@cogitator-ai/voice 0.1.12 → 0.2.0

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 (73) hide show
  1. package/README.md +121 -77
  2. package/dist/audio.d.ts +7 -0
  3. package/dist/audio.d.ts.map +1 -1
  4. package/dist/audio.js +46 -3
  5. package/dist/audio.js.map +1 -1
  6. package/dist/cogitator-runner.d.ts +34 -0
  7. package/dist/cogitator-runner.d.ts.map +1 -0
  8. package/dist/cogitator-runner.js +28 -0
  9. package/dist/cogitator-runner.js.map +1 -0
  10. package/dist/index.d.ts +6 -5
  11. package/dist/index.d.ts.map +1 -1
  12. package/dist/index.js +7 -4
  13. package/dist/index.js.map +1 -1
  14. package/dist/pipeline/index.d.ts +2 -2
  15. package/dist/pipeline/index.d.ts.map +1 -1
  16. package/dist/pipeline/index.js.map +1 -1
  17. package/dist/pipeline/pipeline-session.d.ts +22 -3
  18. package/dist/pipeline/pipeline-session.d.ts.map +1 -1
  19. package/dist/pipeline/pipeline-session.js +189 -61
  20. package/dist/pipeline/pipeline-session.js.map +1 -1
  21. package/dist/pipeline/voice-pipeline.d.ts +12 -7
  22. package/dist/pipeline/voice-pipeline.d.ts.map +1 -1
  23. package/dist/pipeline/voice-pipeline.js +14 -4
  24. package/dist/pipeline/voice-pipeline.js.map +1 -1
  25. package/dist/realtime/gemini-realtime.d.ts +21 -1
  26. package/dist/realtime/gemini-realtime.d.ts.map +1 -1
  27. package/dist/realtime/gemini-realtime.js +173 -65
  28. package/dist/realtime/gemini-realtime.js.map +1 -1
  29. package/dist/realtime/openai-realtime.d.ts +12 -1
  30. package/dist/realtime/openai-realtime.d.ts.map +1 -1
  31. package/dist/realtime/openai-realtime.js +182 -53
  32. package/dist/realtime/openai-realtime.js.map +1 -1
  33. package/dist/realtime/realtime-session.d.ts +14 -0
  34. package/dist/realtime/realtime-session.d.ts.map +1 -1
  35. package/dist/realtime/realtime-session.js +70 -4
  36. package/dist/realtime/realtime-session.js.map +1 -1
  37. package/dist/stt/deepgram-stt.d.ts +3 -0
  38. package/dist/stt/deepgram-stt.d.ts.map +1 -1
  39. package/dist/stt/deepgram-stt.js +128 -49
  40. package/dist/stt/deepgram-stt.js.map +1 -1
  41. package/dist/stt/openai-stt.d.ts +5 -0
  42. package/dist/stt/openai-stt.d.ts.map +1 -1
  43. package/dist/stt/openai-stt.js +75 -10
  44. package/dist/stt/openai-stt.js.map +1 -1
  45. package/dist/tools.d.ts +8 -0
  46. package/dist/tools.d.ts.map +1 -1
  47. package/dist/tools.js +12 -4
  48. package/dist/tools.js.map +1 -1
  49. package/dist/transport/index.d.ts +1 -0
  50. package/dist/transport/index.d.ts.map +1 -1
  51. package/dist/transport/ws-transport.d.ts +14 -0
  52. package/dist/transport/ws-transport.d.ts.map +1 -1
  53. package/dist/transport/ws-transport.js +119 -15
  54. package/dist/transport/ws-transport.js.map +1 -1
  55. package/dist/tts/elevenlabs-tts.d.ts +1 -0
  56. package/dist/tts/elevenlabs-tts.d.ts.map +1 -1
  57. package/dist/tts/elevenlabs-tts.js +36 -17
  58. package/dist/tts/elevenlabs-tts.js.map +1 -1
  59. package/dist/tts/openai-tts.d.ts +1 -0
  60. package/dist/tts/openai-tts.d.ts.map +1 -1
  61. package/dist/tts/openai-tts.js +23 -9
  62. package/dist/tts/openai-tts.js.map +1 -1
  63. package/dist/types.d.ts +39 -16
  64. package/dist/types.d.ts.map +1 -1
  65. package/dist/vad/silero-vad.d.ts +15 -0
  66. package/dist/vad/silero-vad.d.ts.map +1 -1
  67. package/dist/vad/silero-vad.js +109 -20
  68. package/dist/vad/silero-vad.js.map +1 -1
  69. package/dist/voice-agent.d.ts +14 -0
  70. package/dist/voice-agent.d.ts.map +1 -1
  71. package/dist/voice-agent.js +117 -28
  72. package/dist/voice-agent.js.map +1 -1
  73. package/package.json +5 -5
@@ -1,17 +1,31 @@
1
1
  import { EventEmitter } from 'node:events';
2
2
  import { WebSocket } from 'ws';
3
+ import { audioMimeType, detectAudioFormat } from '../audio.js';
3
4
  const DEFAULT_MODEL = 'nova-3';
5
+ const DEFAULT_SAMPLE_RATE = 16000;
4
6
  const BASE_URL = 'https://api.deepgram.com/v1/listen';
5
7
  const WS_URL = 'wss://api.deepgram.com/v1/listen';
8
+ function mapWords(words) {
9
+ if (!words || words.length === 0)
10
+ return undefined;
11
+ return words.map((w) => ({
12
+ word: w.word,
13
+ start: w.start,
14
+ end: w.end,
15
+ confidence: w.confidence,
16
+ }));
17
+ }
6
18
  export class DeepgramSTT {
7
19
  name = 'deepgram';
8
20
  apiKey;
9
21
  model;
10
22
  language;
23
+ sampleRate;
11
24
  constructor(config) {
12
25
  this.apiKey = config.apiKey;
13
26
  this.model = config.model ?? DEFAULT_MODEL;
14
27
  this.language = config.language;
28
+ this.sampleRate = config.sampleRate ?? DEFAULT_SAMPLE_RATE;
15
29
  }
16
30
  async transcribe(audio, options) {
17
31
  const params = new URLSearchParams({
@@ -21,13 +35,25 @@ export class DeepgramSTT {
21
35
  const lang = options?.language ?? this.language;
22
36
  if (lang)
23
37
  params.set('language', lang);
38
+ const format = detectAudioFormat(audio);
39
+ let contentType;
40
+ if (format) {
41
+ contentType = audioMimeType(format);
42
+ }
43
+ else {
44
+ contentType = 'application/octet-stream';
45
+ params.set('encoding', 'linear16');
46
+ params.set('sample_rate', String(this.sampleRate));
47
+ params.set('channels', '1');
48
+ }
24
49
  const response = await fetch(`${BASE_URL}?${params}`, {
25
50
  method: 'POST',
26
51
  headers: {
27
52
  Authorization: `Token ${this.apiKey}`,
28
- 'Content-Type': 'audio/wav',
53
+ 'Content-Type': contentType,
29
54
  },
30
55
  body: new Uint8Array(audio),
56
+ signal: AbortSignal.timeout(30_000),
31
57
  });
32
58
  if (!response.ok) {
33
59
  const body = await response.text();
@@ -37,46 +63,55 @@ export class DeepgramSTT {
37
63
  return this.mapBatchResponse(data);
38
64
  }
39
65
  createStream(options) {
40
- return new DeepgramSTTStream(this.apiKey, this.model, this.language, options);
66
+ return new DeepgramSTTStream(this.apiKey, this.model, this.language, {
67
+ ...options,
68
+ sampleRate: options?.sampleRate ?? this.sampleRate,
69
+ });
41
70
  }
42
71
  mapBatchResponse(data) {
43
- const alt = data.results.channels[0]?.alternatives[0];
72
+ const alt = data.results?.channels?.[0]?.alternatives?.[0];
44
73
  const result = {
45
74
  text: alt?.transcript ?? '',
46
75
  };
47
76
  if (data.metadata?.duration !== undefined) {
48
77
  result.duration = data.metadata.duration;
49
78
  }
50
- if (alt?.words && alt.words.length > 0) {
51
- result.words = alt.words.map((w) => ({
52
- word: w.word,
53
- start: w.start,
54
- end: w.end,
55
- confidence: w.confidence,
56
- }));
57
- }
79
+ const words = mapWords(alt?.words);
80
+ if (words)
81
+ result.words = words;
58
82
  return result;
59
83
  }
60
84
  }
85
+ const CLOSE_TIMEOUT_MS = 10_000;
61
86
  class DeepgramSTTStream extends EventEmitter {
62
87
  ws;
63
88
  ready = false;
89
+ closed = false;
90
+ remoteClosed = false;
91
