
+        ws.onclose = () => {
           resolve();
-        });
+        };
       } catch (err) {
         if (options.onError) options.onError(err);
         reject(err);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lokutor/sdk",
-  "version": "1.0.0",
+  "version": "1.1.0",
   "description": "JavaScript/TypeScript SDK for Lokutor Real-time Voice AI",
   "main": "./dist/index.js",
   "module": "./dist/index.mjs",
@@ -27,14 +27,11 @@
   ],
   "author": "Lokutor AI",
   "license": "MIT",
-  "dependencies": {
-    "ws": "^8.16.0"
-  },
+  "dependencies": {},
   "devDependencies": {
     "@types/node": "^20.10.0",
-    "@types/ws": "^8.5.10",
     "tsup": "^8.0.1",
     "typescript": "^5.3.2",
     "vitest": "^1.0.1"
   }
-}
+}