@cogitator-ai/voice 0.1.0 → 0.1.4

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 (93) hide show
  1. package/dist/audio.d.ts +10 -0
  2. package/dist/audio.d.ts.map +1 -0
  3. package/dist/audio.js +75 -0
  4. package/dist/audio.js.map +1 -0
  5. package/dist/index.d.ts +17 -0
  6. package/dist/index.d.ts.map +1 -0
  7. package/dist/index.js +16 -0
  8. package/dist/index.js.map +1 -0
  9. package/dist/pipeline/index.d.ts +3 -0
  10. package/dist/pipeline/index.d.ts.map +1 -0
  11. package/dist/pipeline/index.js +3 -0
  12. package/dist/pipeline/index.js.map +1 -0
  13. package/dist/pipeline/pipeline-session.d.ts +21 -0
  14. package/dist/pipeline/pipeline-session.d.ts.map +1 -0
  15. package/dist/pipeline/pipeline-session.js +131 -0
  16. package/dist/pipeline/pipeline-session.js.map +1 -0
  17. package/dist/pipeline/voice-pipeline.d.ts +13 -0
  18. package/dist/pipeline/voice-pipeline.d.ts.map +1 -0
  19. package/dist/pipeline/voice-pipeline.js +17 -0
  20. package/dist/pipeline/voice-pipeline.js.map +1 -0
  21. package/dist/realtime/gemini-realtime.d.ts +28 -0
  22. package/dist/realtime/gemini-realtime.d.ts.map +1 -0
  23. package/dist/realtime/gemini-realtime.js +153 -0
  24. package/dist/realtime/gemini-realtime.js.map +1 -0
  25. package/dist/realtime/index.d.ts +4 -0
  26. package/dist/realtime/index.d.ts.map +1 -0
  27. package/dist/realtime/index.js +4 -0
  28. package/dist/realtime/index.js.map +1 -0
  29. package/dist/realtime/openai-realtime.d.ts +27 -0
  30. package/dist/realtime/openai-realtime.d.ts.map +1 -0
  31. package/dist/realtime/openai-realtime.js +138 -0
  32. package/dist/realtime/openai-realtime.js.map +1 -0
  33. package/dist/realtime/realtime-session.d.ts +23 -0
  34. package/dist/realtime/realtime-session.d.ts.map +1 -0
  35. package/dist/realtime/realtime-session.js +47 -0
  36. package/dist/realtime/realtime-session.js.map +1 -0
  37. package/dist/stt/deepgram-stt.d.ts +17 -0
  38. package/dist/stt/deepgram-stt.d.ts.map +1 -0
  39. package/dist/stt/deepgram-stt.js +138 -0
  40. package/dist/stt/deepgram-stt.js.map +1 -0
  41. package/dist/stt/index.d.ts +3 -0
  42. package/dist/stt/index.d.ts.map +1 -0
  43. package/dist/stt/index.js +3 -0
  44. package/dist/stt/index.js.map +1 -0
  45. package/dist/stt/openai-stt.d.ts +17 -0
  46. package/dist/stt/openai-stt.d.ts.map +1 -0
  47. package/dist/stt/openai-stt.js +80 -0
  48. package/dist/stt/openai-stt.js.map +1 -0
  49. package/dist/tools.d.ts +26 -0
  50. package/dist/tools.d.ts.map +1 -0
  51. package/dist/tools.js +36 -0
  52. package/dist/tools.js.map +1 -0
  53. package/dist/transport/index.d.ts +2 -0
  54. package/dist/transport/index.d.ts.map +1 -0
  55. package/dist/transport/index.js +2 -0
  56. package/dist/transport/index.js.map +1 -0
  57. package/dist/transport/ws-transport.d.ts +37 -0
  58. package/dist/transport/ws-transport.d.ts.map +1 -0
  59. package/dist/transport/ws-transport.js +110 -0
  60. package/dist/transport/ws-transport.js.map +1 -0
  61. package/dist/tts/elevenlabs-tts.d.ts +16 -0
  62. package/dist/tts/elevenlabs-tts.d.ts.map +1 -0
  63. package/dist/tts/elevenlabs-tts.js +77 -0
  64. package/dist/tts/elevenlabs-tts.js.map +1 -0
  65. package/dist/tts/index.d.ts +3 -0
  66. package/dist/tts/index.d.ts.map +1 -0
  67. package/dist/tts/index.js +3 -0
  68. package/dist/tts/index.js.map +1 -0
  69. package/dist/tts/openai-tts.d.ts +18 -0
  70. package/dist/tts/openai-tts.d.ts.map +1 -0
  71. package/dist/tts/openai-tts.js +47 -0
  72. package/dist/tts/openai-tts.js.map +1 -0
  73. package/dist/types.d.ts +96 -0
  74. package/dist/types.d.ts.map +1 -0
  75. package/dist/types.js +2 -0
  76. package/dist/types.js.map +1 -0
  77. package/dist/vad/energy-vad.d.ts +19 -0
  78. package/dist/vad/energy-vad.d.ts.map +1 -0
  79. package/dist/vad/energy-vad.js +49 -0
  80. package/dist/vad/energy-vad.js.map +1 -0
  81. package/dist/vad/index.d.ts +3 -0
  82. package/dist/vad/index.d.ts.map +1 -0
  83. package/dist/vad/index.js +3 -0
  84. package/dist/vad/index.js.map +1 -0