+ closePromise = null;
64
92
  pendingChunks = [];
65
- lastResult = { text: '' };
93
+ segments = [];
94
+ words = [];
95
+ audioEnd = 0;
66
96
  constructor(apiKey, model, language, options) {
67
97
  super();
68
98
  const params = new URLSearchParams({
69
99
  model,
70
100
  punctuate: 'true',
71
- interim_results: 'true',
101
+ interim_results: String(options.interimResults ?? true),
102
+ encoding: 'linear16',
103
+ sample_rate: String(options.sampleRate),
104
+ channels: '1',
72
105
  });
73
- const lang = options?.language ?? language;
106
+ const lang = options.language ?? language;
74
107
  if (lang)
75
108
  params.set('language', lang);
76
- if (options?.endpointing !== undefined) {
109
+ if (options.endpointing !== undefined) {
77
110
  params.set('endpointing', String(options.endpointing));
78
111
  }
79
- this.ws = new WebSocket(`${WS_URL}?${params}`, [`token:${apiKey}`]);
112
+ this.ws = new WebSocket(`${WS_URL}?${params}`, {
113
+ headers: { Authorization: `Token ${apiKey}` },
114
+ });
80
115
  this.ws.on('open', () => {
81
116
  this.ready = true;
82
117
  for (const chunk of this.pendingChunks) {
@@ -92,26 +127,56 @@ class DeepgramSTTStream extends EventEmitter {
92
127
  catch {
93
128
  return;
94
129
  }
95
- const alt = msg.channel?.alternatives[0];
96
- if (!alt?.transcript)
130
+ if (msg.type !== undefined && msg.type !== 'Results')
131
+ return;
132
+ const alt = msg.channel?.alternatives?.[0];
133
+ if (!alt)
97
134
  return;
98
135
  if (msg.is_final) {
99
- const result = this.mapStreamResult(msg);
100
- this.lastResult = result;
101
- this.emit('final', result);
136
+ if (typeof msg.start === 'number' && typeof msg.duration === 'number') {
137
+ this.audioEnd = Math.max(this.audioEnd, msg.start + msg.duration);
138
+ }
139
+ if (!alt.transcript)
140
+ return;
141
+ this.segments.push(alt.transcript);
142
+ const words = mapWords(alt.words);
143
+ if (words)
144
+ this.words.push(...words);
145
+ const segment = { text: alt.transcript };
146
+ if (words)
147
+ segment.words = words;
148
+ if (typeof msg.duration === 'number')
149
+ segment.duration = msg.duration;
150
+ this.emit('final', segment);
102
151
  }
103
- else {
152
+ else if (alt.transcript) {
104
153
  this.emit('partial', alt.transcript);
105
154
  }
106
155
  });
107
156
  this.ws.on('error', (err) => {
108
- this.emit('error', err);
157
+ if (this.listenerCount('error') > 0)
158
+ this.emit('error', err);
109
159
  });
110
- this.ws.on('close', () => {
160
+ this.ws.on('close', (code, reason) => {
111
161
  this.ready = false;
162
+ if (!this.closed) {
163
+ this.remoteClosed = true;
164
+ const dropped = this.pendingChunks.length;
165
+ this.pendingChunks = [];
166
+ if (this.listenerCount('error') > 0) {
167
+ const detail = reason.toString() || 'no reason';
168
+ this.emit('error', new Error(`DeepgramSTTStream: connection closed unexpectedly (code ${code}: ${detail})` +
169
+ (dropped > 0 ? `, ${dropped} chunk(s) dropped` : '')));
170
+ }
171
+ }
112
172
  });
113
173
  }
114
174
  write(chunk) {
175
+ if (this.closed) {
176
+ throw new Error('DeepgramSTTStream: cannot write after close');
177
+ }
178
+ if (this.remoteClosed)
179
+ return;
115
180
  if (this.ready) {
116
181
  this.ws.send(chunk);
117
182
  }
@@ -120,33 +185,47 @@ class DeepgramSTTStream extends EventEmitter {
120
185
  }
121
186
  }
122
187
  async close() {
123
- if (!this.ready) {
124
- this.ws.close();
125
- return this.lastResult;
126
- }
127
- return new Promise((resolve) => {
128
- this.ws.on('close', () => resolve(this.lastResult));
129
- this.ws.send(JSON.stringify({ type: 'CloseStream' }));
188
+ if (this.closePromise)
189
+ return this.closePromise;
190
+ this.closed = true;
191
+ this.closePromise = new Promise((resolve) => {
192
+ const state = this.ws.readyState;
193
+ if (state === WebSocket.CLOSED || state === WebSocket.CLOSING) {
194
+ resolve(this.buildResult());
195
+ return;
196
+ }
197
+ const timeout = setTimeout(() => {
198
+ this.ws.terminate();
199
+ resolve(this.buildResult());
200
+ }, CLOSE_TIMEOUT_MS);
201
+ this.ws.once('close', () => {
202
+ clearTimeout(timeout);
203
+ resolve(this.buildResult());
204
+ });
205
+ if (state === WebSocket.OPEN) {
206
+ this.ws.send(JSON.stringify({ type: 'CloseStream' }));
207
+ return;
208
+ }
209
+ if (this.pendingChunks.length === 0) {
210
+ clearTimeout(timeout);
211
+ this.ws.removeAllListeners('close');
212
+ this.ws.on('error', () => { });
213
+ this.ws.terminate();
214
+ resolve(this.buildResult());
215
+ return;
216
+ }
217
+ this.ws.once('open', () => {
218
+ this.ws.send(JSON.stringify({ type: 'CloseStream' }));
219
+ });
130
220
  });
221
+ return this.closePromise;
131
222
  }
132
- mapStreamResult(msg) {
133
- const alt = msg.channel.alternatives[0];
134
- if (!alt)
135
- return { text: '' };
136
- const result = {
137
- text: alt.transcript,
138
- };
139
- if (msg.metadata?.duration !== undefined) {
140
- result.duration = msg.metadata.duration;
141
- }
142
- if (alt.words && alt.words.length > 0) {
143
- result.words = alt.words.map((w) => ({
144
- word: w.word,
145
- start: w.start,
146
- end: w.end,
147
- confidence: w.confidence,
148
- }));
149
- }
223
+ buildResult() {
224
+ const result = { text: this.segments.join(' ').trim() };
225
+ if (this.audioEnd > 0)
226
+ result.duration = this.audioEnd;
227
+ if (this.words.length > 0)
228
+ result.words = [...this.words];
150
229
  return result;
151
230
  }
152
231
  }
@@ -1 +1 @@
1
- {"version":3,"file":"deepgram-stt.js","sourceRoot":"","sources":["../../src/stt/deepgram-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAe/B,MAAM,aAAa,GAAG,QAAQ,CAAC;AAC/B,MAAM,QAAQ,GAAG,oCAAoC,CAAC;AACtD,MAAM,MAAM,GAAG,kCAAkC,CAAC;AA8BlD,MAAM,OAAO,WAAW;IACb,IAAI,GAAG,UAAU,CAAC;IAEV,MAAM,CAAS;IACf,KAAK,CAAS;IACd,QAAQ,CAAU;IAEnC,YAAY,MAAyB;QACnC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;QAC3C,IAAI,CAAC,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC;IAClC,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,KAAa,EAAE,OAAoB;QAClD,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC;YACjC,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,MAAM;SAClB,CAAC,CAAC;QAEH,MAAM,IAAI,GAAG,OAAO,EAAE,QAAQ,IAAI,IAAI,CAAC,QAAQ,CAAC;QAChD,IAAI,IAAI;YAAE,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QAEvC,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,QAAQ,IAAI,MAAM,EAAE,EAAE;YACpD,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,SAAS,IAAI,CAAC,MAAM,EAAE;gBACrC,cAAc,EAAE,WAAW;aAC5B;YACD,IAAI,EAAE,IAAI,UAAU,CAAC,KAAK,CAAC;SAC5B,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnC,MAAM,IAAI,KAAK,CAAC,sBAAsB,QAAQ,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC,CAAC;QACpE,CAAC;QAED,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA0B