@lokutor/sdk 1.0.0 → 1.1.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (6) hide show
  1. package/README.md +2 -0
  2. package/dist/index.d.mts +23 -12
  3. package/dist/index.d.ts +23 -12
  4. package/dist/index.js +79 -41
  5. package/dist/index.mjs +79 -41
  6. package/package.json +3 -6
package/README.md CHANGED
@@ -1,5 +1,7 @@
1
1
  # Lokutor JavaScript SDK
2
2
 
3
+ ![Lokutor Hero](../../assets/img/hero.png)
4
+
3
5
  [![npm version](https://img.shields.io/npm/v/@lokutor/sdk.svg)](https://www.npmjs.com/package/@lokutor/sdk)
4
6
  [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
5
7
 
package/dist/index.d.mts CHANGED
@@ -28,6 +28,7 @@ declare enum Language {
28
28
  */
29
29
  declare const AUDIO_CONFIG: {
30
30
  SAMPLE_RATE: number;
31
+ SPEAKER_SAMPLE_RATE: number;
31
32
  CHANNELS: number;
32
33
  CHUNK_DURATION_MS: number;
33
34
  readonly CHUNK_SIZE: number;
@@ -44,10 +45,9 @@ declare const DEFAULT_URLS: {
44
45
  */
45
46
  interface LokutorConfig {
46
47
  apiKey: string;
47
- serverUrl?: string;
48
48
  onTranscription?: (text: string) => void;
49
49
  onResponse?: (text: string) => void;
50
- onAudio?: (data: Buffer) => void;
50
+ onAudio?: (data: Uint8Array) => void;
51
51
  onStatus?: (status: string) => void;
52
52
  onError?: (error: any) => void;
53
53
  }
@@ -71,7 +71,6 @@ interface SynthesizeOptions {
71
71
  declare class VoiceAgentClient {
72
72
  private ws;
73
73
  private apiKey;
74
- private serverUrl;
75
74
  prompt: string;
76
75
  voice: VoiceStyle;
77
76
  language: Language;
@@ -81,6 +80,7 @@ declare class VoiceAgentClient {
81
80
  private onStatus?;
82
81
  private onError?;
83
82
  private isConnected;
83
+ private messages;
84
84
  constructor(config: LokutorConfig & {
85
85
  prompt: string;
86
86
  voice?: VoiceStyle;
@@ -98,7 +98,7 @@ declare class VoiceAgentClient {
98
98
  * Send raw PCM audio data to the server
99
99
  * @param audioData Int16 PCM audio buffer
100
100
  */
101
- sendAudio(audioData: Buffer | Uint8Array): void;
101
+ sendAudio(audioData: Uint8Array): void;
102
102
  /**
103
103
  * Handle incoming binary data (audio response)
104
104
  */
@@ -109,24 +109,35 @@ declare class VoiceAgentClient {
109
109
  private handleTextMessage;
110
110
  private audioListeners;
111
111
  private emit;
112
- onAudio(callback: (data: Buffer) => void): void;
112
+ onAudio(callback: (data: Uint8Array) => void): void;
113
113
  /**
114
114
  * Disconnect from the server
115
115
  */
116
116
  disconnect(): void;
117
+ /**
118
+ * Update the system prompt mid-conversation
119
+ */
120
+ updatePrompt(newPrompt: string): void;
121
+ /**
122
+ * Get full conversation transcript
123
+ */
124
+ getTranscript(): Array<{
125
+ role: 'user' | 'agent';
126
+ text: string;
127
+ timestamp: number;
128
+ }>;
129
+ /**
130
+ * Get conversation as formatted text
131
+ */
132
+ getTranscriptText(): string;
117
133
  }
118
134
  /**
119
135
  * Client for standalone Text-to-Speech synthesis
120
136
  */
121
137
  declare class TTSClient {
122
138
  private apiKey;
123
- private serverUrl;
124
- private onAudioCallback?;
125
- private onVisemesCallback?;
126
- private onErrorCallback?;
127
139
  constructor(config: {