  85. package/dist/vad/silero-vad.d.ts +27 -0
  86. package/dist/vad/silero-vad.d.ts.map +1 -0
  87. package/dist/vad/silero-vad.js +83 -0
  88. package/dist/vad/silero-vad.js.map +1 -0
  89. package/dist/voice-agent.d.ts +25 -0
  90. package/dist/voice-agent.d.ts.map +1 -0
  91. package/dist/voice-agent.js +139 -0
  92. package/dist/voice-agent.js.map +1 -0
  93. package/package.json +1 -1
@@ -0,0 +1,138 @@
1
+ import { EventEmitter } from 'node:events';
2
+ import { WebSocket } from 'ws';
3
+ const DEFAULT_MODEL = 'gpt-4o-mini-realtime-preview';
4
+ const BASE_URL = 'wss://api.openai.com/v1/realtime';
5
+ export class OpenAIRealtimeAdapter extends EventEmitter {
6
+ config;
7
+ model;
8
+ ws = null;
9
+ constructor(config) {
10
+ super();
11
+ this.config = config;
12
+ this.model = config.model ?? DEFAULT_MODEL;
13
+ }
14
+ async connect() {
15
+ const url = `${BASE_URL}?model=${this.model}`;
16
+ return new Promise((resolve, reject) => {
17
+ this.ws = new WebSocket(url, {
18
+ headers: {
19
+ Authorization: `Bearer ${this.config.apiKey}`,
20
+ 'OpenAI-Beta': 'realtime=v1',
21
+ },
22
+ });
23
+ this.ws.on('open', () => {
24
+ this.sendSessionUpdate();
25
+ this.emit('connected');
26
+ resolve();
27
+ });
28
+ this.ws.on('message', (data) => {
29
+ this.handleMessage(typeof data === 'string' ? data : data.toString());
30
+ });
31
+ this.ws.on('error', (err) => {
32
+ reject(err);
33
+ });
34
+ });
35
+ }
36
+ pushAudio(chunk) {
37
+ this.send({
38
+ type: 'input_audio_buffer.append',
39
+ audio: chunk.toString('base64'),
40
+ });
41
+ }
42
+ sendText(text) {
43
+ this.send({
44
+ type: 'conversation.item.create',
45
+ item: {
46
+ type: 'message',
47
+ role: 'user',
48
+ content: [{ type: 'input_text', text }],
49
+ },
50
+ });
51
+ this.send({ type: 'response.create' });
52
+ }
53
+ interrupt() {
54
+ this.send({ type: 'response.cancel' });
55
+ }
56
+ close() {
57
+ this.ws?.close();
58
+ this.ws = null;
59
+ }
60
+ sendSessionUpdate() {
61
+ const session = {
62
+ voice: this.config.voice ?? 'coral',
63
+ input_audio_format: 'pcm16',
64
+ output_audio_format: 'pcm16',
65
+ turn_detection: { type: 'server_vad' },
66
+ };
67
+ if (this.config.instructions) {
68
+ session.instructions = this.config.instructions;
69
+ }
70
+ if (this.config.tools?.length) {
71
+ session.tools = this.config.tools.map((t) => ({
72
+ type: 'function',
73
+ name: t.name,
74
+ description: t.description,
75
+ parameters: t.parameters,
76
+ }));
77
+ }
78
+ this.send({ type: 'session.update', session });
79
+ }
80
+ handleMessage(raw) {
81
+ const event = JSON.parse(raw);
82
+ const type = event.type;
83
+ switch (type) {
84
+ case 'response.audio.delta':
85
+ this.emit('audio', Buffer.from(event.delta, 'base64'));
86
+ break;
87
+ case 'conversation.item.input_audio_transcription.completed':
88
+ this.emit('transcript', event.transcript, 'user');
89
+ break;
90
+ case 'response.audio_transcript.done':
91
+ this.emit('transcript', event.transcript, 'assistant');
92
+ break;
93
+ case 'input_audio_buffer.speech_started':
94
+ this.emit('speech_start');
95
+ break;
96
+ case 'response.function_call_arguments.done':
97
+ void this.handleToolCall(event);
98
+ break;
99
+ case 'error': {
100
+ const errObj = event.error;
101
+ this.emit('error', new Error(errObj.message));
102
+ break;
103
+ }
104
+ }
105
+ }
106
+ async handleToolCall(event) {
107
+ const name = event.name;
108
+ const callId = event.call_id;
109
+ const args = JSON.parse(event.arguments);
110
+ this.emit('tool_call', name, args);
111
+ const tool = this.config.tools?.find((t) => t.name === name);
112
+ if (!tool)
113
+ return;
114
+ let output;
115
+ try {
116
+ const result = await tool.execute(args);
117
+ output = JSON.stringify(result);
118
+ }
119
+ catch (err) {
120
+ output = JSON.stringify({ error: err instanceof Error ? err.message : String(err) });
121
+ }
122
+ this.send({
123
+ type: 'conversation.item.create',
124
+ item: {
125
+ type: 'function_call_output',
126
+ call_id: callId,
127
+ output,
128
+ },
129
+ });
130