,CAAC;QAC9D,OAAO,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC;IACrC,CAAC;IAED,YAAY,CAAC,OAA0B;QACrC,OAAO,IAAI,iBAAiB,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAC;IAChF,CAAC;IAEO,gBAAgB,CAAC,IAA2B;QAClD,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,YAAY,CAAC,CAAC,CAAC,CAAC;QAEtD,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,GAAG,EAAE,UAAU,IAAI,EAAE;SAC5B,CAAC;QAEF,IAAI,IAAI,CAAC,QAAQ,EAAE,QAAQ,KAAK,SAAS,EAAE,CAAC;YAC1C,MAAM,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC;QAC3C,CAAC;QAED,IAAI,GAAG,EAAE,KAAK,IAAI,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACvC,MAAM,CAAC,KAAK,GAAG,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBACnC,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,KAAK,EAAE,CAAC,CAAC,KAAK;gBACd,GAAG,EAAE,CAAC,CAAC,GAAG;gBACV,UAAU,EAAE,CAAC,CAAC,UAAU;aACzB,CAAC,CAAC,CAAC;QACN,CAAC;QAED,OAAO,MAAM,CAAC;IAChB,CAAC;CACF;AAED,MAAM,iBAAkB,SAAQ,YAAY;IAClC,EAAE,CAAY;IACd,KAAK,GAAG,KAAK,CAAC;IACd,aAAa,GAAa,EAAE,CAAC;IAC7B,UAAU,GAAqB,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;IAEpD,YACE,MAAc,EACd,KAAa,EACb,QAA4B,EAC5B,OAA0B;QAE1B,KAAK,EAAE,CAAC;QAER,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC;YACjC,KAAK;YACL,SAAS,EAAE,MAAM;YACjB,eAAe,EAAE,MAAM;SACxB,CAAC,CAAC;QAEH,MAAM,IAAI,GAAG,OAAO,EAAE,QAAQ,IAAI,QAAQ,CAAC;QAC3C,IAAI,IAAI;YAAE,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QACvC,IAAI,OAAO,EAAE,WAAW,KAAK,SAAS,EAAE,CAAC;YACvC,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,CAAC,OAAO,CAAC,WAAW,CAAC,CAAC,CAAC;QACzD,CAAC;QAED,IAAI,CAAC,EAAE,GAAG,IAAI,SAAS,CAAC,GAAG,MAAM,IAAI,MAAM,EAAE,EAAE,CAAC,SAAS,MAAM,EAAE,CAAC,CAAC,CAAC;QAEpE,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;YACtB,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;YAClB,KAAK,MAAM,KAAK,IAAI,IAAI,CAAC,aAAa,EAAE,CAAC;gBACvC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACtB,CAAC;YACD,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;QAC1B,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAoB,EAAE,EAAE;YAC7C,IAAI,GAA0B,CAAC;YAC/B,IAAI,CAAC;gBACH,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAA0B,CAAC;YACzD,CAAC;YAAC,MAAM,CAAC;gBACP,OAAO;YACT,CAAC;YACD,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,EAAE,YAAY,CAAC,CAAC,CAAC,CAAC;YACzC,IAAI,CAAC,GAAG,EAAE,UAAU;gBAAE,OAAO;YAE7B,IAAI,GAAG,CAAC,QAAQ,EAAE,CAAC;gBACjB,MAAM,MAAM,GAAG,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC;gBACzC,IAAI,CAAC,UAAU,GAAG,MAAM,CAAC;gBACzB,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;YAC7B,CAAC;iBAAM,CAAC;gBACN,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,GAAG,CAAC,UAAU,CAAC,CAAC;YACvC,CAAC;QACH,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;YACjC,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;QAC1B,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YACvB,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC;QACrB,CAAC,CAAC,CAAC;IACL,CAAC;IAED,KAAK,CAAC,KAAa;QACjB,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;YACf,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACtB,CAAC;aAAM,CAAC;YACN,IAAI,CAAC,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACjC,CAAC;IACH,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE,CAAC;YAChB,IAAI,CAAC,EAAE,CAAC,KAAK,EAAE,CAAC;YAChB,OAAO,IAAI,CAAC,UAAU,CAAC;QACzB,CAAC;QAED,OAAO,IAAI,OAAO,CAAmB,CAAC,OAAO,EAAE,EAAE;YAC/C,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;YACpD,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;QACxD,CAAC,CAAC,CAAC;IACL,CAAC;IAEO,eAAe,CAAC,GAA0B;QAChD,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC;QACxC,IAAI,CAAC,GAAG;YAAE,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;QAC9B,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,GAAG,CAAC,UAAU;SACrB,CAAC;QAEF,IAAI,GAAG,CAAC,QAAQ,EAAE,QAAQ,KAAK,SAAS,EAAE,CAAC;YACzC,MAAM,CAAC,QAAQ,GAAG,GAAG,CAAC,QAAQ,CAAC,QAAQ,CAAC;QAC1C,CAAC;QAED,IAAI,GAAG,CAAC,KAAK,IAAI,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACtC,MAAM,CAAC,KAAK,GAAG,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBACnC,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,KAAK,EAAE,CAAC,CAAC,KAAK;gBACd,GAAG,EAAE,CAAC,CAAC,GAAG;gBACV,UAAU,EAAE,CAAC,CAAC,UAAU;aACzB,CAAC,CAAC,CAAC;QACN,CAAC;QAED,OAAO,MAAM,CAAC;IAChB,CAAC;CACF"}
1
+ {"version":3,"file":"deepgram-stt.js","sourceRoot":"","sources":["../../src/stt/deepgram-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAC/B,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAiB/D,MAAM,aAAa,GAAG,QAAQ,CAAC;AAC/B,MAAM,mBAAmB,GAAG,KAAK,CAAC;AAClC,MAAM,QAAQ,GAAG,oCAAoC,CAAC;AACtD,MAAM,MAAM,GAAG,kCAAkC,CAAC;AAgClD,SAAS,QAAQ,CAAC,KAAiC;IACjD,IAAI,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IACnD,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;QACvB,IAAI,EAAE,CAAC,CAAC,IAAI;QACZ,KAAK,EAAE,CAAC,CAAC,KAAK;QACd,GAAG,EAAE,CAAC,CAAC,GAAG;QACV,UAAU,EAAE,CAAC,CAAC,UAAU;KACzB,CAAC,CAAC,CAAC;AACN,CAAC;AAED,MAAM,OAAO,WAAW;IACb,IAAI,GAAG,UAAU,CAAC;IAEV,MAAM,CAAS;IACf,KAAK,CAAS;IACd,QAAQ,CAAU;IAClB,UAAU,CAAS;IAEpC,YAAY,MAAyB;QACnC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;QAC3C,IAAI,CAAC,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC;QAChC,IAAI,CAAC,UAAU,GAAG,MAAM,CAAC,UAAU,IAAI,mBAAmB,CAAC;IAC7D,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,KAAa,EAAE,OAAoB;QAClD,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC;YACjC,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,MAAM;SAClB,CAAC,CAAC;QAEH,MAAM,IAAI,GAAG,OAAO,EAAE,QAAQ,IAAI,IAAI,CAAC,QAAQ,CAAC;QAChD,IAAI,IAAI;YAAE,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QAEvC,MAAM,MAAM,GAAG,iBAAiB,CAAC,KAAK,CAAC,CAAC;QACxC,IAAI,WAAmB,CAAC;QACxB,IAAI,MAAM,EAAE,CAAC;YACX,WAAW,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;QACtC,CAAC;aAAM,CAAC;YACN,WAAW,GAAG,0BAA0B,CAAC;YACzC,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,UAAU,CAAC,CAAC;YACnC,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;YACnD,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,GAAG,CAAC,CAAC;QAC9B,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,QAAQ,IAAI,MAAM,EAAE,EAAE;YACpD,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,SAAS,IAAI,CAAC,MAAM,EAAE;gBACrC,cAAc,EAAE,WAAW;aAC5B;YACD,IAAI,EAAE,IAAI,UAAU,CAAC,KAAK,CAAC;YAC3B,MAAM,EAAE,WAAW,CAAC,OAAO,CAAC,MAAM,CAAC;SACpC,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnC,MAAM,IAAI,KAAK,CAAC,sBAAsB,QAAQ,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC,CAAC;QACpE,CAAC;QAED,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA0B,CAAC;QAC9D,OAAO,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC;IACrC,CAAC;IAED,YAAY,CAAC,OAA0B;QACrC,OAAO,IAAI,iBAAiB,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,QAAQ,EAAE;YACnE,GAAG,OAAO;YACV,UAAU,EAAE,OAAO,EAAE,UAAU,IAAI,IAAI,CAAC,UAAU;SACnD,CAAC,CAAC;IACL,CAAC;IAEO,gBAAgB,CAAC,IAA2B;QAClD,