128
140
  apiKey: string;
129
- serverUrl?: string;
130
141
  });
131
142
  /**
132
143
  * Synthesize text to speech
@@ -141,7 +152,7 @@ declare class TTSClient {
141
152
  speed?: number;
142
153
  steps?: number;
143
154
  visemes?: boolean;
144
- onAudio?: (data: Buffer) => void;
155
+ onAudio?: (data: Uint8Array) => void;
145
156
  onVisemes?: (visemes: any[]) => void;
146
157
  onError?: (error: any) => void;
147
158
  }): Promise<void>;
@@ -157,7 +168,7 @@ declare function simpleConversation(config: LokutorConfig & {
157
168
  */
158
169
  declare function simpleTTS(options: SynthesizeOptions & {
159
170
  apiKey: string;
160
- onAudio: (buf: Buffer) => void;
171
+ onAudio: (buf: Uint8Array) => void;
161
172
  }): Promise<void>;
162
173
 
163
174
  export { AUDIO_CONFIG, DEFAULT_URLS, Language, type LokutorConfig, type SynthesizeOptions, TTSClient, VoiceAgentClient, VoiceStyle, simpleConversation, simpleTTS };
package/dist/index.d.ts CHANGED
@@ -28,6 +28,7 @@ declare enum Language {
28
28
  */
29
29
  declare const AUDIO_CONFIG: {
30
30
  SAMPLE_RATE: number;
31
+ SPEAKER_SAMPLE_RATE: number;
31
32
  CHANNELS: number;
32
33
  CHUNK_DURATION_MS: number;
33
34
  readonly CHUNK_SIZE: number;
@@ -44,10 +45,9 @@ declare const DEFAULT_URLS: {
44
45
  */
45
46
  interface LokutorConfig {
46
47
  apiKey: string;
47
- serverUrl?: string;
48
48
  onTranscription?: (text: string) => void;
49
49
  onResponse?: (text: string) => void;
50
- onAudio?: (data: Buffer) => void;
50
+ onAudio?: (data: Uint8Array) => void;
51
51
  onStatus?: (status: string) => void;
52
52
  onError?: (error: any) => void;
53
53
  }
@@ -71,7 +71,6 @@ interface SynthesizeOptions {
71
71
  declare class VoiceAgentClient {
72
72
  private ws;
73
73
  private apiKey;
74
- private serverUrl;
75
74
  prompt: string;
76
75
  voice: VoiceStyle;
77
76
  language: Language;
@@ -81,6 +80,7 @@ declare class VoiceAgentClient {
81
80
  private onStatus?;
82
81
  private onError?;
83
82
  private isConnected;
83
+ private messages;
84
84
  constructor(config: LokutorConfig & {
85
85
  prompt: string;
86
86
  voice?: VoiceStyle;
@@ -98,7 +98,7 @@ declare class VoiceAgentClient {
98
98
  * Send raw PCM audio data to the server
99
99
  * @param audioData Int16 PCM audio buffer
100
100
  */
101
- sendAudio(audioData: Buffer | Uint8Array): void;
101
+ sendAudio(audioData: Uint8Array): void;
102
102
  /**
103
103
  * Handle incoming binary data (audio response)
104
104
  */
@@ -109,24 +109,35 @@ declare class VoiceAgentClient {
109
109
  private handleTextMessage;
110
110
  private audioListeners;
111
111
  private emit;
112
- onAudio(callback: (data: Buffer) => void): void;
112
+ onAudio(callback: (data: Uint8Array) => void): void;
113
113
  /**
114
114
  * Disconnect from the server
115
115
  */
116
116
  disconnect(): void;
117
+ /**
118
+ * Update the system prompt mid-conversation
119
+ */
120
+ updatePrompt(newPrompt: string): void;
121
+ /**
122
+ * Get full conversation transcript
123
+ */
124
+ getTranscript(): Array<{
125
+ role: 'user' | 'agent';
126
+ text: string;
127
+ timestamp: number;
128
+ }>;
129
+ /**
130
+ * Get conversation as formatted text
131
+ */
132
+ getTranscriptText(): string;