+ this.send({ type: 'response.create' });
131
+ }
132
+ send(data) {
133
+ if (this.ws?.readyState === this.ws?.OPEN) {
134
+ this.ws.send(JSON.stringify(data));
135
+ }
136
+ }
137
+ }
138
+ //# sourceMappingURL=openai-realtime.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"openai-realtime.js","sourceRoot":"","sources":["../../src/realtime/openai-realtime.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAG/B,MAAM,aAAa,GAAG,8BAA8B,CAAC;AACrD,MAAM,QAAQ,GAAG,kCAAkC,CAAC;AAWpD,MAAM,OAAO,qBAAsB,SAAQ,YAAkC;IAC1D,MAAM,CAAwB;IAC9B,KAAK,CAAS;IACvB,EAAE,GAAqB,IAAI,CAAC;IAEpC,YAAY,MAA6B;QACvC,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED,KAAK,CAAC,OAAO;QACX,MAAM,GAAG,GAAG,GAAG,QAAQ,UAAU,IAAI,CAAC,KAAK,EAAE,CAAC;QAE9C,OAAO,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YAC3C,IAAI,CAAC,EAAE,GAAG,IAAI,SAAS,CAAC,GAAG,EAAE;gBAC3B,OAAO,EAAE;oBACP,aAAa,EAAE,UAAU,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE;oBAC7C,aAAa,EAAE,aAAa;iBAC7B;aACF,CAAC,CAAC;YAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;gBACtB,IAAI,CAAC,iBAAiB,EAAE,CAAC;gBACzB,IAAI,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;gBACvB,OAAO,EAAE,CAAC;YACZ,CAAC,CAAC,CAAC;YAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,IAAqB,EAAE,EAAE;gBAC9C,IAAI,CAAC,aAAa,CAAC,OAAO,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,EAAE,CAAC,CAAC;YACxE,CAAC,CAAC,CAAC;YAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;gBACjC,MAAM,CAAC,GAAG,CAAC,CAAC;YACd,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;IACL,CAAC;IAED,SAAS,CAAC,KAAa;QACrB,IAAI,CAAC,IAAI,CAAC;YACR,IAAI,EAAE,2BAA2B;YACjC,KAAK,EAAE,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC;SAChC,CAAC,CAAC;IACL,CAAC;IAED,QAAQ,CAAC,IAAY;QACnB,IAAI,CAAC,IAAI,CAAC;YACR,IAAI,EAAE,0BAA0B;YAChC,IAAI,EAAE;gBACJ,IAAI,EAAE,SAAS;gBACf,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,CAAC,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,CAAC;aACxC;SACF,CAAC,CAAC;QACH,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,iBAAiB,EAAE,CAAC,CAAC;IACzC,CAAC;IAED,SAAS;QACP,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,iBAAiB,EAAE,CAAC,CAAC;IACzC,CAAC;IAED,KAAK;QACH,IAAI,CAAC,EAAE,EAAE,KAAK,EAAE,CAAC;QACjB,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC;IACjB,CAAC;IAEO,iBAAiB;QACvB,MAAM,OAAO,GAA4B;YACvC,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,OAAO;YACnC,kBAAkB,EAAE,OAAO;YAC3B,mBAAmB,EAAE,OAAO;YAC5B,cAAc,EAAE,EAAE,IAAI,EAAE,YAAY,EAAE;SACvC,CAAC;QAEF,IAAI,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;YAC7B,OAAO,CAAC,YAAY,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,CAAC;QAClD,CAAC;QAED,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,CAAC;YAC9B,OAAO,CAAC,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBAC5C,IAAI,EAAE,UAAU;gBAChB,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,WAAW,EAAE,CAAC,CAAC,WAAW;gBAC1B,UAAU,EAAE,CAAC,CAAC,UAAU;aACzB,CAAC,CAAC,CAAC;QACN,CAAC;QAED,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,gBAAgB,EAAE,OAAO,EAAE,CAAC,CAAC;IACjD,CAAC;IAEO,aAAa,CAAC,GAAW;QAC/B,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAA4B,CAAC;QACzD,MAAM,IAAI,GAAG,KAAK,CAAC,IAAc,CAAC;QAElC,QAAQ,IAAI,EAAE,CAAC;YACb,KAAK,sBAAsB;gBACzB,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,KAAe,EAAE,QAAQ,CAAC,CAAC,CAAC;gBACjE,MAAM;YAER,KAAK,uDAAuD;gBAC1D,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,KAAK,CAAC,UAAoB,EAAE,MAAM,CAAC,CAAC;gBAC5D,MAAM;YAER,KAAK,gCAAgC;gBACnC,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,KAAK,CAAC,UAAoB,EAAE,WAAW,CAAC,CAAC;gBACjE,MAAM;YAER,KAAK,mCAAmC;gBACtC,IAAI,CAAC,IAAI,CAAC,cAAc,CAAC,CAAC;gBAC1B,MAAM;YAER,KAAK,uCAAuC;gBAC1C,KAAK,IAAI,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC;gBAChC,MAAM;YAER,KAAK,OAAO,CAAC,CAAC,CAAC;gBACb,MAAM,MAAM,GAAG,KAAK,CAAC,KAA2C,CAAC;gBACjE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC;gBAC9C,MAAM;YACR,CAAC;QACH,CAAC;IACH,CAAC;IAEO,KAAK,CAAC,cAAc,CAAC,KAA8B;QACzD,MAAM,IAAI,GAAG,KAAK,CAAC,IAAc,CAAC;QAClC,MAAM,MAAM,GAAG,KAAK,CAAC,OAAiB,CAAC;QACvC,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