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,YAAY,EAAE,CAAC,CAAC,CAAC,CAAC;QAE3D,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,GAAG,EAAE,UAAU,IAAI,EAAE;SAC5B,CAAC;QAEF,IAAI,IAAI,CAAC,QAAQ,EAAE,QAAQ,KAAK,SAAS,EAAE,CAAC;YAC1C,MAAM,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC;QAC3C,CAAC;QAED,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;QACnC,IAAI,KAAK;YAAE,MAAM,CAAC,KAAK,GAAG,KAAK,CAAC;QAEhC,OAAO,MAAM,CAAC;IAChB,CAAC;CACF;AAED,MAAM,gBAAgB,GAAG,MAAM,CAAC;AAEhC,MAAM,iBAAkB,SAAQ,YAAY;IAClC,EAAE,CAAY;IACd,KAAK,GAAG,KAAK,CAAC;IACd,MAAM,GAAG,KAAK,CAAC;IACf,YAAY,GAAG,KAAK,CAAC;IACrB,YAAY,GAAqC,IAAI,CAAC;IACtD,aAAa,GAAa,EAAE,CAAC;IAC7B,QAAQ,GAAa,EAAE,CAAC;IACxB,KAAK,GAA2C,EAAE,CAAC;IACnD,QAAQ,GAAG,CAAC,CAAC;IAErB,YACE,MAAc,EACd,KAAa,EACb,QAA4B,EAC5B,OAAkD;QAElD,KAAK,EAAE,CAAC;QAER,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC;YACjC,KAAK;YACL,SAAS,EAAE,MAAM;YACjB,eAAe,EAAE,MAAM,CAAC,OAAO,CAAC,cAAc,IAAI,IAAI,CAAC;YACvD,QAAQ,EAAE,UAAU;YACpB,WAAW,EAAE,MAAM,CAAC,OAAO,CAAC,UAAU,CAAC;YACvC,QAAQ,EAAE,GAAG;SACd,CAAC,CAAC;QAEH,MAAM,IAAI,GAAG,OAAO,CAAC,QAAQ,IAAI,QAAQ,CAAC;QAC1C,IAAI,IAAI;YAAE,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QACvC,IAAI,OAAO,CAAC,WAAW,KAAK,SAAS,EAAE,CAAC;YACtC,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,CAAC,OAAO,CAAC,WAAW,CAAC,CAAC,CAAC;QACzD,CAAC;QAED,IAAI,CAAC,EAAE,GAAG,IAAI,SAAS,CAAC,GAAG,MAAM,IAAI,MAAM,EAAE,EAAE;YAC7C,OAAO,EAAE,EAAE,aAAa,EAAE,SAAS,MAAM,EAAE,EAAE;SAC9C,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;YACtB,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;YAClB,KAAK,MAAM,KAAK,IAAI,IAAI,CAAC,aAAa,EAAE,CAAC;gBACvC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACtB,CAAC;YACD,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;QAC1B,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAsB,EAAE,EAAE;YAC/C,IAAI,GAA0B,CAAC;YAC/B,IAAI,CAAC;gBACH,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAA0B,CAAC;YACzD,CAAC;YAAC,MAAM,CAAC;gBACP,OAAO;YACT,CAAC;YACD,IAAI,GAAG,CAAC,IAAI,KAAK,SAAS,IAAI,GAAG,CAAC,IAAI,KAAK,SAAS;gBAAE,OAAO;YAE7D,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,EAAE,YAAY,EAAE,CAAC,CAAC,CAAC,CAAC;YAC3C,IAAI,CAAC,GAAG;gBAAE,OAAO;YAEjB,IAAI,GAAG,CAAC,QAAQ,EAAE,CAAC;gBACjB,IAAI,OAAO,GAAG,CAAC,KAAK,KAAK,QAAQ,IAAI,OAAO,GAAG,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;oBACtE,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,GAAG,GAAG,CAAC,QAAQ,CAAC,CAAC;gBACpE,CAAC;gBACD,IAAI,CAAC,GAAG,CAAC,UAAU;oBAAE,OAAO;gBAC5B,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,UAAU,CAAC,CAAC;gBACnC,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;gBAClC,IAAI,KAAK;oBAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,KAAK,CAAC,CAAC;gBAErC,MAAM,OAAO,GAAqB,EAAE,IAAI,EAAE,GAAG,CAAC,UAAU,EAAE,CAAC;gBAC3D,IAAI,KAAK;oBAAE,OAAO,CAAC,KAAK,GAAG,KAAK,CAAC;gBACjC,IAAI,OAAO,GAAG,CAAC,QAAQ,KAAK,QAAQ;oBAAE,OAAO,CAAC,QAAQ,GAAG,GAAG,CAAC,QAAQ,CAAC;gBACtE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;YAC9B,CAAC;iBAAM,IAAI,GAAG,CAAC,UAAU,EAAE,CAAC;gBAC1B,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,GAAG,CAAC,UAAU,CAAC,CAAC;YACvC,CAAC;QACH,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;YACjC,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC;gBAAE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;QAC/D,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,IAAY,EAAE,MAAc,EAAE,EAAE;YACnD,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC;YACnB,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;gBACjB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC;gBACzB,MAAM,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC;gBAC1C,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;gBACxB,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC;oBACpC,MAAM,MAAM,GAAG,MAAM,CAAC,QAAQ,EAAE,IAAI,WAAW,CAAC;oBAChD,IAAI,CAAC,IAAI,CACP,OAAO,EACP,IAAI,KAAK,CACP,2DAA2D,IAAI,KAAK,MAAM,GAAG;wBAC3E,CAAC,OAAO,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,OAAO,mBAAmB,CAAC,CAAC,CAAC,EAAE,CAAC,CACvD,CACF,CAAC;gBACJ,CAAC;YACH,CAAC;QACH,CAAC,CAAC,CAAC;IACL,CAAC;IAED,KAAK,CAAC,KAAa;QACjB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,MAAM,IAAI,KAAK,CAAC,6CAA6C,CAAC,CAAC;QACjE,CAAC;QACD,IAAI,IAAI,CAAC,YAAY;YAAE,OAAO;QAC9B,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;YACf,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACtB,CAAC;aAAM,CAAC;YACN,IAAI,CAAC,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACjC,CAAC;IACH,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,IAAI,CAAC,YAAY;YAAE,OAAO,IAAI,CAAC,YAAY,CAAC;QAChD,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;QAEnB,IAAI,CAAC,YAAY,GAAG,IAAI,OAAO,CAAmB,CAAC,OAAO,EAAE,EAAE;YAC5D,MAAM,KAAK,GAAG,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC;YACjC,IAAI,KAAK,KAAK,SAAS,CAAC,MAAM,IAAI,KAAK,KAAK,SAAS,CAAC,OAAO,EAAE,CAAC;gBAC9D,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;gBAC5B,OAAO;YACT,CAAC;YAED,MAAM,OAAO,GAAG,UAAU,CAAC,GAAG,EAAE;gBAC9B,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC;gBACpB,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;YAC9B,CAAC,EAAE,gBAAgB,CAAC,CAAC;YAErB,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,EAAE;gBACzB,YAAY,CAAC,OAAO,CAAC,CAAC;gBACtB,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;YAC9B,CAAC,CAAC,CAAC;YAEH,IAAI,KAAK,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;gBAC7B,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;gBACtD,OAAO;YACT,CAAC;YAED,IAAI,IAAI,CAAC,aAAa,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;gBACpC,YAAY,CAAC,OAAO,CAAC,CAAC;gBACtB,IAAI,CAAC,EAAE,CAAC,kBAAkB,CAAC,OAAO,CAAC,CAAC;gBACpC,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;gBAC9B,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC;gBACpB,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;gBAC5B,OAAO;YACT,CAAC;YAED,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,EAAE;gBACxB,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;YACxD,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;QAEH,OAAO,IAAI,CAAC,YAAY,CAAC;IAC3B,CAAC;IAEO,WAAW;QACjB,MAAM,MAAM,GAAqB,EAAE,IAAI,EAAE,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC;QAC1E,IAAI,IAAI,CAAC,QAAQ,GAAG,CAAC;YAAE,MAAM,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC;QACvD,IAAI,IAAI,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC;YAAE,MAAM,CAAC,KAAK,GAAG,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC;QAC1D,OAAO,MAAM,CAAC;IAChB,CAAC;CACF"}
@@ -9,8 +9,13 @@ export declare class OpenAISTT implements STTProvider {
9
9
  private readonly client;
10
10
  private readonly model;
11
11
  constructor(config: OpenAISTTConfig);
12
+ /**
13
+ * Transcribe a complete audio file (wav, mp3, ogg, flac, webm, mp4). Headerless input is
14
+ * treated as raw PCM16 mono at 16kHz and wrapped into a WAV container.
