@livekit/agents 1.0.24 → 1.0.27

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 (184) hide show
  1. package/dist/inference/llm.cjs +1 -2
  2. package/dist/inference/llm.cjs.map +1 -1
  3. package/dist/inference/llm.d.ts.map +1 -1
  4. package/dist/inference/llm.js +1 -2
  5. package/dist/inference/llm.js.map +1 -1
  6. package/dist/inference/stt.cjs +1 -1
  7. package/dist/inference/stt.cjs.map +1 -1
  8. package/dist/inference/stt.d.ts.map +1 -1
  9. package/dist/inference/stt.js +1 -1
  10. package/dist/inference/stt.js.map +1 -1
  11. package/dist/inference/tts.cjs +4 -4
  12. package/dist/inference/tts.cjs.map +1 -1
  13. package/dist/inference/tts.d.cts +0 -1
  14. package/dist/inference/tts.d.ts +0 -1
  15. package/dist/inference/tts.d.ts.map +1 -1
  16. package/dist/inference/tts.js +4 -4
  17. package/dist/inference/tts.js.map +1 -1
  18. package/dist/ipc/job_proc_lazy_main.cjs +1 -1
  19. package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
  20. package/dist/ipc/job_proc_lazy_main.js +1 -1
  21. package/dist/ipc/job_proc_lazy_main.js.map +1 -1
  22. package/dist/job.cjs +29 -2
  23. package/dist/job.cjs.map +1 -1
  24. package/dist/job.d.cts +6 -0
  25. package/dist/job.d.ts +6 -0
  26. package/dist/job.d.ts.map +1 -1
  27. package/dist/job.js +19 -2
  28. package/dist/job.js.map +1 -1
  29. package/dist/llm/llm.cjs +2 -1
  30. package/dist/llm/llm.cjs.map +1 -1
  31. package/dist/llm/llm.d.cts +1 -1
  32. package/dist/llm/llm.d.ts +1 -1
  33. package/dist/llm/llm.d.ts.map +1 -1
  34. package/dist/llm/llm.js +2 -1
  35. package/dist/llm/llm.js.map +1 -1
  36. package/dist/stream/deferred_stream.cjs +12 -4
  37. package/dist/stream/deferred_stream.cjs.map +1 -1
  38. package/dist/stream/deferred_stream.d.cts +6 -1
  39. package/dist/stream/deferred_stream.d.ts +6 -1
  40. package/dist/stream/deferred_stream.d.ts.map +1 -1
  41. package/dist/stream/deferred_stream.js +12 -4
  42. package/dist/stream/deferred_stream.js.map +1 -1
  43. package/dist/stream/deferred_stream.test.cjs +2 -2
  44. package/dist/stream/deferred_stream.test.cjs.map +1 -1
  45. package/dist/stream/deferred_stream.test.js +2 -2
  46. package/dist/stream/deferred_stream.test.js.map +1 -1
  47. package/dist/stt/stream_adapter.cjs +15 -8
  48. package/dist/stt/stream_adapter.cjs.map +1 -1
  49. package/dist/stt/stream_adapter.d.cts +7 -3
  50. package/dist/stt/stream_adapter.d.ts +7 -3
  51. package/dist/stt/stream_adapter.d.ts.map +1 -1
  52. package/dist/stt/stream_adapter.js +15 -8
  53. package/dist/stt/stream_adapter.js.map +1 -1
  54. package/dist/stt/stt.cjs +8 -3
  55. package/dist/stt/stt.cjs.map +1 -1
  56. package/dist/stt/stt.d.cts +9 -3
  57. package/dist/stt/stt.d.ts +9 -3
  58. package/dist/stt/stt.d.ts.map +1 -1
  59. package/dist/stt/stt.js +9 -4
  60. package/dist/stt/stt.js.map +1 -1
  61. package/dist/telemetry/traces.cjs +23 -2
  62. package/dist/telemetry/traces.cjs.map +1 -1
  63. package/dist/telemetry/traces.d.ts.map +1 -1
  64. package/dist/telemetry/traces.js +23 -2
  65. package/dist/telemetry/traces.js.map +1 -1
  66. package/dist/tts/stream_adapter.cjs +10 -7
  67. package/dist/tts/stream_adapter.cjs.map +1 -1
  68. package/dist/tts/stream_adapter.d.cts +6 -3
  69. package/dist/tts/stream_adapter.d.ts +6 -3
  70. package/dist/tts/stream_adapter.d.ts.map +1 -1
  71. package/dist/tts/stream_adapter.js +10 -7
  72. package/dist/tts/stream_adapter.js.map +1 -1
  73. package/dist/tts/tts.cjs +27 -16
  74. package/dist/tts/tts.cjs.map +1 -1
  75. package/dist/tts/tts.d.cts +12 -5
  76. package/dist/tts/tts.d.ts +12 -5
  77. package/dist/tts/tts.d.ts.map +1 -1
  78. package/dist/tts/tts.js +28 -17
  79. package/dist/tts/tts.js.map +1 -1
  80. package/dist/types.cjs +21 -32
  81. package/dist/types.cjs.map +1 -1
  82. package/dist/types.d.cts +41 -10
  83. package/dist/types.d.ts +41 -10
  84. package/dist/types.d.ts.map +1 -1
  85. package/dist/types.js +18 -30
  86. package/dist/types.js.map +1 -1
  87. package/dist/voice/agent.cjs +54 -19
  88. package/dist/voice/agent.cjs.map +1 -1
  89. package/dist/voice/agent.d.ts.map +1 -1
  90. package/dist/voice/agent.js +54 -19
  91. package/dist/voice/agent.js.map +1 -1
  92. package/dist/voice/agent_activity.cjs +0 -3
  93. package/dist/voice/agent_activity.cjs.map +1 -1
  94. package/dist/voice/agent_activity.d.ts.map +1 -1
  95. package/dist/voice/agent_activity.js +0 -3
  96. package/dist/voice/agent_activity.js.map +1 -1
  97. package/dist/voice/agent_session.cjs +107 -27
  98. package/dist/voice/agent_session.cjs.map +1 -1
  99. package/dist/voice/agent_session.d.cts +16 -2
  100. package/dist/voice/agent_session.d.ts +16 -2
  101. package/dist/voice/agent_session.d.ts.map +1 -1
  102. package/dist/voice/agent_session.js +110 -27
  103. package/dist/voice/agent_session.js.map +1 -1
  104. package/dist/voice/events.cjs.map +1 -1
  105. package/dist/voice/events.d.cts +4 -4
  106. package/dist/voice/events.d.ts +4 -4
  107. package/dist/voice/events.d.ts.map +1 -1
  108. package/dist/voice/events.js.map +1 -1
  109. package/dist/voice/generation.cjs +6 -7