,SAAmB,CAAY,CAAC;QAE9D,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC;QAEnC,MAAM,IAAI,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,IAAI,CAAC,CAAC;QAC7D,IAAI,CAAC,IAAI;YAAE,OAAO;QAElB,IAAI,MAAc,CAAC;QACnB,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;YACxC,MAAM,GAAG,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC;QAClC,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,GAAG,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QACvF,CAAC;QAED,IAAI,CAAC,IAAI,CAAC;YACR,IAAI,EAAE,0BAA0B;YAChC,IAAI,EAAE;gBACJ,IAAI,EAAE,sBAAsB;gBAC5B,OAAO,EAAE,MAAM;gBACf,MAAM;aACP;SACF,CAAC,CAAC;QACH,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,iBAAiB,EAAE,CAAC,CAAC;IACzC,CAAC;IAEO,IAAI,CAAC,IAA6B;QACxC,IAAI,IAAI,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC,EAAE,EAAE,IAAI,EAAE,CAAC;YAC1C,IAAI,CAAC,EAAG,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;QACtC,CAAC;IACH,CAAC;CACF"}
@@ -0,0 +1,23 @@
1
+ import { EventEmitter } from 'node:events';
2
+ import type { RealtimeSessionConfig } from '../types.js';
3
+ interface RealtimeSessionEvents {
4
+ connected: [];
5
+ audio: [chunk: Buffer];
6
+ transcript: [text: string, role: 'user' | 'assistant'];
7
+ tool_call: [name: string, args: unknown];
8
+ speech_start: [];
9
+ error: [error: Error];
10
+ }
11
+ export declare class RealtimeSession extends EventEmitter<RealtimeSessionEvents> {
12
+ private readonly adapter;
13
+ private readonly _provider;
14
+ constructor(config: RealtimeSessionConfig);
15
+ get provider(): RealtimeSessionConfig['provider'];
16
+ connect(): Promise<void>;
17
+ pushAudio(chunk: Buffer): void;
18
+ sendText(text: string): void;
19
+ interrupt(): void;
20
+ close(): void;
21
+ }
22
+ export {};
23
+ //# sourceMappingURL=realtime-session.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"realtime-session.d.ts","sourceRoot":"","sources":["../../src/realtime/realtime-session.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,qBAAqB,EAAE,MAAM,aAAa,CAAC;AAezD,UAAU,qBAAqB;IAC7B,SAAS,EAAE,EAAE,CAAC;IACd,KAAK,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;IACvB,UAAU,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,GAAG,WAAW,CAAC,CAAC;IACvD,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,CAAC,CAAC;IACzC,YAAY,EAAE,EAAE,CAAC;IACjB,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;AAED,qBAAa,eAAgB,SAAQ,YAAY,CAAC,qBAAqB,CAAC;IACtE,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAkB;IAC1C,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAoC;gBAElD,MAAM,EAAE,qBAAqB;IAezC,IAAI,QAAQ,IAAI,qBAAqB,CAAC,UAAU,CAAC,CAEhD;IAEK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAI9B,SAAS,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAI9B,QAAQ,CAAC,IAAI,EAAE,MAAM,GAAG,IAAI;IAI5B,SAAS,IAAI,IAAI;IAIjB,KAAK,IAAI,IAAI;CAGd"}
@@ -0,0 +1,47 @@
1
+ import { EventEmitter } from 'node:events';
2
+ import { OpenAIRealtimeAdapter } from './openai-realtime.js';
3
+ import { GeminiRealtimeAdapter } from './gemini-realtime.js';
4
+ const FORWARDED_EVENTS = [
5
+ 'connected',
6
+ 'speech_start',
7
+ 'transcript',
8
+ 'audio',
9
+ 'tool_call',
10
+ 'error',
11
+ ];
12
+ export class RealtimeSession extends EventEmitter {
13
+ adapter;
14
+ _provider;
15
+ constructor(config) {
16
+ super();
17
+ this._provider = config.provider;
18
+ this.adapter =
19
+ config.provider === 'openai'
20
+ ? new OpenAIRealtimeAdapter(config)
21
+ : new GeminiRealtimeAdapter(config);
22
+ for (const event of FORWARDED_EVENTS) {
23
+ this.adapter.on(event, (...args) => {
24
+ this.emit(event, ...args);
25
+ });
26
+ }
27
+ }
28
+ get provider() {
29
+ return this._provider;
30
+ }
31
+ async connect() {
32
+ return this.adapter.connect();
33
+ }
34
+ pushAudio(chunk) {
35
+ this.adapter.pushAudio(chunk);
36
+ }
37
+ sendText(text) {
38
+ this.adapter.sendText(text);
39
+ }
40
+ interrupt() {
41
+ this.adapter.interrupt();
42
+ }
43
+ close() {
44
+ this.adapter.close();
45
+ }
46
+ }
47