15
+ */
12
16
  transcribe(audio: Buffer, options?: STTOptions): Promise<TranscribeResult>;
13
17
  createStream(options?: STTStreamOptions): STTStream;
18
+ private transcribeFile;
14
19
  private mapResponse;
15
20
  }
16
21
  //# sourceMappingURL=openai-stt.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"openai-stt.d.ts","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EACV,WAAW,EACX,UAAU,EACV,gBAAgB,EAChB,SAAS,EACT,gBAAgB,EACjB,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAID,qBAAa,SAAU,YAAW,WAAW;IAC3C,QAAQ,CAAC,IAAI,YAAY;IAEzB,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;gBAEnB,MAAM,EAAE,eAAe;IAQ7B,UAAU,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,gBAAgB,CAAC;IAehF,YAAY,CAAC,OAAO,CAAC,EAAE,gBAAgB,GAAG,SAAS;IAInD,OAAO,CAAC,WAAW;CAwBpB"}
1
+ {"version":3,"file":"openai-stt.d.ts","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EACV,WAAW,EACX,UAAU,EACV,gBAAgB,EAChB,SAAS,EACT,gBAAgB,EACjB,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AA0BD,qBAAa,SAAU,YAAW,WAAW;IAC3C,QAAQ,CAAC,IAAI,YAAY;IAEzB,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;gBAEnB,MAAM,EAAE,eAAe;IAQnC;;;OAGG;IACG,UAAU,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,gBAAgB,CAAC;IAIhF,YAAY,CAAC,OAAO,CAAC,EAAE,gBAAgB,GAAG,SAAS;YAYrC,cAAc;IA0B5B,OAAO,CAAC,WAAW;CAwBpB"}
@@ -1,6 +1,28 @@
1
1
  import { EventEmitter } from 'node:events';
2
2
  import OpenAI from 'openai';
3
+ import { audioMimeType, detectAudioFormat, pcmToWav } from '../audio.js';
3
4
  const DEFAULT_MODEL = 'gpt-4o-mini-transcribe';
5
+ const DEFAULT_SAMPLE_RATE = 16000;
6
+ function supportsVerboseJson(model) {
7
+ return model.startsWith('whisper');
8
+ }
9
+ function toUploadFile(audio, sampleRate) {
10
+ const format = detectAudioFormat(audio);
11
+ if (!format) {
12
+ return new File([new Uint8Array(pcmToWav(audio, sampleRate))], 'audio.wav', {
13
+ type: 'audio/wav',
14
+ });
15
+ }
16
+ return new File([new Uint8Array(audio)], `audio.${format}`, { type: audioMimeType(format) });
17
+ }
18
+ function hasContainerHeader(audio) {
19
+ const format = detectAudioFormat(audio);
20
+ if (format === null)
21
+ return false;
22
+ if (format === 'mp3')
23
+ return audio.toString('ascii', 0, 3) === 'ID3';
24
+ return true;
25
+ }
4
26
  export class OpenAISTT {
5
27
  name = 'openai';
6
28
  client;
@@ -12,20 +34,47 @@ export class OpenAISTT {
12
34
  });
13
35
  this.model = config.model ?? DEFAULT_MODEL;
14
36
  }
37
+ /**
38
+ * Transcribe a complete audio file (wav, mp3, ogg, flac, webm, mp4). Headerless input is
39
+ * treated as raw PCM16 mono at 16kHz and wrapped into a WAV container.
40
+ */
15
41
  async transcribe(audio, options) {
16
- const file = new File([new Uint8Array(audio)], 'audio.wav', { type: 'audio/wav' });
17
- const response = await this.client.audio.transcriptions.create({
42
+ return this.transcribeFile(toUploadFile(audio, DEFAULT_SAMPLE_RATE), options);
43
+ }
44
+ createStream(options) {
45
+ const sampleRate = options?.sampleRate ?? DEFAULT_SAMPLE_RATE;
46
+ return new OpenAISTTStream((audio, opts) => {
47
+ const file = hasContainerHeader(audio)
48
+ ? toUploadFile(audio, sampleRate)
49
+ : new File([new Uint8Array(pcmToWav(audio, sampleRate))], 'audio.wav', {
50
+ type: 'audio/wav',
51
+ });
52
+ return this.transcribeFile(file, opts);
53
+ }, options);
54
+ }
55
+ async transcribeFile(file, options) {
56
+ const common = {
18
57
  file,
19
58
  model: this.model,
20
- response_format: 'verbose_json',
21
- timestamp_granularities: ['word'],
22
59
  ...(options?.language && { language: options.language }),
23
60
  ...(options?.prompt && { prompt: options.prompt }),
61
+ };
62
+ if (supportsVerboseJson(this.model)) {
63
+ const response = await this.client.audio.transcriptions.create({
64
+ ...common,
65
+ response_format: 'verbose_json',
66
+ timestamp_granularities: ['word'],
67
+ });
68
+ return this.mapResponse(response);
69
+ }
70
+ const response = await this.client.audio.transcriptions.create({
71
+ ...common,
72
+ response_format: 'json',
24
73
  });
25
- return this.mapResponse(response);
26
- }
27
- createStream(options) {
28
- return new OpenAISTTStream((audio, opts) => this.transcribe(audio, opts), options);
74
+ const result = { text: response.text };
75
+ if (options?.language)
76
+ result.language = options.language;
77
+ return result;
29
78
  }
30
79
  mapResponse(response) {
31
80
  const result = {
@@ -50,6 +99,8 @@ export class OpenAISTT {
50
99
  }
51
100
  class OpenAISTTStream extends EventEmitter {
52
101
  chunks = [];
102
+ closed = false;
103
+ closePromise = null;
53
104
  transcribeFn;
54
105
  options;
55
106
  constructor(transcribeFn, options) {
@@ -58,11 +109,24 @@ class OpenAISTTStream extends EventEmitter {
58
109
  this.options = options;
59
110
  }
60
111
  write(chunk) {
112
+ if (this.closed) {
113
+ throw new Error('OpenAISTTStream: cannot write after close');
114
+ }
61
115
  this.chunks.push(chunk);
62
116
  }
63
- async close() {
117
+ close() {
118
+ if (!this.closePromise) {
119
+ this.closed = true;
120
+ this.closePromise = this.finish();
121
+ }
122
+ return this.closePromise;
123
+ }
124
+ async finish() {
64
125
  const combined = Buffer.concat(this.chunks);
65
126
  this.chunks = [];
127
+ if (combined.length === 0) {
128
+ return { text: '' };
129
+ }
66
130
  try {
67
131
  const result = await this.transcribeFn(combined, this.options);
68
132
  this.emit('final', result);
@@ -70,7 +134,8 @@ class OpenAISTTStream extends EventEmitter {
70
134
  }
71
135
  catch (err) {
72
136
  const error = err instanceof Error ? err : new Error(String(err));
73
- this.emit('error', error);
137
+ if (this.listenerCount('error') > 0)
138
+ this.emit('error', error);
74
139
  throw error;
75
140
  }
76
141
  }
@@ -1 +1 @@
1