  110. package/dist/voice/generation.cjs.map +1 -1
  111. package/dist/voice/generation.d.ts.map +1 -1
  112. package/dist/voice/generation.js +7 -8
  113. package/dist/voice/generation.js.map +1 -1
  114. package/dist/voice/io.cjs +16 -0
  115. package/dist/voice/io.cjs.map +1 -1
  116. package/dist/voice/io.d.cts +8 -0
  117. package/dist/voice/io.d.ts +8 -0
  118. package/dist/voice/io.d.ts.map +1 -1
  119. package/dist/voice/io.js +16 -0
  120. package/dist/voice/io.js.map +1 -1
  121. package/dist/voice/recorder_io/index.cjs +23 -0
  122. package/dist/voice/recorder_io/index.cjs.map +1 -0
  123. package/dist/voice/recorder_io/index.d.cts +2 -0
  124. package/dist/voice/recorder_io/index.d.ts +2 -0
  125. package/dist/voice/recorder_io/index.d.ts.map +1 -0
  126. package/dist/voice/recorder_io/index.js +2 -0
  127. package/dist/voice/recorder_io/index.js.map +1 -0
  128. package/dist/voice/recorder_io/recorder_io.cjs +542 -0
  129. package/dist/voice/recorder_io/recorder_io.cjs.map +1 -0
  130. package/dist/voice/recorder_io/recorder_io.d.cts +100 -0
  131. package/dist/voice/recorder_io/recorder_io.d.ts +100 -0
  132. package/dist/voice/recorder_io/recorder_io.d.ts.map +1 -0
  133. package/dist/voice/recorder_io/recorder_io.js +508 -0
  134. package/dist/voice/recorder_io/recorder_io.js.map +1 -0
  135. package/dist/voice/report.cjs +7 -2
  136. package/dist/voice/report.cjs.map +1 -1
  137. package/dist/voice/report.d.cts +11 -1
  138. package/dist/voice/report.d.ts +11 -1
  139. package/dist/voice/report.d.ts.map +1 -1
  140. package/dist/voice/report.js +7 -2
  141. package/dist/voice/report.js.map +1 -1
  142. package/dist/voice/room_io/_input.cjs +2 -1
  143. package/dist/voice/room_io/_input.cjs.map +1 -1
  144. package/dist/voice/room_io/_input.d.ts.map +1 -1
  145. package/dist/voice/room_io/_input.js +2 -1
  146. package/dist/voice/room_io/_input.js.map +1 -1
  147. package/dist/voice/room_io/_output.cjs +8 -7
  148. package/dist/voice/room_io/_output.cjs.map +1 -1
  149. package/dist/voice/room_io/_output.d.cts +2 -1
  150. package/dist/voice/room_io/_output.d.ts +2 -1
  151. package/dist/voice/room_io/_output.d.ts.map +1 -1
  152. package/dist/voice/room_io/_output.js +8 -7
  153. package/dist/voice/room_io/_output.js.map +1 -1
  154. package/dist/worker.cjs +4 -3
  155. package/dist/worker.cjs.map +1 -1
  156. package/dist/worker.js +4 -3
  157. package/dist/worker.js.map +1 -1
  158. package/package.json +1 -1
  159. package/src/inference/llm.ts +0 -1
  160. package/src/inference/stt.ts +1 -2
  161. package/src/inference/tts.ts +5 -4
  162. package/src/ipc/job_proc_lazy_main.ts +1 -1
  163. package/src/job.ts +21 -2
  164. package/src/llm/llm.ts +2 -2
  165. package/src/stream/deferred_stream.test.ts +3 -3
  166. package/src/stream/deferred_stream.ts +22 -5
  167. package/src/stt/stream_adapter.ts +18 -8
  168. package/src/stt/stt.ts +19 -6
  169. package/src/telemetry/traces.ts +25 -3
  170. package/src/tts/stream_adapter.ts +15 -7
  171. package/src/tts/tts.ts +46 -21
  172. package/src/types.ts +57 -33
  173. package/src/voice/agent.ts +59 -19
  174. package/src/voice/agent_activity.ts +0 -3
  175. package/src/voice/agent_session.ts +142 -35
  176. package/src/voice/events.ts +6 -3
  177. package/src/voice/generation.ts +10 -8
  178. package/src/voice/io.ts +19 -0
  179. package/src/voice/recorder_io/index.ts +4 -0
  180. package/src/voice/recorder_io/recorder_io.ts +690 -0
  181. package/src/voice/report.ts +20 -3
  182. package/src/voice/room_io/_input.ts +2 -1
  183. package/src/voice/room_io/_output.ts +10 -7
  184. package/src/worker.ts +1 -1
package/dist/stt/stt.cjs CHANGED
@@ -51,9 +51,9 @@ class STT extends import_node_events.EventEmitter {
51
51
  return this.#capabilities;
52
52
  }
53
53
  /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
54
- async recognize(frame) {
54
+ async recognize(frame, abortSignal) {
55
55
  const startTime = process.hrtime.bigint();
56
- const event = await this._recognize(frame);
56
+ const event = await this._recognize(frame, abortSignal);
57
57
  const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1e6));
58
58
  this.emit("metrics_collected", {
59
59
  type: "stt_metrics",
@@ -82,6 +82,7 @@ class SpeechStream {
82
82
  deferredInputStream;
83
83
  logger = (0, import_log.log)();
84
84
  _connOptions;
85
+ abortController = new AbortController();
85
86
  constructor(stt, sampleRate, connectionOptions = import_types.DEFAULT_API_CONNECT_OPTIONS) {
86
87
  this.#stt = stt;
87
88
  this._connOptions = connectionOptions;
@@ -97,7 +98,7 @@ class SpeechStream {
97
98
  return await this.run();
98
99
  } catch (error) {
99
100
  if (error instanceof import_exceptions.APIError) {
100
- const retryInterval = this._connOptions._intervalForRetry(i);
101
+ const retryInterval = (0, import_types.intervalForRetry)(this._connOptions, i);
101
102
  if (this._connOptions.maxRetry === 0 || !error.retryable) {
102
103
  this.emitError({ error, recoverable: false });
103
104
  throw error;
@@ -177,6 +178,9 @@ class SpeechStream {
177
178
  this.output.close();
178
179
  }
179
180
  }
181
+ get abortSignal() {