+ //# sourceMappingURL=realtime-session.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"realtime-session.js","sourceRoot":"","sources":["../../src/realtime/realtime-session.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,EAAE,qBAAqB,EAAE,MAAM,sBAAsB,CAAC;AAC7D,OAAO,EAAE,qBAAqB,EAAE,MAAM,sBAAsB,CAAC;AAI7D,MAAM,gBAAgB,GAAG;IACvB,WAAW;IACX,cAAc;IACd,YAAY;IACZ,OAAO;IACP,WAAW;IACX,OAAO;CACC,CAAC;AAWX,MAAM,OAAO,eAAgB,SAAQ,YAAmC;IACrD,OAAO,CAAkB;IACzB,SAAS,CAAoC;IAE9D,YAAY,MAA6B;QACvC,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,QAAQ,CAAC;QACjC,IAAI,CAAC,OAAO;YACV,MAAM,CAAC,QAAQ,KAAK,QAAQ;gBAC1B,CAAC,CAAC,IAAI,qBAAqB,CAAC,MAAM,CAAC;gBACnC,CAAC,CAAC,IAAI,qBAAqB,CAAC,MAAM,CAAC,CAAC;QAExC,KAAK,MAAM,KAAK,IAAI,gBAAgB,EAAE,CAAC;YACrC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,CAAC,GAAG,IAAe,EAAE,EAAE;gBAC5C,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE,GAAI,IAAc,CAAC,CAAC;YACvC,CAAC,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IAED,IAAI,QAAQ;QACV,OAAO,IAAI,CAAC,SAAS,CAAC;IACxB,CAAC;IAED,KAAK,CAAC,OAAO;QACX,OAAO,IAAI,CAAC,OAAO,CAAC,OAAO,EAAE,CAAC;IAChC,CAAC;IAED,SAAS,CAAC,KAAa;QACrB,IAAI,CAAC,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;IAChC,CAAC;IAED,QAAQ,CAAC,IAAY;QACnB,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;IAC9B,CAAC;IAED,SAAS;QACP,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,CAAC;IAC3B,CAAC;IAED,KAAK;QACH,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC;IACvB,CAAC;CACF"}
@@ -0,0 +1,17 @@
1
+ import type { STTProvider, STTOptions, STTStreamOptions, STTStream, TranscribeResult } from '../types.js';
2
+ export interface DeepgramSTTConfig {
3
+ apiKey: string;
4
+ model?: string;
5
+ language?: string;
6
+ }
7
+ export declare class DeepgramSTT implements STTProvider {
8
+ readonly name = "deepgram";
9
+ private readonly apiKey;
10
+ private readonly model;
11
+ private readonly language?;
12
+ constructor(config: DeepgramSTTConfig);
13
+ transcribe(audio: Buffer, options?: STTOptions): Promise<TranscribeResult>;
14
+ createStream(options?: STTStreamOptions): STTStream;
15
+ private mapBatchResponse;
16
+ }
17
+ //# sourceMappingURL=deepgram-stt.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"deepgram-stt.d.ts","sourceRoot":"","sources":["../../src/stt/deepgram-stt.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EACV,WAAW,EACX,UAAU,EACV,gBAAgB,EAChB,SAAS,EACT,gBAAgB,EACjB,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,iBAAiB;IAChC,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAkCD,qBAAa,WAAY,YAAW,WAAW;IAC7C,QAAQ,CAAC,IAAI,cAAc;IAE3B,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;IAC/B,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAS;gBAEvB,MAAM,EAAE,iBAAiB;IAM/B,UAAU,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,gBAAgB,CAAC;IA2BhF,YAAY,CAAC,OAAO,CAAC,EAAE,gBAAgB,GAAG,SAAS;IAInD,OAAO,CAAC,gBAAgB;CAsBzB"}
@@ -0,0 +1,138 @@
1
+ import { EventEmitter } from 'node:events';
2
+ import { WebSocket } from 'ws';
3
+ const DEFAULT_MODEL = 'nova-3';
4
+ const BASE_URL = 'https://api.deepgram.com/v1/listen';
5
+ const WS_URL = 'wss://api.deepgram.com/v1/listen';
6
+ export class DeepgramSTT {
7
+ name = 'deepgram';
8
+ apiKey;
9
+ model;
10
+ language;
11
+ constructor(config) {
12
+ this.apiKey = config.apiKey;
13
+ this.model = config.model ?? DEFAULT_MODEL;
14
+ this.language = config.language;
15
+ }
16
+ async transcribe(audio, options) {
17
+ const params = new URLSearchParams({
18
+ model: this.model,
19
+ punctuate: 'true',
20
+ });
21
+ const lang = options?.language ?? this.language;
22
+ if (lang)
23
+ params.set('language', lang);
24
+ const response = await fetch(`${BASE_URL}?${params}`, {
25
+ method: 'POST',
26
+ headers: {
27
+ Authorization: `Token ${this.apiKey}`,
28
+ 'Content-Type': 'audio/wav',
29
+ },
30
+ body: new Uint8Array(audio).buffer,
31
+ });
32
+ if (!response.ok) {
33
+ const body = await response.text();
34
+ throw new Error(`Deepgram API error ${response.status}: ${body}`);
35
+ }
36
+ const data = (await response.json());
37
+ return this.mapBatchResponse(data);
38
+ }
39
+ createStream(options) {
40