- {"version":3,"file":"openai-stt.js","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,MAAM,MAAM,QAAQ,CAAC;AAgB5B,MAAM,aAAa,GAAG,wBAAwB,CAAC;AAE/C,MAAM,OAAO,SAAS;IACX,IAAI,GAAG,QAAQ,CAAC;IAER,MAAM,CAAS;IACf,KAAK,CAAS;IAE/B,YAAY,MAAuB;QACjC,IAAI,CAAC,MAAM,GAAG,IAAI,MAAM,CAAC;YACvB,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,GAAG,CAAC,MAAM,CAAC,OAAO,IAAI,EAAE,OAAO,EAAE,MAAM,CAAC,OAAO,EAAE,CAAC;SACnD,CAAC,CAAC;QACH,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,KAAa,EAAE,OAAoB;QAClD,MAAM,IAAI,GAAG,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,KAAK,CAAC,CAAC,EAAE,WAAW,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,CAAC,CAAC;QAEnF,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,cAAc,CAAC,MAAM,CAAC;YAC7D,IAAI;YACJ,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,eAAe,EAAE,cAAc;YAC/B,uBAAuB,EAAE,CAAC,MAAM,CAAC;YACjC,GAAG,CAAC,OAAO,EAAE,QAAQ,IAAI,EAAE,QAAQ,EAAE,OAAO,CAAC,QAAQ,EAAE,CAAC;YACxD,GAAG,CAAC,OAAO,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC;SACnD,CAAC,CAAC;QAEH,OAAO,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,CAAC;IACpC,CAAC;IAED,YAAY,CAAC,OAA0B;QACrC,OAAO,IAAI,eAAe,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,UAAU,CAAC,KAAK,EAAE,IAAI,CAAC,EAAE,OAAO,CAAC,CAAC;IACrF,CAAC;IAEO,WAAW,CAAC,QAA8B;QAChD,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,QAAQ,CAAC,IAAI;SACpB,CAAC;QAEF,IAAI,QAAQ,CAAC,QAAQ,EAAE,CAAC;YACtB,MAAM,CAAC,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC;QACtC,CAAC;QAED,IAAI,QAAQ,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;YACpC,MAAM,CAAC,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC;QACtC,CAAC;QAED,IAAI,QAAQ,CAAC,KAAK,IAAI,QAAQ,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAChD,MAAM,CAAC,KAAK,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBACxC,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,KAAK,EAAE,CAAC,CAAC,KAAK;gBACd,GAAG,EAAE,CAAC,CAAC,GAAG;gBACV,UAAU,EAAE,CAAC;aACd,CAAC,CAAC,CAAC;QACN,CAAC;QAED,OAAO,MAAM,CAAC;IAChB,CAAC;CACF;AAID,MAAM,eAAgB,SAAQ,YAAY;IAChC,MAAM,GAAa,EAAE,CAAC;IACb,YAAY,CAAe;IAC3B,OAAO,CAAoB;IAE5C,YAAY,YAA0B,EAAE,OAA0B;QAChE,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,YAAY,GAAG,YAAY,CAAC;QACjC,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;IACzB,CAAC;IAED,KAAK,CAAC,KAAa;QACjB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAC1B,CAAC;IAED,KAAK,CAAC,KAAK;QACT,MAAM,QAAQ,GAAG,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC5C,IAAI,CAAC,MAAM,GAAG,EAAE,CAAC;QAEjB,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;YAC/D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;YAC3B,OAAO,MAAM,CAAC;QAChB,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,KAAK,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;YAClE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;YAC1B,MAAM,KAAK,CAAC;QACd,CAAC;IACH,CAAC;CACF"}
1
+ {"version":3,"file":"openai-stt.js","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,MAAM,MAAM,QAAQ,CAAC;AAE5B,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAezE,MAAM,aAAa,GAAG,wBAAwB,CAAC;AAC/C,MAAM,mBAAmB,GAAG,KAAK,CAAC;AAElC,SAAS,mBAAmB,CAAC,KAAa;IACxC,OAAO,KAAK,CAAC,UAAU,CAAC,SAAS,CAAC,CAAC;AACrC,CAAC;AAED,SAAS,YAAY,CAAC,KAAa,EAAE,UAAkB;IACrD,MAAM,MAAM,GAAG,iBAAiB,CAAC,KAAK,CAAC,CAAC;IACxC,IAAI,CAAC,MAAM,EAAE,CAAC;QACZ,OAAO,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE;YAC1E,IAAI,EAAE,WAAW;SAClB,CAAC,CAAC;IACL,CAAC;IACD,OAAO,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,KAAK,CAAC,CAAC,EAAE,SAAS,MAAM,EAAE,EAAE,EAAE,IAAI,EAAE,aAAa,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;AAC/F,CAAC;AAED,SAAS,kBAAkB,CAAC,KAAa;IACvC,MAAM,MAAM,GAAG,iBAAiB,CAAC,KAAK,CAAC,CAAC;IACxC,IAAI,MAAM,KAAK,IAAI;QAAE,OAAO,KAAK,CAAC;IAClC,IAAI,MAAM,KAAK,KAAK;QAAE,OAAO,KAAK,CAAC,QAAQ,CAAC,OAAO,EAAE,CAAC,EAAE,CAAC,CAAC,KAAK,KAAK,CAAC;IACrE,OAAO,IAAI,CAAC;AACd,CAAC;AAED,MAAM,OAAO,SAAS;IACX,IAAI,GAAG,QAAQ,CAAC;IAER,MAAM,CAAS;IACf,KAAK,CAAS;IAE/B,YAAY,MAAuB;QACjC,IAAI,CAAC,MAAM,GAAG,IAAI,MAAM,CAAC;YACvB,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,GAAG,CAAC,MAAM,CAAC,OAAO,IAAI,EAAE,OAAO,EAAE,MAAM,CAAC,OAAO,EAAE,CAAC;SACnD,CAAC,CAAC;QACH,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,UAAU,CAAC,KAAa,EAAE,OAAoB;QAClD,OAAO,IAAI,CAAC,cAAc,CAAC,YAAY,CAAC,KAAK,EAAE,mBAAmB,CAAC,EAAE,OAAO,CAAC,CAAC;IAChF,CAAC;IAED,YAAY,CAAC,OAA0B;QACrC,MAAM,UAAU,GAAG,OAAO,EAAE,UAAU,IAAI,mBAAmB,CAAC;QAC9D,OAAO,IAAI,eAAe,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE;YACzC,MAAM,IAAI,GAAG,kBAAkB,CAAC,KAAK,CAAC;gBACpC,CAAC,CAAC,YAAY,CAAC,KAAK,EAAE,UAAU,CAAC;gBACjC,CAAC,CAAC,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE;oBACnE,IAAI,EAAE,WAAW;iBAClB,CAAC,CAAC;YACP,OAAO,IAAI,CAAC,cAAc,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;QACzC,CAAC,EAAE,OAAO,CAAC,CAAC;IACd,CAAC;IAEO,KAAK,CAAC,cAAc,CAAC,IAAU,EAAE,OAAoB;QAC3D,MAAM,MAAM,GAAG;YACb,IAAI;YACJ,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,GAAG,CAAC,OAAO,EAAE,QAAQ,IAAI,EAAE,QAAQ,EAAE,OAAO,CAAC,QAAQ,EAAE,CAAC;YACxD,GAAG,CAAC,OAAO,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC;SACnD,CAAC;QAEF,IAAI,mBAAmB,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YACpC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,cAAc,CAAC,MAAM,CAAC;gBAC7D,GAAG,MAAM;gBACT,eAAe,EAAE,cAAc;gBAC/B,uBAAuB,EAAE,CAAC,MAAM,CAAC;aAClC,CAAC,CAAC;YACH,OAAO,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,CAAC;QACpC,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,cAAc,CAAC,MAAM,CAAC;YAC7D,GAAG,MAAM;YACT,eAAe,EAAE,MAAM;SACxB,CAAC,CAAC;QACH,MAAM,MAAM,GAAqB,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,EAAE,CAAC;QACzD,IAAI,OAAO,EAAE,QAAQ;YAAE,MAAM,CAAC,QAAQ,GAAG,OAAO,CAAC,QAAQ,CAAC;QAC1D,OAAO,MAAM,CAAC;IAChB,CAAC;IAEO,WAAW,CAAC,QAA8B;QAChD,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,QAAQ,CAAC,IAAI;SACpB,CAAC;QAEF,IAAI,QAAQ,CAAC,QAAQ,EAAE,CAAC;YACtB,MAAM,CAAC,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC;QACtC,CAAC;QAED,IAAI,QAAQ,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;YACpC,MAAM,CAAC,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC;QACtC,CAAC;QAED,IAAI,QAAQ,CAAC,KAAK,IAAI,QAAQ,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAChD,MAAM,CAAC,KAAK,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBACxC,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,KAAK,EAAE,CAAC,CAAC,KAAK;gBACd,GAAG,EAAE,CAAC,CAAC,GAAG;gBACV,UAAU,EAAE,CAAC;aACd,CAAC,CAAC,CAAC;QACN,CAAC;QAED,OAAO,MAAM,CAAC;IAChB,CAAC;CACF;AAID,MAAM,eAAgB,SAAQ,YAAY;IAChC,MAAM,GAAa,EAAE,CAAC;IACtB,MAAM,GAAG,KAAK,CAAC;IACf,YAAY,GAAqC,IAAI,CAAC;IAC7C,YAAY,CAAe;IAC3B,OAAO,CAAoB;IAE5C,YAAY,YAA0B,EAAE,OAA0B;QAChE,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,YAAY,GAAG,YAAY,CAAC;QACjC,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;IACzB,CAAC;IAED,KAAK,CAAC,KAAa;QACjB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAC;QAC/D,CAAC;QACD,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAC1B,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC;YACvB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;YACnB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QACpC,CAAC;QACD,OAAO,IAAI,CAAC,YAAY,CAAC;IAC3B,CAAC;IAEO,KAAK,CAAC,MAAM;QAClB,MAAM,QAAQ,GAAG,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC5C,IAAI,CAAC,MAAM,GAAG,EAAE,CAAC;QAEjB,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC1B,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;QACtB,CAAC;QAED,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;YAC/D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;YAC3B,OAAO,MAAM,CAAC;QAChB,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,KAAK,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;YAClE,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC;gBAAE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;YAC/D,MAAM,KAAK,CAAC;QACd,CAAC;IACH,CAAC;CACF"}