182
+ return this.abortController.signal;
183
+ }
180
184
  updateInputStream(audioStream) {
181
185
  this.deferredInputStream.setSource(audioStream);
182
186
  }
@@ -233,6 +237,7 @@ class SpeechStream {
233
237
  if (!this.input.closed) this.input.close();
234
238
  if (!this.queue.closed) this.queue.close();
235
239
  if (!this.output.closed) this.output.close();
240
+ if (!this.abortController.signal.aborted) this.abortController.abort();
236
241
  this.closed = true;
237
242
  }
238
243
  [Symbol.asyncIterator]() {
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n });\n return event;\n }\n protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAEhD,yBAA6B;AAE7B,wBAA6C;AAC7C,mBAA8C;AAC9C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAoE;AAEpE,mBAA8D;AAGvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AAmFL,MAAe,YAAa,gCAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,UAAM,4CAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EASA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,gCAAoE;AAAA,EAChF,SAAS,IAAI,gCAAgC;AAAA,EAC7C,QAAQ,IAAI,gCAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EACb;AAAA,EAER,YACE,KACA,YACA,oBAAuC,0CACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,gCAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,+BAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
1
+ {"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame, abortSignal);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n });\n return event;\n }\n\n protected abstract _recognize(\n frame: AudioBuffer,\n abortSignal?: AbortSignal,\n ): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n\n protected abortController = new AbortController();\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n protected get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAEhD,yBAA6B;AAE7B,wBAA6C;AAC7C,mBAA8C;AAC9C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAsF;AAEtF,mBAA8D;AAGvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AAmFL,MAAe,YAAa,gCAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAAoB,aAAiD;AACnF,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,OAAO,WAAW;AACtD,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,UAAM,4CAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAeA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,gCAAoE;AAAA,EAChF,SAAS,IAAI,gCAAgC;AAAA,EAC7C,QAAQ,IAAI,gCAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EACb;AAAA,EAEE,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,KACA,YACA,oBAAuC,0CACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,gCAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,oBAAgB,+BAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,IAAc,cAA2B;AACvC,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,+BAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
@@ -91,13 +91,17 @@ export declare abstract class STT extends STT_base {
91
91
  /** Returns this STT's capabilities */
92
92
  get capabilities(): STTCapabilities;
93
93
  /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
94
- recognize(frame: AudioBuffer): Promise<SpeechEvent>;
95
- protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;
94
+ recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent>;
95
+ protected abstract _recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent>;
96
96
  /**
97
97
  * Returns a {@link SpeechStream} that can be used to push audio frames and receive
98
98
  * transcriptions
99
+ *
100
+ * @param options - Optional configuration including connection options
99
101
  */
100
- abstract stream(): SpeechStream;
102
+ abstract stream(options?: {
103
+ connOptions?: APIConnectOptions;
104
+ }): SpeechStream;
101
105
  close(): Promise<void>;
102
106
  }
103
107
  /**
@@ -129,12 +133,14 @@ export declare abstract class SpeechStream implements AsyncIterableIterator<Spee
129
133
  private deferredInputStream;
130
134
  private logger;
131
135
  private _connOptions;
136
+ protected abortController: AbortController;
132
137
  constructor(stt: STT, sampleRate?: number, connectionOptions?: APIConnectOptions);
133
138
  private mainTask;
134
139
  private emitError;
135
140
  protected pumpInput(): Promise<void>;
136
141
  protected monitorMetrics(): Promise<void>;
137
142
  protected abstract run(): Promise<void>;
143
+ protected get abortSignal(): AbortSignal;
138
144
  updateInputStream(audioStream: ReadableStream<AudioFrame>): void;
139
145
  detachInputStream(): void;
140
146
  /** Push an audio frame to the STT */
package/dist/stt/stt.d.ts CHANGED
@@ -91,13 +91,17 @@ export declare abstract class STT extends STT_base {
91
91
  /** Returns this STT's capabilities */
92
92
  get capabilities(): STTCapabilities;
93
93
  /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
94
- recognize(frame: AudioBuffer): Promise<SpeechEvent>;
95
- protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;
94
+ recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent>;
95
+ protected abstract _recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent>;
96
96
  /**
97
97
  * Returns a {@link SpeechStream} that can be used to push audio frames and receive