+ return new DeepgramSTTStream(this.apiKey, this.model, this.language, options);
41
+ }
42
+ mapBatchResponse(data) {
43
+ const alt = data.results.channels[0]?.alternatives[0];
44
+ const result = {
45
+ text: alt?.transcript ?? '',
46
+ };
47
+ if (data.metadata?.duration !== undefined) {
48
+ result.duration = data.metadata.duration;
49
+ }
50
+ if (alt?.words && alt.words.length > 0) {
51
+ result.words = alt.words.map((w) => ({
52
+ word: w.word,
53
+ start: w.start,
54
+ end: w.end,
55
+ confidence: w.confidence,
56
+ }));
57
+ }
58
+ return result;
59
+ }
60
+ }
61
+ class DeepgramSTTStream extends EventEmitter {
62
+ ws;
63
+ ready = false;
64
+ pendingChunks = [];
65
+ lastResult = { text: '' };
66
+ constructor(apiKey, model, language, options) {
67
+ super();
68
+ const params = new URLSearchParams({
69
+ model,
70
+ punctuate: 'true',
71
+ interim_results: 'true',
72
+ });
73
+ const lang = options?.language ?? language;
74
+ if (lang)
75
+ params.set('language', lang);
76
+ if (options?.endpointing !== undefined) {
77
+ params.set('endpointing', String(options.endpointing));
78
+ }
79
+ this.ws = new WebSocket(`${WS_URL}?${params}`, [`token:${apiKey}`]);
80
+ this.ws.on('open', () => {
81
+ this.ready = true;
82
+ for (const chunk of this.pendingChunks) {
83
+ this.ws.send(chunk);
84
+ }
85
+ this.pendingChunks = [];
86
+ });
87
+ this.ws.on('message', (raw) => {
88
+ const msg = JSON.parse(String(raw));
89
+ const alt = msg.channel?.alternatives[0];
90
+ if (!alt?.transcript)
91
+ return;
92
+ if (msg.is_final) {
93
+ const result = this.mapStreamResult(msg);
94
+ this.lastResult = result;
95
+ this.emit('final', result);
96
+ }
97
+ else {
98
+ this.emit('partial', alt.transcript);
99
+ }
100
+ });
101
+ this.ws.on('error', (err) => {
102
+ this.emit('error', err);
103
+ });
104
+ }
105
+ write(chunk) {
106
+ if (this.ready) {
107
+ this.ws.send(chunk);
108
+ }
109
+ else {
110
+ this.pendingChunks.push(chunk);
111
+ }
112
+ }
113
+ async close() {
114
+ if (this.ready) {
115
+ this.ws.send(JSON.stringify({ type: 'CloseStream' }));
116
+ }
117
+ return this.lastResult;
118
+ }
119
+ mapStreamResult(msg) {
120
+ const alt = msg.channel.alternatives[0];
121
+ const result = {
122
+ text: alt.transcript,
123
+ };
124
+ if (msg.metadata?.duration !== undefined) {
125
+ result.duration = msg.metadata.duration;
126
+ }
127
+ if (alt.words && alt.words.length > 0) {
128
+ result.words = alt.words.map((w) => ({
129
+ word: w.word,
130
+ start: w.start,
131
+ end: w.end,
132
+ confidence: w.confidence,
133
+ }));
134
+ }
135
+ return result;
136
+ }
137
+ }
138
+ //# sourceMappingURL=deepgram-stt.js.map
@@ -0,0 +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,CAAC,MAAqB;SAClD,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,MAAM,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAA0B,CAAC;YAC7D,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;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,IAAI,CAAC,KAAK,EAAE,CAAC;YACf,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;QACxD,CAAC;QACD,OAAO,IAAI,CAAC,UAAU,CAAC;IACzB,CAAC;IAEO,eAAe,CAAC,GAA0B;QAChD,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,CAAC,YAAY,CAAC,CAAC,CAAE,CAAC;QACzC,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"}
@@ -0,0 +1,3 @@
1
+ export { OpenAISTT, type OpenAISTTConfig } from './openai-stt.js';
2
+ export { DeepgramSTT, type DeepgramSTTConfig } from './deepgram-stt.js';
3
+ //# sourceMappingURL=index.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/stt/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,KAAK,eAAe,EAAE,MAAM,iBAAiB,CAAC;AAClE,OAAO,EAAE,WAAW,EAAE,KAAK,iBAAiB,EAAE,MAAM,mBAAmB,CAAC"}
@@ -0,0 +1,3 @@
1
+ export { OpenAISTT } from './openai-stt.js';
2
+ export { DeepgramSTT } from './deepgram-stt.js';
3
+ //# sourceMappingURL=index.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/stt/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAwB,MAAM,iBAAiB,CAAC;AAClE,OAAO,EAAE,WAAW,EAA0B,MAAM,mBAAmB,CAAC"}
@@ -0,0 +1,17 @@
1