117
133
  }
118
134
  /**
119
135
  * Client for standalone Text-to-Speech synthesis
120
136
  */
121
137
  declare class TTSClient {
122
138
  private apiKey;
123
- private serverUrl;
124
- private onAudioCallback?;
125
- private onVisemesCallback?;
126
- private onErrorCallback?;
127
139
  constructor(config: {
128
140
  apiKey: string;
129
- serverUrl?: string;
130
141
  });
131
142
  /**
132
143
  * Synthesize text to speech
@@ -141,7 +152,7 @@ declare class TTSClient {
141
152
  speed?: number;
142
153
  steps?: number;
143
154
  visemes?: boolean;
144
- onAudio?: (data: Buffer) => void;
155
+ onAudio?: (data: Uint8Array) => void;
145
156
  onVisemes?: (visemes: any[]) => void;
146
157
  onError?: (error: any) => void;
147
158
  }): Promise<void>;
@@ -157,7 +168,7 @@ declare function simpleConversation(config: LokutorConfig & {
157
168
  */
158
169
  declare function simpleTTS(options: SynthesizeOptions & {
159
170
  apiKey: string;
160
- onAudio: (buf: Buffer) => void;
171
+ onAudio: (buf: Uint8Array) => void;
161
172
  }): Promise<void>;
162
173
 
163
174
  export { AUDIO_CONFIG, DEFAULT_URLS, Language, type LokutorConfig, type SynthesizeOptions, TTSClient, VoiceAgentClient, VoiceStyle, simpleConversation, simpleTTS };
package/dist/index.js CHANGED
@@ -55,6 +55,7 @@ var Language = /* @__PURE__ */ ((Language2) => {
55
55
  })(Language || {});
56
56
  var AUDIO_CONFIG = {
57
57
  SAMPLE_RATE: 44100,
58
+ SPEAKER_SAMPLE_RATE: 16e3,
58
59
  CHANNELS: 1,
59
60
  CHUNK_DURATION_MS: 20,
60
61
  get CHUNK_SIZE() {
@@ -67,11 +68,17 @@ var DEFAULT_URLS = {
67
68
  };
68
69
 
69
70
  // src/client.ts
70
- var import_ws = require("ws");
71
+ function base64ToUint8Array(base64) {
72
+ const binaryString = atob(base64);
73
+ const bytes = new Uint8Array(binaryString.length);
74
+ for (let i = 0; i < binaryString.length; i++) {
75
+ bytes[i] = binaryString.charCodeAt(i);
76
+ }
77
+ return bytes;
78
+ }
71
79
  var VoiceAgentClient = class {
72
80
  ws = null;
73
81
  apiKey;
74
- serverUrl;
75
82
  prompt;
76
83
  voice;
77
84
  language;
@@ -82,9 +89,9 @@ var VoiceAgentClient = class {
82
89
  onStatus;
83
90
  onError;
84
91
  isConnected = false;
92
+ messages = [];
85
93
  constructor(config) {
86
94
  this.apiKey = config.apiKey;
87
- this.serverUrl = config.serverUrl || DEFAULT_URLS.VOICE_AGENT;
88
95
  this.prompt = config.prompt;
89
96
  this.voice = config.voice || "F1" /* F1 */;
90
97
  this.language = config.language || "en" /* ENGLISH */;
@@ -100,36 +107,36 @@ var VoiceAgentClient = class {
100
107
  async connect() {
101
108
  return new Promise((resolve, reject) => {
102
109
  try {
103
- console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
104
- const headers = {};
110
+ let url = DEFAULT_URLS.VOICE_AGENT;
105
111
  if (this.apiKey) {
106
- headers["X-API-Key"] = this.apiKey;
112
+ const separator = url.includes("?") ? "&" : "?";
113
+ url += `${separator}api_key=${this.apiKey}`;
107
114
  }
108
- this.ws = new import_ws.WebSocket(this.serverUrl, {
109
- headers
110
- });
111
- this.ws.on("open", () => {
115
+ console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
116
+ this.ws = new WebSocket(url);
117
+ this.ws.binaryType = "arraybuffer";