package/dist/tools.d.ts CHANGED
@@ -14,6 +14,14 @@ type TranscribeParams = z.infer<typeof TranscribeParamsSchema>;
14
14
  declare const SpeakParamsSchema: z.ZodObject<{
15
15
  text: z.ZodString;
16
16
  voice: z.ZodOptional<z.ZodString>;
17
+ format: z.ZodOptional<z.ZodEnum<{
18
+ wav: "wav";
19
+ mp3: "mp3";
20
+ flac: "flac";
21
+ pcm16: "pcm16";
22
+ opus: "opus";
23
+ aac: "aac";
24
+ }>>;
17
25
  }, z.core.$strip>;
18
26
  type SpeakParams = z.infer<typeof SpeakParamsSchema>;
19
27
  export declare function transcribeTool(stt: STTProvider): VoiceTool<TranscribeParams>;
@@ -1 +1 @@
1
- {"version":3,"file":"tools.d.ts","sourceRoot":"","sources":["../src/tools.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,KAAK,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AAE3D,MAAM,WAAW,SAAS,CAAC,OAAO,GAAG,OAAO;IAC1C,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC;IAC/B,OAAO,EAAE,CAAC,MAAM,EAAE,OAAO,KAAK,OAAO,CAAC,OAAO,CAAC,CAAC;CAChD;AAED,QAAA,MAAM,sBAAsB;;;iBAG1B,CAAC;AAEH,KAAK,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,sBAAsB,CAAC,CAAC;AAE/D,QAAA,MAAM,iBAAiB;;;iBAGrB,CAAC;AAEH,KAAK,WAAW,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,iBAAiB,CAAC,CAAC;AAErD,wBAAgB,cAAc,CAAC,GAAG,EAAE,WAAW,GAAG,SAAS,CAAC,gBAAgB,CAAC,CAW5E;AAED,wBAAgB,SAAS,CAAC,GAAG,EAAE,WAAW,GAAG,SAAS,CAAC,WAAW,CAAC,CAUlE;AAED,wBAAgB,UAAU,CAAC,MAAM,EAAE;IACjC,GAAG,EAAE,WAAW,CAAC;IACjB,GAAG,EAAE,WAAW,CAAC;CAClB,GAAG,CAAC,SAAS,CAAC,gBAAgB,CAAC,EAAE,SAAS,CAAC,WAAW,CAAC,CAAC,CAExD"}
1
+ {"version":3,"file":"tools.d.ts","sourceRoot":"","sources":["../src/tools.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,KAAK,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AAE3D,MAAM,WAAW,SAAS,CAAC,OAAO,GAAG,OAAO;IAC1C,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC;IAC/B,OAAO,EAAE,CAAC,MAAM,EAAE,OAAO,KAAK,OAAO,CAAC,OAAO,CAAC,CAAC;CAChD;AAID,QAAA,MAAM,sBAAsB;;;iBAM1B,CAAC;AAEH,KAAK,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,sBAAsB,CAAC,CAAC;AAE/D,QAAA,MAAM,iBAAiB;;;;;;;;;;;iBAOrB,CAAC;AAEH,KAAK,WAAW,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,iBAAiB,CAAC,CAAC;AAErD,wBAAgB,cAAc,CAAC,GAAG,EAAE,WAAW,GAAG,SAAS,CAAC,gBAAgB,CAAC,CAW5E;AAED,wBAAgB,SAAS,CAAC,GAAG,EAAE,WAAW,GAAG,SAAS,CAAC,WAAW,CAAC,CAUlE;AAED,wBAAgB,UAAU,CAAC,MAAM,EAAE;IACjC,GAAG,EAAE,WAAW,CAAC;IACjB,GAAG,EAAE,WAAW,CAAC;CAClB,GAAG,CAAC,SAAS,CAAC,gBAAgB,CAAC,EAAE,SAAS,CAAC,WAAW,CAAC,CAAC,CAExD"}
package/dist/tools.js CHANGED
@@ -1,11 +1,19 @@
1
1
  import { z } from 'zod';
2
+ const MAX_AUDIO_BASE64_LENGTH = 50 * 1024 * 1024 * 1.37;
2
3
  const TranscribeParamsSchema = z.object({
3
- audioBase64: z.string().describe('Base64-encoded audio data'),
4
+ audioBase64: z
5
+ .string()
6
+ .max(Math.ceil(MAX_AUDIO_BASE64_LENGTH))
7
+ .describe('Base64-encoded audio data'),
4
8
  language: z.string().optional().describe('Language code (e.g., "en", "es")'),
5
9
  });
6
10
  const SpeakParamsSchema = z.object({
7
11
  text: z.string().describe('Text to convert to speech'),
8
12
  voice: z.string().optional().describe('Voice to use'),
13
+ format: z
14
+ .enum(['pcm16', 'mp3', 'opus', 'aac', 'flac', 'wav'])
15
+ .optional()
16
+ .describe('Output audio format'),
9
17
  });
10
18
  export function transcribeTool(stt) {
11
19
  return {
@@ -24,9 +32,9 @@ export function speakTool(tts) {
24
32
  name: 'speak_text',
25
33
  description: 'Convert text to speech audio',
26
34
  parameters: SpeakParamsSchema,
27
- execute: async ({ text, voice }) => {
28
- const audio = await tts.synthesize(text, { voice });
29
- return { audioBase64: audio.toString('base64'), format: 'mp3' };
35
+ execute: async ({ text, voice, format }) => {
36
+ const audio = await tts.synthesize(text, { voice, format });
37
+ return { audioBase64: audio.toString('base64'), format: format ?? 'mp3' };
30
38
  },
31
39
  };
32
40
  }
package/dist/tools.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"tools.js","sourceRoot":"","sources":["../src/tools.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAUxB,MAAM,sBAAsB,GAAG,CAAC,CAAC,MAAM,CAAC;IACtC,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,2BAA2B,CAAC;IAC7D,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,kCAAkC,CAAC;CAC7E,CAAC,CAAC;AAIH,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACjC,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,2BAA2B,CAAC;IACtD,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,cAAc,CAAC;CACtD,CAAC,CAAC;AAIH,MAAM,UAAU,cAAc,CAAC,GAAgB;IAC7C,OAAO;QACL,IAAI,EAAE,kBAAkB;QACxB,WAAW,EAAE,yDAAyD;QACtE,UAAU,EAAE,sBAAsB;QAClC,OAAO,EAAE,KAAK,EAAE,EAAE,WAAW,EAAE,QAAQ,EAAE,EAAE,EAAE;YAC3C,MAAM,KAAK,GAAG,MAAM,CAAC,IAAI,CAAC,WAAW,EAAE,QAAQ,CAAC,CAAC;YACjD,MAAM,MAAM,GAAG,MAAM,GAAG,CAAC,UAAU,CAAC,KAAK,EAAE,EAAE,QAAQ,EAAE,CAAC,CAAC;YACzD,OAAO,EAAE,IAAI,EAAE,MAAM,CAAC,IAAI,EAAE,QAAQ,EAAE,MAAM,CAAC,QAAQ,EAAE,QAAQ,EAAE,MAAM,CAAC,QAAQ,EAAE,CAAC;QACrF,CAAC;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,SAAS,CAAC,GAAgB;IACxC,OAAO;QACL,IAAI,EAAE,YAAY;QAClB,WAAW,EAAE,8BAA8B;QAC3C,UAAU,EAAE,iBAAiB;QAC7B,OAAO,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,KAAK,EAAE,EAAE,EAAE;YACjC,MAAM,KAAK,GAAG,MAAM,GAAG,CAAC,UAAU,CAAC,IAAI,EAAE,EAAE,KAAK,EAAE,CAAC,CAAC;YACpD,OAAO,EAAE,WAAW,EAAE,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC;QAClE,CAAC;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,MAG1B;IACC,OAAO,CAAC,cAAc,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,SAAS,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;AAC7D,CAAC"}
1