+ import type { STTProvider, STTOptions, STTStreamOptions, STTStream, TranscribeResult } from '../types.js';
2
+ export interface OpenAISTTConfig {
3
+ apiKey: string;
4
+ model?: string;
5
+ baseURL?: string;
6
+ }
7
+ export declare class OpenAISTT implements STTProvider {
8
+ readonly name = "openai";
9
+ private readonly client;
10
+ private readonly model;
11
+ constructor(config: OpenAISTTConfig);
12
+ transcribe(audio: Buffer, options?: STTOptions): Promise<TranscribeResult>;
13
+ createStream(options?: STTStreamOptions): STTStream;
14
+ _transcribeBuffer(audio: Buffer, options?: STTOptions): Promise<TranscribeResult>;
15
+ private mapResponse;
16
+ }
17
+ //# sourceMappingURL=openai-stt.d.ts.map
@@ -0,0 +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,iBAAiB,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,gBAAgB,CAAC;IAIjF,OAAO,CAAC,WAAW;CAwBpB"}
@@ -0,0 +1,80 @@
1
+ import { EventEmitter } from 'node:events';
2
+ import OpenAI from 'openai';
3
+ const DEFAULT_MODEL = 'gpt-4o-mini-transcribe';
4
+ export class OpenAISTT {
5
+ name = 'openai';
6
+ client;
7
+ model;
8
+ constructor(config) {
9
+ this.client = new OpenAI({
10
+ apiKey: config.apiKey,
11
+ ...(config.baseURL && { baseURL: config.baseURL }),
12
+ });
13
+ this.model = config.model ?? DEFAULT_MODEL;
14
+ }
15
+ 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({
18
+ file,
19
+ model: this.model,
20
+ response_format: 'verbose_json',
21
+ timestamp_granularities: ['word'],
22
+ ...(options?.language && { language: options.language }),
23
+ ...(options?.prompt && { prompt: options.prompt }),
24
+ });
25
+ return this.mapResponse(response);
26
+ }
27
+ createStream(options) {
28
+ return new OpenAISTTStream(this, options);
29
+ }
30
+ _transcribeBuffer(audio, options) {
31
+ return this.transcribe(audio, options);
32
+ }
33
+ mapResponse(response) {
34
+ const result = {
35
+ text: response.text,
36
+ };
37
+ if (response.language) {
38
+ result.language = response.language;
39
+ }
40
+ if (response.duration !== undefined) {
41
+ result.duration = response.duration;
42
+ }
43
+ if (response.words && response.words.length > 0) {
44
+ result.words = response.words.map((w) => ({
45
+ word: w.word,
46
+ start: w.start,
47
+ end: w.end,
48
+ confidence: 1,
49
+ }));
50
+ }
51
+ return result;
52
+ }
53
+ }
54
+ class OpenAISTTStream extends EventEmitter {
55
+ chunks = [];
56
+ provider;
57
+ options;
58
+ constructor(provider, options) {
59
+ super();
60
+ this.provider = provider;
61
+ this.options = options;
62
+ }
63
+ write(chunk) {
64
+ this.chunks.push(chunk);
65
+ }
66
+ async close() {
67
+ const combined = Buffer.concat(this.chunks);
68
+ try {
69
+ const result = await this.provider._transcribeBuffer(combined, this.options);
70
+ this.emit('final', result);
71
+ return result;
72
+ }
73
+ catch (err) {
74
+ const error = err instanceof Error ? err : new Error(String(err));
75
+ this.emit('error', error);
76
+ throw error;
77
+ }
78
+ }
79
+ }
80
+ //# sourceMappingURL=openai-stt.js.map
@@ -0,0 +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,IAAI,EAAE,OAAO,CAAC,CAAC;IAC5C,CAAC;IAED,iBAAiB,CAAC,KAAa,EAAE,OAAoB;QACnD,OAAO,IAAI,CAAC,UAAU,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;IACzC,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;AAED,MAAM,eAAgB,SAAQ,YAAY;IAChC,MAAM,GAAa,EAAE,CAAC;IACb,QAAQ,CAAY;IACpB,OAAO,CAAoB;IAE5C,YAAY,QAAmB,EAAE,OAA0B;QACzD,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC;QACzB,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;QAE5C,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,QAAQ,CAAC,iBAAiB,CAAC,QAAQ,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;YAC7E,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"}
@@ -0,0 +1,26 @@
1
+ import { z } from 'zod';
2
+ import type { STTProvider, TTSProvider } from './types.js';
3
+ export interface VoiceTool<TParams = unknown> {
4
+ name: string;
5
+ description: string;
6
+ parameters: z.ZodType<TParams>;
7
+ execute: (params: TParams) => Promise<unknown>;
8
+ }
9
+ declare const TranscribeParamsSchema: z.ZodObject<{
10