118
+ this.ws.onopen = () => {
112
119
  this.isConnected = true;
113
120
  console.log("\u2705 Connected to voice agent!");
114
121
  this.sendConfig();
115
122
  resolve(true);
116
- });
117
- this.ws.on("message", (data, isBinary) => {
118
- if (isBinary) {
119
- this.handleBinaryMessage(data);
123
+ };
124
+ this.ws.onmessage = async (event) => {
125
+ if (event.data instanceof ArrayBuffer) {
126
+ this.handleBinaryMessage(new Uint8Array(event.data));
120
127
  } else {
121
- this.handleTextMessage(data.toString());
128
+ this.handleTextMessage(event.data.toString());
122
129
  }
123
- });
124
- this.ws.on("error", (err) => {
130
+ };
131
+ this.ws.onerror = (err) => {
125
132
  console.error("\u274C WebSocket error:", err);
126
133
  if (this.onError) this.onError(err);
127
134
  if (!this.isConnected) reject(err);
128
- });
129
- this.ws.on("close", () => {
135
+ };
136
+ this.ws.onclose = () => {
130
137
  this.isConnected = false;
131
138
  console.log("Disconnected");
132
- });
139
+ };
133
140
  } catch (err) {
134
141
  if (this.onError) this.onError(err);
135
142
  reject(err);
@@ -152,7 +159,7 @@ var VoiceAgentClient = class {
152
159
  */
153
160
  sendAudio(audioData) {
154
161
  if (this.ws && this.isConnected) {
155
- this.ws.send(audioData, { binary: true });
162
+ this.ws.send(audioData);
156
163
  }
157
164
  }
158
165
  /**
@@ -170,11 +177,17 @@ var VoiceAgentClient = class {
170
177
  switch (msg.type) {
171
178
  case "audio":
172
179
  if (msg.data) {
173
- const buffer = Buffer.from(msg.data, "base64");
180
+ const buffer = base64ToUint8Array(msg.data);
174
181
  this.handleBinaryMessage(buffer);
175
182
  }
176
183
  break;
177
184
  case "transcript":
185
+ const role = msg.role === "user" ? "user" : "agent";
186
+ this.messages.push({
187
+ role,
188
+ text: msg.data,
189
+ timestamp: Date.now()
190
+ });
178
191
  if (msg.role === "user") {
179
192
  if (this.onTranscription) this.onTranscription(msg.data);
180
193
  console.log(`\u{1F4AC} You: ${msg.data}`);
@@ -220,16 +233,39 @@ var VoiceAgentClient = class {
220
233
  this.ws = null;
221
234
  }
222
235
  }
236
+ /**
237
+ * Update the system prompt mid-conversation
238
+ */
239
+ updatePrompt(newPrompt) {
240
+ this.prompt = newPrompt;
241
+ if (this.ws && this.isConnected) {
242
+ try {
243
+ this.ws.send(JSON.stringify({ type: "prompt", data: newPrompt }));
244
+ console.log(`\u2699\uFE0F Updated prompt: ${newPrompt.substring(0, 50)}...`);
245
+ } catch (error) {
246
+ console.error("Error updating prompt:", error);
247
+ }
248
+ } else {
249
+ console.warn("Not connected - prompt will be updated on next connection");
250
+ }
251
+ }
252
+ /**
253
+ * Get full conversation transcript
254
+ */
255
+ getTranscript() {
256
+ return this.messages.slice();
257
+ }
258
+ /**
259
+ * Get conversation as formatted text
260
+ */
261
+ getTranscriptText() {
262
+ return this.messages.map((msg) => `${msg.role === "user" ? "You" : "Agent"}: ${msg.text}`).join("\n");
263
+ }
223
264
  };
224
265
  var TTSClient = class {
225
266
  apiKey;
226
- serverUrl;
227
- onAudioCallback;
228
- onVisemesCallback;
229
- onErrorCallback;
230
267
  constructor(config) {
231