98
98
  * transcriptions
99
+ *
100
+ * @param options - Optional configuration including connection options
99
101
  */
100
- abstract stream(): SpeechStream;
102
+ abstract stream(options?: {
103
+ connOptions?: APIConnectOptions;
104
+ }): SpeechStream;
101
105
  close(): Promise<void>;
102
106
  }
103
107
  /**
@@ -129,12 +133,14 @@ export declare abstract class SpeechStream implements AsyncIterableIterator<Spee
129
133
  private deferredInputStream;
130
134
  private logger;
131
135
  private _connOptions;
136
+ protected abortController: AbortController;
132
137
  constructor(stt: STT, sampleRate?: number, connectionOptions?: APIConnectOptions);
133
138
  private mainTask;
134
139
  private emitError;
135
140
  protected pumpInput(): Promise<void>;
136
141
  protected monitorMetrics(): Promise<void>;
137
142
  protected abstract run(): Promise<void>;
143
+ protected get abortSignal(): AbortSignal;
138
144
  updateInputStream(audioStream: ReadableStream<AudioFrame>): void;
139
145
  detachInputStream(): void;
140
146
  /** Push an audio frame to the STT */
@@ -1 +1 @@
1
- {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/stt/stt.ts"],"names":[],"mappings":";AAGA,OAAO,EAAE,KAAK,UAAU,EAAE,cAAc,EAAE,MAAM,mBAAmB,CAAC;AACpE,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAItD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAErD,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAC/C,OAAO,EAAE,kBAAkB,EAA6B,MAAM,aAAa,CAAC;AAE5E,2CAA2C;AAC3C,oBAAY,eAAe;IACzB;;;;OAIG;IACH,eAAe,IAAI;IACnB;;OAEG;IACH,kBAAkB,IAAI;IACtB;;;OAGG;IACH,gBAAgB,IAAI;IACpB;;;OAGG;IACH,aAAa,IAAI;IACjB,mEAAmE;IACnE,iBAAiB,IAAI;IACrB;;;;OAIG;IACH,oBAAoB,IAAI;CACzB;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,gBAAgB;IAC/B,aAAa,EAAE,MAAM,CAAC;CACvB;AAED,sDAAsD;AACtD,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,eAAe,CAAC;IACtB,YAAY,CAAC,EAAE,CAAC,UAAU,EAAE,GAAG,UAAU,EAAE,CAAC,CAAC;IAC7C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,gBAAgB,CAAC,EAAE,gBAAgB,CAAC;CACrC;AAED;;;;;GAKG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;IACnB,cAAc,EAAE,OAAO,CAAC;CACzB;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IACtF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAGX,YAAY,EAAE,eAAe;IAKzC,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,8FAA8F;IACxF,SAAS,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAezD,SAAS,CAAC,QAAQ,CAAC,UAAU,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAEvE;;;OAGG;IACH,QAAQ,CAAC,MAAM,IAAI,YAAY;IAEzB,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;;;GAeG;AACH,8BAAsB,YAAa,YAAW,qBAAqB,CAAC,WAAW,CAAC;;IAC9E,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,MAAM,kCAAyC;IACzD,SAAS,CAAC,KAAK,kCAAyC;IACxD,SAAS,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;IACpC,SAAS,CAAC,SAAS,CAAC,EAAE,cAAc,CAAC;IACrC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,SAAS,CAAC,MAAM,UAAS;IAEzB,OAAO,CAAC,mBAAmB,CAAqC;IAChE,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,YAAY,CAAoB;gBAGtC,GAAG,EAAE,GAAG,EACR,UAAU,CAAC,EAAE,MAAM,EACnB,iBAAiB,GAAE,iBAA+C;YAgBtD,QAAQ;IAqCtB,OAAO,CAAC,SAAS;cAUD,SAAS;cAkBT,cAAc;IA+B9B,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,iBAAiB,CAAC,WAAW,EAAE,cAAc,CAAC,UAAU,CAAC;IAIzD,iBAAiB;IAIjB,qCAAqC;IACrC,SAAS,CAAC,KAAK,EAAE,UAAU;IAwB3B,4DAA4D;IAC5D,KAAK;IAUL,2DAA2D;IAC3D,QAAQ;IAUR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,WAAW,CAAC,CAAC;IAI5C,wDAAwD;IACxD,KAAK;IAOL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,YAAY;CAGvC"}
1
+ {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/stt/stt.ts"],"names":[],"mappings":";AAGA,OAAO,EAAE,KAAK,UAAU,EAAE,cAAc,EAAE,MAAM,mBAAmB,CAAC;AACpE,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAItD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAErD,OAAO,EAAE,KAAK,iBAAiB,EAAiD,MAAM,aAAa,CAAC;AACpG,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAC/C,OAAO,EAAE,kBAAkB,EAA6B,MAAM,aAAa,CAAC;AAE5E,2CAA2C;AAC3C,oBAAY,eAAe;IACzB;;;;OAIG;IACH,eAAe,IAAI;IACnB;;OAEG;IACH,kBAAkB,IAAI;IACtB;;;OAGG;IACH,gBAAgB,IAAI;IACpB;;;OAGG;IACH,aAAa,IAAI;IACjB,mEAAmE;IACnE,iBAAiB,IAAI;IACrB;;;;OAIG;IACH,oBAAoB,IAAI;CACzB;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,gBAAgB;IAC/B,aAAa,EAAE,MAAM,CAAC;CACvB;AAED,sDAAsD;AACtD,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,eAAe,CAAC;IACtB,YAAY,CAAC,EAAE,CAAC,UAAU,EAAE,GAAG,UAAU,EAAE,CAAC,CAAC;IAC7C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,gBAAgB,CAAC,EAAE,gBAAgB,CAAC;CACrC;AAED;;;;;GAKG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;IACnB,cAAc,EAAE,OAAO,CAAC;CACzB;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IACtF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAGX,YAAY,EAAE,eAAe;IAKzC,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,8FAA8F;IACxF,SAAS,CAAC,KAAK,EAAE,WAAW,EAAE,WAAW,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAgBpF,SAAS,CAAC,QAAQ,CAAC,UAAU,CAC3B,KAAK,EAAE,WAAW,EAClB,WAAW,CAAC,EAAE,WAAW,GACxB,OAAO,CAAC,WAAW,CAAC;IAEvB;;;;;OAKG;IACH,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,YAAY;IAEtE,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;;;GAeG;AACH,8BAAsB,