+ {"version":3,"file":"tools.js","sourceRoot":"","sources":["../src/tools.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAUxB,MAAM,uBAAuB,GAAG,EAAE,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,CAAC;AAExD,MAAM,sBAAsB,GAAG,CAAC,CAAC,MAAM,CAAC;IACtC,WAAW,EAAE,CAAC;SACX,MAAM,EAAE;SACR,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,uBAAuB,CAAC,CAAC;SACvC,QAAQ,CAAC,2BAA2B,CAAC;IACxC,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,kCAAkC,CAAC;CAC7E,CAAC,CAAC;AAIH,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACjC,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,2BAA2B,CAAC;IACtD,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,cAAc,CAAC;IACrD,MAAM,EAAE,CAAC;SACN,IAAI,CAAC,CAAC,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC;SACpD,QAAQ,EAAE;SACV,QAAQ,CAAC,qBAAqB,CAAC;CACnC,CAAC,CAAC;AAIH,MAAM,UAAU,cAAc,CAAC,GAAgB;IAC7C,OAAO;QACL,IAAI,EAAE,kBAAkB;QACxB,WAAW,EAAE,yDAAyD;QACtE,UAAU,EAAE,sBAAsB;QAClC,OAAO,EAAE,KAAK,EAAE,EAAE,WAAW,EAAE,QAAQ,EAAE,EAAE,EAAE;YAC3C,MAAM,KAAK,GAAG,MAAM,CAAC,IAAI,CAAC,WAAW,EAAE,QAAQ,CAAC,CAAC;YACjD,MAAM,MAAM,GAAG,MAAM,GAAG,CAAC,UAAU,CAAC,KAAK,EAAE,EAAE,QAAQ,EAAE,CAAC,CAAC;YACzD,OAAO,EAAE,IAAI,EAAE,MAAM,CAAC,IAAI,EAAE,QAAQ,EAAE,MAAM,CAAC,QAAQ,EAAE,QAAQ,EAAE,MAAM,CAAC,QAAQ,EAAE,CAAC;QACrF,CAAC;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,SAAS,CAAC,GAAgB;IACxC,OAAO;QACL,IAAI,EAAE,YAAY;QAClB,WAAW,EAAE,8BAA8B;QAC3C,UAAU,EAAE,iBAAiB;QAC7B,OAAO,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,KAAK,EAAE,MAAM,EAAE,EAAE,EAAE;YACzC,MAAM,KAAK,GAAG,MAAM,GAAG,CAAC,UAAU,CAAC,IAAI,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,CAAC,CAAC;YAC5D,OAAO,EAAE,WAAW,EAAE,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,MAAM,IAAI,KAAK,EAAE,CAAC;QAC5E,CAAC;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,MAG1B;IACC,OAAO,CAAC,cAAc,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,SAAS,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;AAC7D,CAAC"}
@@ -1,2 +1,3 @@
1
1
  export { WebSocketTransport, VoiceClient } from './ws-transport.js';
2
+ export type { WebSocketTransportOptions } from './ws-transport.js';
2
3
  //# sourceMappingURL=index.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/transport/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC"}
1
+ {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/transport/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACpE,YAAY,EAAE,yBAAyB,EAAE,MAAM,mBAAmB,CAAC"}
@@ -2,6 +2,8 @@ import { EventEmitter } from 'node:events';
2
2
  import http from 'node:http';
3
3
  import { WebSocket } from 'ws';
4
4
  import type { WebSocketTransportConfig } from '../types.js';
5
+ /** @deprecated Use `WebSocketTransportConfig`. */
6
+ export type WebSocketTransportOptions = WebSocketTransportConfig;
5
7
  interface VoiceClientEvents {
6
8
  audio: [chunk: Buffer];
7
9
  message: [msg: Record<string, unknown>];
@@ -11,10 +13,15 @@ interface VoiceClientEvents {
11
13
  export declare class VoiceClient extends EventEmitter<VoiceClientEvents> {
12
14
  private readonly ws;
13
15
  readonly id: string;
16
+ private alive;
14
17
  constructor(ws: WebSocket);
18
+ get isAlive(): boolean;
19
+ markDead(): void;
20
+ ping(): void;
15
21
  sendAudio(chunk: Buffer): void;
16
22
  sendMessage(msg: Record<string, unknown>): void;
17
23
  close(code?: number, reason?: string): void;
24
+ terminate(): void;
18
25
  }
19
26
  interface TransportEvents {
20
27
  connection: [client: VoiceClient];
@@ -22,16 +29,23 @@ interface TransportEvents {
22
29
  export declare class WebSocketTransport extends EventEmitter<TransportEvents> {
23
30
  private readonly path;
24
31
  private readonly maxConnections;
32
+ private readonly verifyClient?;
25
33
  private wss;
26
34
  private server;
27
35
  private ownsServer;
28
36
  private clients;
37
+ private upgradeHandler;
38
+ private pingInterval;
39
+ private pendingUpgrades;
29
40
  constructor(config?: WebSocketTransportConfig);
30
41
  listen(port: number): Promise<void>;
31
42
  attachToServer(server: http.Server): void;
32
43
  get port(): number | undefined;
33
44
  close(): Promise<void>;
34
45
  private setupWss;
46
+ private handleUpgrade;
47
+ private startKeepalive;
48
+ private stopKeepalive;
35
49
  }
36
50
  export {};
37
51
  //# sourceMappingURL=ws-transport.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"ws-transport.d.ts","sourceRoot":"","sources":["../../src/transport/ws-transport.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,IAAI,MAAM,WAAW,CAAC;AAC7B,OAAO,EAAmB,SAAS,EAAE,MAAM,IAAI,CAAC;AAEhD,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAE5D,UAAU,iBAAiB;IACzB,KAAK,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;IACvB,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;IACxC,KAAK,EAAE,EAAE,CAAC;IACV,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;AAED,qBAAa,WAAY,SAAQ,YAAY,CAAC,iBAAiB,CAAC;IAGlD,OAAO,CAAC,QAAQ,CAAC,EAAE;IAF/B,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;gBAES,EAAE,EAAE,SAAS;IAqB1C,SAAS,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAM9B,WAAW,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI;IAM/C,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,IAAI;CAG5C;AAED,UAAU,eAAe;IACvB,UAAU,EAAE,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC;CACnC;AAED,qBAAa,kBAAmB,SAAQ,YAAY,CAAC,eAAe,CAAC;IACnE,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAS;IAC9B,OAAO,CAAC,QAAQ,CAAC,cAAc,CAAS;IACxC,OAAO,CAAC,GAAG,CAAgC;IAC3C,OAAO,CAAC,MAAM,CAA4B;IAC1C,OAAO,CAAC,UAAU,CAAS;IAC3B,OAAO,CAAC,OAAO,CAA0B;gBAE7B,MAAM,CAAC,EAAE,wBAAwB;IAMvC,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAczC,cAAc,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,GAAG,IAAI;IAMzC,IAAI,IAAI,IAAI,MAAM,GAAG,SAAS,CAI7B;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAmB5B,OAAO,CAAC,QAAQ;CA4BjB"}
1
+ {"version":3,"file":"ws-transport.d.ts","sourceRoot":"","sources":["../../src/transport/ws-transport.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,IAAI,MAAM,WAAW,CAAC;AAE7B,OAAO,EAAmB,SAAS,EAAE,MAAM,IAAI,CAAC;AAEhD,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAE5D,kDAAkD;AAClD,MAAM,MAAM,yBAAyB,GAAG,wBAAwB,CAAC;AAWjE,UAAU,iBAAiB;IACzB,KAAK,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;IACvB,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;IACxC,KAAK,EAAE,EAAE,CAAC;IACV,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;AAED,qBAAa,WAAY,SAAQ,YAAY,CAAC,iBAAiB,CAAC;IAIlD,OAAO,CAAC,QAAQ,CAAC,EAAE;IAH/B,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,OAAO,CAAC,KAAK,CAAQ;gBAEQ,EAAE,EAAE,SAAS;IA+B1C,IAAI,OAAO,IAAI,OAAO,CAErB;IAED,QAAQ,IAAI,IAAI;IAIhB,IAAI,IAAI,IAAI;IAMZ,SAAS,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAM9B,WAAW,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI;IAM/C,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,IAAI;IAI3C,SAAS,IAAI,IAAI;CAGlB;AAED,UAAU,eAAe;IACvB,UAAU,EAAE,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC;CACnC;AAED,qBAAa,kBAAmB,SAAQ,YAAY,CAAC,eAAe,CAAC;IACnE,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAS;IAC9B,OAAO,CAAC,QAAQ,CAAC,cAAc,CAAS;IACxC,OAAO,CAAC,QAAQ,CAAC,YAAY,CAAC,CAA2C;IACzE,OAAO,CAAC,GAAG,CAAgC;IAC3C,OAAO,CAAC,MAAM,CAA4B;IAC1C,OAAO,CAAC,UAAU,CAAS;IAC3B,OAAO,CAAC,OAAO,CAA0B;IACzC,OAAO,CAAC,cAAc,CAEN;IAChB,OAAO,CAAC,YAAY,CAA+C;IACnE,OAAO,CAAC,eAAe,CAAK;gBAEhB,MAAM,CAAC,EAAE,wBAAwB;IAOvC,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAczC,cAAc,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,GAAG,IAAI;IAOzC,IAAI,IAAI,IAAI,MAAM,GAAG,SAAS,CAI7B;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IA2B5B,OAAO,CAAC,QAAQ;YA0BF,aAAa;IAgD3B,OAAO,CAAC,cAAc;IAetB,OAAO,CAAC,aAAa;CAMtB"}