268
  this.apiKey = config.apiKey;
232
- this.serverUrl = config.serverUrl || DEFAULT_URLS.TTS;
233
269
  }
234
270
  /**
235
271
  * Synthesize text to speech
@@ -240,12 +276,14 @@ var TTSClient = class {
240
276
  synthesize(options) {
241
277
  return new Promise((resolve, reject) => {
242
278
  try {
243
- const headers = {};
279
+ let url = DEFAULT_URLS.TTS;
244
280
  if (this.apiKey) {
245
- headers["X-API-Key"] = this.apiKey;
281
+ const separator = url.includes("?") ? "&" : "?";
282
+ url += `${separator}api_key=${this.apiKey}`;
246
283
  }
247
- const ws = new import_ws.WebSocket(this.serverUrl, { headers });
248
- ws.on("open", () => {
284
+ const ws = new WebSocket(url);
285
+ ws.binaryType = "arraybuffer";
286
+ ws.onopen = () => {
249
287
  const req = {
250
288
  text: options.text,
251
289
  voice: options.voice || "F1" /* F1 */,
@@ -255,27 +293,27 @@ var TTSClient = class {
255
293
  visemes: options.visemes || false
256
294
  };
257
295
  ws.send(JSON.stringify(req));
258
- });
259
- ws.on("message", (data, isBinary) => {
260
- if (isBinary) {
261
- if (options.onAudio) options.onAudio(data);
296
+ };
297
+ ws.onmessage = async (event) => {
298
+ if (event.data instanceof ArrayBuffer) {
299
+ if (options.onAudio) options.onAudio(new Uint8Array(event.data));
262
300
  } else {
263
301
  try {
264
- const msg = JSON.parse(data.toString());
302
+ const msg = JSON.parse(event.data.toString());
265
303
  if (Array.isArray(msg) && options.onVisemes) {
266
304
  options.onVisemes(msg);
267
305
  }
268
306
  } catch (e) {
269
307
  }
270
308
  }
271
- });
272
- ws.on("error", (err) => {
309
+ };
310
+ ws.onerror = (err) => {
273
311
  if (options.onError) options.onError(err);
274
312
  reject(err);
275
- });
276
- ws.on("close", () => {
313
+ };
314
+ ws.onclose = () => {
277
315
  resolve();
278
- });
316
+ };
279
317
  } catch (err) {
280
318
  if (options.onError) options.onError(err);
281
319
  reject(err);
package/dist/index.mjs CHANGED
@@ -22,6 +22,7 @@ var Language = /* @__PURE__ */ ((Language2) => {
22
22
  })(Language || {});
23
23
  var AUDIO_CONFIG = {
24
24
  SAMPLE_RATE: 44100,
25
+ SPEAKER_SAMPLE_RATE: 16e3,
25
26
  CHANNELS: 1,
26
27
  CHUNK_DURATION_MS: 20,
27
28
  get CHUNK_SIZE() {
@@ -34,11 +35,17 @@ var DEFAULT_URLS = {
34
35
  };
35
36
 
36
37
  // src/client.ts
37
- import { WebSocket } from "ws";
38
+ function base64ToUint8Array(base64) {
39
+ const binaryString = atob(base64);
40
+ const bytes = new Uint8Array(binaryString.length);
41
+ for (let i = 0; i < binaryString.length; i++) {
42
+ bytes[i] = binaryString.charCodeAt(i);
43
+ }
44
+ return bytes;
45
+ }
38
46
  var VoiceAgentClient = class {
39
47
  ws = null;
40
48
  apiKey;
41
- serverUrl;
42
49
  prompt;
43
50
  voice;
44
51
  language;
@@ -49,9 +56,9 @@ var VoiceAgentClient = class {
49
56
  onStatus;
50
57
  onError;
51
58
  isConnected = false;
59
+ messages = [];
52
60
  constructor(config) {
53
61
  this.apiKey = config.apiKey;
54
- this.serverUrl = config.serverUrl || DEFAULT_URLS.VOICE_AGENT;
55
62
  this.prompt = config.prompt;
56
63
  this.voice = config.voice || "F1" /* F1 */;
57
64
  this.language = config.language || "en" /* ENGLISH */;