YAAa,YAAW,qBAAqB,CAAC,WAAW,CAAC;;IAC9E,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,MAAM,kCAAyC;IACzD,SAAS,CAAC,KAAK,kCAAyC;IACxD,SAAS,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;IACpC,SAAS,CAAC,SAAS,CAAC,EAAE,cAAc,CAAC;IACrC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,SAAS,CAAC,MAAM,UAAS;IAEzB,OAAO,CAAC,mBAAmB,CAAqC;IAChE,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,YAAY,CAAoB;IAExC,SAAS,CAAC,eAAe,kBAAyB;gBAGhD,GAAG,EAAE,GAAG,EACR,UAAU,CAAC,EAAE,MAAM,EACnB,iBAAiB,GAAE,iBAA+C;YAgBtD,QAAQ;IAqCtB,OAAO,CAAC,SAAS;cAUD,SAAS;cAkBT,cAAc;IA+B9B,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,SAAS,KAAK,WAAW,IAAI,WAAW,CAEvC;IAED,iBAAiB,CAAC,WAAW,EAAE,cAAc,CAAC,UAAU,CAAC;IAIzD,iBAAiB;IAIjB,qCAAqC;IACrC,SAAS,CAAC,KAAK,EAAE,UAAU;IAwB3B,4DAA4D;IAC5D,KAAK;IAUL,2DAA2D;IAC3D,QAAQ;IAUR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,WAAW,CAAC,CAAC;IAI5C,wDAAwD;IACxD,KAAK;IAQL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,YAAY;CAGvC"}
package/dist/stt/stt.js CHANGED
@@ -4,7 +4,7 @@ import { APIConnectionError, APIError } from "../_exceptions.js";
4
4
  import { calculateAudioDurationSeconds } from "../audio.js";
5
5
  import { log } from "../log.js";
6
6
  import { DeferredReadableStream } from "../stream/deferred_stream.js";
7
- import { DEFAULT_API_CONNECT_OPTIONS } from "../types.js";
7
+ import { DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from "../types.js";
8
8
  import { AsyncIterableQueue, delay, startSoon, toError } from "../utils.js";
9
9
  var SpeechEventType = /* @__PURE__ */ ((SpeechEventType2) => {
10
10
  SpeechEventType2[SpeechEventType2["START_OF_SPEECH"] = 0] = "START_OF_SPEECH";
@@ -26,9 +26,9 @@ class STT extends EventEmitter {
26
26
  return this.#capabilities;
27
27
  }
28
28
  /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
29
- async recognize(frame) {
29
+ async recognize(frame, abortSignal) {
30
30
  const startTime = process.hrtime.bigint();
31
- const event = await this._recognize(frame);
31
+ const event = await this._recognize(frame, abortSignal);
32
32
  const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1e6));
33
33
  this.emit("metrics_collected", {
34
34
  type: "stt_metrics",
@@ -57,6 +57,7 @@ class SpeechStream {
57
57
  deferredInputStream;
58
58
  logger = log();
59
59
  _connOptions;
60
+ abortController = new AbortController();
60
61
  constructor(stt, sampleRate, connectionOptions = DEFAULT_API_CONNECT_OPTIONS) {
61
62
  this.#stt = stt;
62
63
  this._connOptions = connectionOptions;
@@ -72,7 +73,7 @@ class SpeechStream {
72
73
  return await this.run();
73
74
  } catch (error) {
74
75
  if (error instanceof APIError) {
75
- const retryInterval = this._connOptions._intervalForRetry(i);
76
+ const retryInterval = intervalForRetry(this._connOptions, i);
76
77
  if (this._connOptions.maxRetry === 0 || !error.retryable) {
77
78
  this.emitError({ error, recoverable: false });
78
79
  throw error;
@@ -152,6 +153,9 @@ class SpeechStream {
152
153
  this.output.close();
153
154
  }
154
155
  }
156
+ get abortSignal() {
157
+ return this.abortController.signal;
158
+ }
155
159
  updateInputStream(audioStream) {
156
160
  this.deferredInputStream.setSource(audioStream);
157
161
  }
@@ -208,6 +212,7 @@ class SpeechStream {
208
212
  if (!this.input.closed) this.input.close();
209
213
  if (!this.queue.closed) this.queue.close();
210
214
  if (!this.output.closed) this.output.close();
215
+ if (!this.abortController.signal.aborted) this.abortController.abort();
211
216
  this.closed = true;
212
217
  }
213
218
  [Symbol.asyncIterator]() {
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n });\n return event;\n }\n protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":"AAGA,SAA0B,sBAAsB;AAEhD,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,qCAAqC;AAC9C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAiC,mCAAmC;AAEpE,SAAS,oBAAoB,OAAO,WAAW,eAAe;AAGvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AAmFL,MAAe,YAAa,aAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,MAAM,8BAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EASA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,mBAAoE;AAAA,EAChF,SAAS,IAAI,mBAAgC;AAAA,EAC7C,QAAQ,IAAI,mBAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EACb;AAAA,EAER,YACE,KACA,YACA,oBAAuC,6BACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,uBAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,eAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
1