+ audioBase64: z.ZodString;
11
+ language: z.ZodOptional<z.ZodString>;
12
+ }, z.core.$strip>;
13
+ type TranscribeParams = z.infer<typeof TranscribeParamsSchema>;
14
+ declare const SpeakParamsSchema: z.ZodObject<{
15
+ text: z.ZodString;
16
+ voice: z.ZodOptional<z.ZodString>;
17
+ }, z.core.$strip>;
18
+ type SpeakParams = z.infer<typeof SpeakParamsSchema>;
19
+ export declare function transcribeTool(stt: STTProvider): VoiceTool<TranscribeParams>;
20
+ export declare function speakTool(tts: TTSProvider): VoiceTool<SpeakParams>;
21
+ export declare function voiceTools(config: {
22
+ stt: STTProvider;
23
+ tts: TTSProvider;
24
+ }): [VoiceTool<TranscribeParams>, VoiceTool<SpeakParams>];
25
+ export {};
26
+ //# sourceMappingURL=tools.d.ts.map
@@ -0,0 +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"}
package/dist/tools.js ADDED
@@ -0,0 +1,36 @@
1
+ import { z } from 'zod';
2
+ const TranscribeParamsSchema = z.object({
3
+ audioBase64: z.string().describe('Base64-encoded audio data'),
4
+ language: z.string().optional().describe('Language code (e.g., "en", "es")'),
5
+ });
6
+ const SpeakParamsSchema = z.object({
7
+ text: z.string().describe('Text to convert to speech'),
8
+ voice: z.string().optional().describe('Voice to use'),
9
+ });
10
+ export function transcribeTool(stt) {
11
+ return {
12
+ name: 'transcribe_audio',
13
+ description: 'Transcribe audio input to text using speech recognition',
14
+ parameters: TranscribeParamsSchema,
15
+ execute: async ({ audioBase64, language }) => {
16
+ const audio = Buffer.from(audioBase64, 'base64');
17
+ const result = await stt.transcribe(audio, { language });
18
+ return { text: result.text, language: result.language, duration: result.duration };
19
+ },
20
+ };
21
+ }
22
+ export function speakTool(tts) {
23
+ return {
24
+ name: 'speak_text',
25
+ description: 'Convert text to speech audio',
26
+ parameters: SpeakParamsSchema,
27
+ execute: async ({ text, voice }) => {
28
+ const audio = await tts.synthesize(text, { voice });
29
+ return { audioBase64: audio.toString('base64'), format: 'mp3' };
30
+ },
31
+ };
32
+ }
33
+ export function voiceTools(config) {
34
+ return [transcribeTool(config.stt), speakTool(config.tts)];
35
+ }
36
+ //# sourceMappingURL=tools.js.map
@@ -0,0 +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"}
@@ -0,0 +1,2 @@
1
+ export { WebSocketTransport, VoiceClient } from './ws-transport.js';
2
+ //# sourceMappingURL=index.d.ts.map
@@ -0,0 +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"}
@@ -0,0 +1,2 @@
1
+ export { WebSocketTransport, VoiceClient } from './ws-transport.js';
2
+ //# sourceMappingURL=index.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/transport/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC"}
@@ -0,0 +1,37 @@
1
+ import { EventEmitter } from 'node:events';
2
+ import http from 'node:http';
3
+ import { WebSocket } from 'ws';
4
+ import type { WebSocketTransportConfig } from '../types.js';
5
+ interface VoiceClientEvents {
6
+ audio: [chunk: Buffer];
7
+ message: [msg: Record<string, unknown>];
8
+ close: [];
9
+ error: [error: Error];
10
+ }
11
+ export declare class VoiceClient extends EventEmitter<VoiceClientEvents> {
12
+ private readonly ws;
13
+ readonly id: string;
14
+ constructor(ws: WebSocket);
15
+ sendAudio(chunk: Buffer): void;
16
+ sendMessage(msg: Record<string, unknown>): void;
17
+ close(code?: number, reason?: string): void;
18
+ }
19
+ interface TransportEvents {
20
+ connection: [client: VoiceClient];
21
+ }
22
+ export declare class WebSocketTransport extends EventEmitter<TransportEvents> {
23
+ private readonly path;
24
+ private readonly maxConnections;
25
+ private wss;
26
+ private server;
27
+ private ownsServer;
28
+ private clients;
29
+ constructor(config?: WebSocketTransportConfig);
30
+ listen(port: number): Promise<void>;
31
+ attachToServer(server: http.Server): void;
32
+ get port(): number | undefined;
33
+ close(): Promise<void>;
34
+ private setupWss;
35
+ }
36
+ export {};
37
+ //# sourceMappingURL=ws-transport.d.ts.map