@@ -67,36 +74,36 @@ var VoiceAgentClient = class {
67
74
  async connect() {
68
75
  return new Promise((resolve, reject) => {
69
76
  try {
70
- console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
71
- const headers = {};
77
+ let url = DEFAULT_URLS.VOICE_AGENT;
72
78
  if (this.apiKey) {
73
- headers["X-API-Key"] = this.apiKey;
79
+ const separator = url.includes("?") ? "&" : "?";
80
+ url += `${separator}api_key=${this.apiKey}`;
74
81
  }
75
- this.ws = new WebSocket(this.serverUrl, {
76
- headers
77
- });
78
- this.ws.on("open", () => {
82
+ console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
83
+ this.ws = new WebSocket(url);
84
+ this.ws.binaryType = "arraybuffer";
85
+ this.ws.onopen = () => {
79
86
  this.isConnected = true;
80
87
  console.log("\u2705 Connected to voice agent!");
81
88
  this.sendConfig();
82
89
  resolve(true);
83
- });
84
- this.ws.on("message", (data, isBinary) => {
85
- if (isBinary) {
86
- this.handleBinaryMessage(data);
90
+ };
91
+ this.ws.onmessage = async (event) => {
92
+ if (event.data instanceof ArrayBuffer) {
93
+ this.handleBinaryMessage(new Uint8Array(event.data));
87
94
  } else {
88
- this.handleTextMessage(data.toString());
95
+ this.handleTextMessage(event.data.toString());
89
96
  }
90
- });
91
- this.ws.on("error", (err) => {
97
+ };
98
+ this.ws.onerror = (err) => {
92
99
  console.error("\u274C WebSocket error:", err);
93
100
  if (this.onError) this.onError(err);
94
101
  if (!this.isConnected) reject(err);
95
- });
96
- this.ws.on("close", () => {
102
+ };
103
+ this.ws.onclose = () => {
97
104
  this.isConnected = false;
98
105
  console.log("Disconnected");
99
- });
106
+ };
100
107
  } catch (err) {
101
108
  if (this.onError) this.onError(err);
102
109
  reject(err);
@@ -119,7 +126,7 @@ var VoiceAgentClient = class {
119
126
  */
120
127
  sendAudio(audioData) {
121
128
  if (this.ws && this.isConnected) {
122
- this.ws.send(audioData, { binary: true });
129
+ this.ws.send(audioData);
123
130
  }
124
131
  }
125
132
  /**
@@ -137,11 +144,17 @@ var VoiceAgentClient = class {
137
144
  switch (msg.type) {
138
145
  case "audio":
139
146
  if (msg.data) {
140
- const buffer = Buffer.from(msg.data, "base64");
147
+ const buffer = base64ToUint8Array(msg.data);
141
148
  this.handleBinaryMessage(buffer);
142
149
  }
143
150
  break;
144
151
  case "transcript":
152
+ const role = msg.role === "user" ? "user" : "agent";
153
+ this.messages.push({
154
+ role,
155
+ text: msg.data,
156
+ timestamp: Date.now()
157
+ });
145
158
  if (msg.role === "user") {
146
159
  if (this.onTranscription) this.onTranscription(msg.data);
147
160
  console.log(`\u{1F4AC} You: ${msg.data}`);
@@ -187,16 +200,39 @@ var VoiceAgentClient = class {
187
200
  this.ws = null;
188
201
  }
189
202
  }
203
+ /**
204
+ * Update the system prompt mid-conversation
205
+ */
206
+ updatePrompt(newPrompt) {
207
+ this.prompt = newPrompt;
208
+ if (this.ws && this.isConnected) {
209
+ try {
210
+ this.ws.send(JSON.stringify({ type: "prompt", data: newPrompt }));
211
+ console.log(`\u2699\uFE0F Updated prompt: ${newPrompt.substring(0, 50)}...`);
212
+ } catch (error) {
213
+ console.error("Error updating prompt:", error);
214
+ }
215