+ {"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame, abortSignal);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n });\n return event;\n }\n\n protected abstract _recognize(\n frame: AudioBuffer,\n abortSignal?: AbortSignal,\n ): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n\n protected abortController = new AbortController();\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n protected get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":"AAGA,SAA0B,sBAAsB;AAEhD,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,qCAAqC;AAC9C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAiC,6BAA6B,wBAAwB;AAEtF,SAAS,oBAAoB,OAAO,WAAW,eAAe;AAGvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AAmFL,MAAe,YAAa,aAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAAoB,aAAiD;AACnF,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,OAAO,WAAW;AACtD,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,MAAM,8BAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAeA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,mBAAoE;AAAA,EAChF,SAAS,IAAI,mBAAgC;AAAA,EAC7C,QAAQ,IAAI,mBAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EACb;AAAA,EAEE,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,KACA,YACA,oBAAuC,6BACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,uBAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,IAAc,cAA2B;AACvC,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,eAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
@@ -45,6 +45,7 @@ var import_sdk_trace_node = require("@opentelemetry/sdk-trace-node");
45
45
  var import_semantic_conventions = require("@opentelemetry/semantic-conventions");
46
46
  var import_form_data = __toESM(require("form-data"), 1);
47
47
  var import_livekit_server_sdk = require("livekit-server-sdk");
48
+ var import_promises = __toESM(require("node:fs/promises"), 1);
48
49
  var import_log = require("../log.cjs");
49
50
  var import_otel_http_exporter = require("./otel_http_exporter.cjs");
50
51
  var import_pino_otel_transport = require("./pino_otel_transport.cjs");
@@ -336,13 +337,14 @@ async function uploadSessionReport(options) {
336
337
  token.addObservabilityGrant({ write: true });
337
338
  const jwt = await token.toJwt();
338
339
  const formData = new import_form_data.default();
340
+ const audioStartTime = report.audioRecordingStartedAt ?? 0;
339
341
  const headerMsg = new import_protocol.MetricsRecordingHeader({
340
342
  roomId: report.roomId,
341
343
  duration: BigInt(0),
342
344
  // TODO: Calculate actual duration from report
343
345
  startTime: {
344
- seconds: BigInt(Math.floor(report.timestamp / 1e3)),
345
- nanos: Math.floor(report.timestamp % 1e3 * 1e6)
346
+ seconds: BigInt(Math.floor(audioStartTime / 1e3)),
347
+ nanos: Math.floor(audioStartTime % 1e3 * 1e6)
346
348
  }
347
349
  });
348
350
  const headerBytes = Buffer.from(headerMsg.toBinary());
@@ -366,6 +368,25 @@ async function uploadSessionReport(options) {
366
368
  "Content-Length": chatHistoryBuffer.length.toString()
367
369
  }
368
370
  });
371
+ if (report.audioRecordingPath && report.audioRecordingStartedAt) {
372
+ let audioBytes;
373
+ try {
374
+ audioBytes = await import_promises.default.readFile(report.audioRecordingPath);
375
+ } catch {
376
+ audioBytes = Buffer.alloc(0);
377
+ }
378
+ if (audioBytes.length > 0) {
379
+ formData.append("audio", audioBytes, {
380
+ filename: "recording.ogg",
381
+ contentType: "audio/ogg",
382
+ knownLength: audioBytes.length,
383
+ header: {
384
+ "Content-Type": "audio/ogg",
385
+ "Content-Length": audioBytes.length.toString()
386
+ }
387
+ });
388
+ }
389
+ }
369
390
  return new Promise((resolve, reject) => {
370
391
  formData.submit(
371
392
  {
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/telemetry/traces.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { MetricsRecordingHeader } from '@livekit/protocol';\nimport {\n type Attributes,\n type Context,\n type Span,\n type SpanOptions,\n type Tracer,\n type TracerProvider,\n context as otelContext,\n trace,\n} from '@opentelemetry/api';\nimport { SeverityNumber } from '@opentelemetry/api-logs';\nimport { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-proto';\nimport { CompressionAlgorithm } from '@opentelemetry/otlp-exporter-base';\nimport { Resource } from '@opentelemetry/resources';\nimport type { ReadableSpan, SpanProcessor } from '@opentelemetry/sdk-trace-base';\nimport { BatchSpanProcessor, NodeTracerProvider } from '@opentelemetry/sdk-trace-node';\nimport { ATTR_SERVICE_NAME } from '@opentelemetry/semantic-conventions';\nimport FormData from 'form-data';\nimport { AccessToken } from 'livekit-server-sdk';\nimport type { ChatContent, ChatItem } from '../llm/index.js';\nimport { enableOtelLogging } from '../log.js';\nimport type { SessionReport } from '../voice/report.js';\nimport { type SimpleLogRecord, SimpleOTLPHttpLogExporter } from './otel_http_exporter.js';\nimport { flushPinoLogs, initPinoCloudExporter } from './pino_otel_transport.js';\n\nexport interface StartSpanOptions {\n /** Name of the span */\n name: string;\n /** Optional parent context to use for this span */\n context?: Context;\n /** Attributes to set on the span when it starts */\n attributes?: Attributes;\n /** Whether to end the span when the function exits (default: true) */\n endOnExit?: boolean;\n}\n\n/**\n * A dynamic tracer that allows the tracer provider to be changed at runtime.\n */\nclass DynamicTracer {\n private tracerProvider: TracerProvider;\n private tracer: Tracer;\n private readonly instrumentingModuleName: string;\n\n constructor(instrumentingModuleName: string) {\n this.instrumentingModuleName = instrumentingModuleName;\n this.tracerProvider = trace.getTracerProvider();\n this.tracer = trace.getTracer(instrumentingModuleName);\n }\n\n /**\n * Set a new tracer provider. This updates the underlying tracer instance.\n * @param provider - The new tracer provider to use\n */\n setProvider(provider: TracerProvider): void {\n this.tracerProvider = provider;\n this.tracer = this.tracerProvider.getTracer(this.instrumentingModuleName);\n }\n\n /**\n * Get the underlying OpenTelemetry tracer.\n * Use this to access the full Tracer API when needed.\n */\n getTracer(): Tracer {\n return this.tracer;\n }\n\n /**\n * Start a span manually (without making it active).\n * You must call span.end() when done.\n *\n * @param options - Span configuration including name\n * @returns The created span\n */\n startSpan(options: StartSpanOptions): Span {\n const ctx = options.context || otelContext.active();\n const span = this.tracer.startSpan(\n options.name,\n {\n attributes: options.attributes,\n },\n ctx,\n );\n\n return span;\n }\n\n /**\n * Start a new span and make it active in the current context.\n * The span will automatically be ended when the provided function completes (unless endOnExit=false).\n *\n * @param fn - The function to execute within the span context\n * @param options - Span configuration including name\n * @returns The result of the provided function\n */\n async startActiveSpan<T>(fn: (span: Span) => Promise<T>, options: StartSpanOptions): Promise<T> {\n const ctx = options.context || otelContext.active();\n const endOnExit = options.endOnExit === undefined ? true : options.endOnExit; // default true\n const opts: SpanOptions = { attributes: options.attributes };\n\n // Directly return the tracer's startActiveSpan result - it handles async correctly\n return await this.tracer.startActiveSpan(options.name, opts, ctx, async (span) => {\n try {\n return await fn(span);\n } finally {\n if (endOnExit) {\n span.end();\n }\n }\n });\n }\n\n /**\n * Synchronous version of startActiveSpan for non-async operations.\n *\n * @param fn - The function to execute within the span context\n * @param options - Span configuration including name\n * @returns The result of the provided function\n */\n startActiveSpanSync<T>(fn: (span: Span) => T, options: StartSpanOptions): T {\n const ctx = options.context || otelContext.active();\n const endOnExit = options.endOnExit === undefined ? true : options.endOnExit; // default true\n const opts: SpanOptions = { attributes: options.attributes };\n\n return this.tracer.startActiveSpan(options.name, opts, ctx, (span) => {\n try {\n return fn(span);\n } finally {\n if (endOnExit) {\n span.end();\n }\n }\n });\n }\n}\n\n/**\n * The global tracer instance used throughout the agents framework.\n * This tracer can have its provider updated at runtime via setTracerProvider().\n */\nexport const tracer = new DynamicTracer('livekit-agents');\n\nclass MetadataSpanProcessor implements SpanProcessor {\n private metadata: Attributes;\n\n constructor(metadata: Attributes) {\n this.metadata = metadata;\n }\n\n onStart(span: Span, _parentContext: Context): void {\n span.setAttributes(this.metadata);\n }\n\n onEnd(_span: ReadableSpan): void {}\n\n shutdown(): Promise<void> {\n return Promise.resolve();\n }\n\n forceFlush(): Promise<void> {\n return Promise.resolve();\n }\n}\n\n/**\n * Set the tracer provider for the livekit-agents framework.\n * This should be called before agent session start if using custom tracer providers.\n *\n * @param provider - The tracer provider to use (must be a NodeTracerProvider)\n * @param options - Optional configuration with metadata property to inject into all spans\n *\n * @example\n * ```typescript\n * import { NodeTracerProvider } from '@opentelemetry/sdk-trace-node';\n * import { setTracerProvider } from '@livekit/agents/telemetry';\n *\n * const provider = new NodeTracerProvider();\n * setTracerProvider(provider, {\n * metadata: { room_id: 'room123', job_id: 'job456' }\n * });\n * ```\n */\nexport function setTracerProvider(\n provider: NodeTracerProvider,\n options?: { metadata?: Attributes },\n): void {\n if (options?.metadata) {\n provider.addSpanProcessor(new MetadataSpanProcessor(options.metadata));\n }\n\n tracer.setProvider(provider);\n}\n\n/**\n * Setup OpenTelemetry tracer for LiveKit Cloud observability.\n * This configures OTLP exporters to send traces to LiveKit Cloud.\n *\n * @param options - Configuration for cloud tracer with roomId, jobId, and cloudHostname properties\n *\n * @internal\n */\nexport async function setupCloudTracer(options: {\n roomId: string;\n jobId: string;\n cloudHostname: string;\n}): Promise<void> {\n const { roomId, jobId, cloudHostname } = options;\n\n const apiKey = process.env.LIVEKIT_API_KEY;\n const apiSecret = process.env.LIVEKIT_API_SECRET;\n\n if (!apiKey || !apiSecret) {\n throw new Error('LIVEKIT_API_KEY and LIVEKIT_API_SECRET must be set for cloud tracing');\n }\n\n const token = new AccessToken(apiKey, apiSecret, {\n identity: 'livekit-agents-telemetry',\n ttl: '6h',\n });\n token.addObservabilityGrant({ write: true });\n\n try {\n const jwt = await token.toJwt();\n\n const headers = {\n Authorization: `Bearer ${jwt}`,\n };\n\n