+ } else {
216
+ console.warn("Not connected - prompt will be updated on next connection");
217
+ }
218
+ }
219
+ /**
220
+ * Get full conversation transcript
221
+ */
222
+ getTranscript() {
223
+ return this.messages.slice();
224
+ }
225
+ /**
226
+ * Get conversation as formatted text
227
+ */
228
+ getTranscriptText() {
229
+ return this.messages.map((msg) => `${msg.role === "user" ? "You" : "Agent"}: ${msg.text}`).join("\n");
230
+ }
190
231
  };
191
232
  var TTSClient = class {
192
233
  apiKey;
193
- serverUrl;
194
- onAudioCallback;
195
- onVisemesCallback;
196
- onErrorCallback;
197
234
  constructor(config) {
198
235
  this.apiKey = config.apiKey;
199
- this.serverUrl = config.serverUrl || DEFAULT_URLS.TTS;
200
236
  }
201
237
  /**
202
238
  * Synthesize text to speech
@@ -207,12 +243,14 @@ var TTSClient = class {
207
243
  synthesize(options) {
208
244
  return new Promise((resolve, reject) => {
209
245
  try {
210
- const headers = {};
246
+ let url = DEFAULT_URLS.TTS;
211
247
  if (this.apiKey) {
212
- headers["X-API-Key"] = this.apiKey;
248
+ const separator = url.includes("?") ? "&" : "?";
249
+ url += `${separator}api_key=${this.apiKey}`;
213
250
  }
214
- const ws = new WebSocket(this.serverUrl, { headers });
215
- ws.on("open", () => {
251
+ const ws = new WebSocket(url);
252
+ ws.binaryType = "arraybuffer";
253
+ ws.onopen = () => {
216
254
  const req = {
217
255
  text: options.text,
218
256
  voice: options.voice || "F1" /* F1 */,
@@ -222,27 +260,27 @@ var TTSClient = class {
222
260
  visemes: options.visemes || false
223
261
  };
224
262
  ws.send(JSON.stringify(req));
225
- });
226
- ws.on("message", (data, isBinary) => {
227
- if (isBinary) {
228
- if (options.onAudio) options.onAudio(data);
263
+ };
264
+ ws.onmessage = async (event) => {
265
+ if (event.data instanceof ArrayBuffer) {
266
+ if (options.onAudio) options.onAudio(new Uint8Array(event.data));
229
267
  } else {
230
268
  try {
231
- const msg = JSON.parse(data.toString());
269
+ const msg = JSON.parse(event.data.toString());
232
270
  if (Array.isArray(msg) && options.onVisemes) {
233
271
  options.onVisemes(msg);
234
272
  }
235
273
  } catch (e) {
236
274
  }
237
275
  }
238
- });
239
- ws.on("error", (err) => {
276
+ };
277
+ ws.onerror = (err) => {
240
278
  if (options.onError) options.onError(err);
241
279
  reject(err);
242
- });
243
- ws.on("close", () => {
280
+ };
281
+ ws.onclose = () => {
244
282
  resolve();
245
- });
283
+ };
246
284
  } catch (err) {
247
285
  if (options.onError) options.onError(err);
248
286
  reject(err);
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@lokutor/sdk",
3
- "version": "1.0.0",
3
+ "version": "1.1.1",
4
4
  "description": "JavaScript/TypeScript SDK for Lokutor Real-time Voice AI",
5
5
  "main": "./dist/index.js",
6
6
  "module": "./dist/index.mjs",
@@ -27,14 +27,11 @@
27
27
  ],
28
28
  "author": "Lokutor AI",
29
29
  "license": "MIT",
30
- "dependencies": {
31
- "ws": "^8.16.0"
32
- },
30
+ "dependencies": {},
33
31
  "devDependencies": {
34
32
  "@types/node": "^20.10.0",
35
- "@types/ws": "^8.5.10",
36
33
  "tsup": "^8.0.1",
37
34
  "typescript": "^5.3.2",
38
35
  "vitest": "^1.0.1"
39
36
  }
40
- }
37
+ }