const metadata: Attributes = {\n room_id: roomId,\n job_id: jobId,\n };\n\n const resource = new Resource({\n [ATTR_SERVICE_NAME]: 'livekit-agents',\n room_id: roomId,\n job_id: jobId,\n });\n\n // Configure OTLP exporter to send traces to LiveKit Cloud\n const spanExporter = new OTLPTraceExporter({\n url: `https://${cloudHostname}/observability/traces/otlp/v0`,\n headers,\n compression: CompressionAlgorithm.GZIP,\n });\n\n const tracerProvider = new NodeTracerProvider({\n resource,\n spanProcessors: [new MetadataSpanProcessor(metadata), new BatchSpanProcessor(spanExporter)],\n });\n tracerProvider.register();\n\n setTracerProvider(tracerProvider);\n\n // Initialize standalone Pino cloud exporter (no OTEL SDK dependency)\n initPinoCloudExporter({\n cloudHostname,\n roomId,\n jobId,\n });\n\n enableOtelLogging();\n } catch (error) {\n console.error('Failed to setup cloud tracer:', error);\n throw error;\n }\n}\n\n/**\n * Flush all pending Pino logs to ensure they are exported.\n * Call this before session/job ends to ensure all logs are sent.\n *\n * @internal\n */\nexport async function flushOtelLogs(): Promise<void> {\n await flushPinoLogs();\n}\n\n/**\n * Convert ChatItem to proto-compatible dictionary format.\n * TODO: Use actual agent_session proto types once @livekit/protocol v1.43.1+ is published\n */\nfunction chatItemToProto(item: ChatItem): Record<string, any> {\n const itemDict: Record<string, any> = {};\n\n if (item.type === 'message') {\n const roleMap: Record<string, string> = {\n developer: 'DEVELOPER',\n system: 'SYSTEM',\n user: 'USER',\n assistant: 'ASSISTANT',\n };\n\n const msg: Record<string, any> = {\n id: item.id,\n role: roleMap[item.role] || item.role.toUpperCase(),\n content: item.content.map((c: ChatContent) => ({ text: c })),\n createdAt: toRFC3339(item.createdAt),\n };\n\n if (item.interrupted) {\n msg.interrupted = item.interrupted;\n }\n\n // TODO(brian): Add extra and transcriptConfidence to ChatMessage\n // if (item.extra && Object.keys(item.extra).length > 0) {\n // msg.extra = item.extra;\n // }\n\n // if (item.transcriptConfidence !== undefined && item.transcriptConfidence !== null) {\n // msg.transcriptConfidence = item.transcriptConfidence;\n // }\n\n // TODO(brian): Add metrics to ChatMessage\n // const metrics = item.metrics || {};\n // if (Object.keys(metrics).length > 0) {\n // msg.metrics = {};\n // if (metrics.started_speaking_at) {\n // msg.metrics.startedSpeakingAt = toRFC3339(metrics.started_speaking_at);\n // }\n // if (metrics.stopped_speaking_at) {\n // msg.metrics.stoppedSpeakingAt = toRFC3339(metrics.stopped_speaking_at);\n // }\n // if (metrics.transcription_delay !== undefined) {\n // msg.metrics.transcriptionDelay = metrics.transcription_delay;\n // }\n // if (metrics.end_of_turn_delay !== undefined) {\n // msg.metrics.endOfTurnDelay = metrics.end_of_turn_delay;\n // }\n // if (metrics.on_user_turn_completed_delay !== undefined) {\n // msg.metrics.onUserTurnCompletedDelay = metrics.on_user_turn_completed_delay;\n // }\n // if (metrics.llm_node_ttft !== undefined) {\n // msg.metrics.llmNodeTtft = metrics.llm_node_ttft;\n // }\n // if (metrics.tts_node_ttfb !== undefined) {\n // msg.metrics.ttsNodeTtfb = metrics.tts_node_ttfb;\n // }\n // if (metrics.e2e_latency !== undefined) {\n // msg.metrics.e2eLatency = metrics.e2e_latency;\n // }\n // }\n\n itemDict.message = msg;\n } else if (item.type === 'function_call') {\n itemDict.functionCall = {\n id: item.id,\n callId: item.callId,\n arguments: item.args,\n name: item.name,\n createdAt: toRFC3339(item.createdAt),\n };\n } else if (item.type === 'function_call_output') {\n itemDict.functionCallOutput = {\n id: item.id,\n name: item.name,\n callId: item.callId,\n output: item.output,\n isError: item.isError,\n createdAt: toRFC3339(item.createdAt),\n };\n } else if (item.type === 'agent_handoff') {\n const handoff: Record<string, any> = {\n id: item.id,\n newAgentId: item.newAgentId,\n createdAt: toRFC3339(item.createdAt),\n };\n if (item.oldAgentId !== undefined && item.oldAgentId !== null && item.oldAgentId !== '') {\n handoff.oldAgentId = item.oldAgentId;\n }\n itemDict.agentHandoff = handoff;\n }\n\n try {\n if (item.type === 'function_call' && typeof itemDict.functionCall?.arguments === 'string') {\n itemDict.functionCall.arguments = JSON.parse(itemDict.functionCall.arguments);\n } else if (\n item.type === 'function_call_output' &&\n typeof itemDict.functionCallOutput?.output === 'string'\n ) {\n itemDict.functionCallOutput.output = JSON.parse(itemDict.functionCallOutput.output);\n }\n } catch {\n // ignore parsing errors\n }\n\n return itemDict;\n}\n\n/**\n * Convert timestamp to RFC3339 format matching Python's _to_rfc3339.\n * Note: TypeScript createdAt is in milliseconds (Date.now()), not seconds like Python.\n * @internal\n */\nfunction toRFC3339(valueMs: number | Date): string {\n // valueMs is already in milliseconds (from Date.now())\n const dt = valueMs instanceof Date ? valueMs : new Date(valueMs);\n // Truncate sub-millisecond precision\n const truncated = new Date(Math.floor(dt.getTime()));\n return truncated.toISOString();\n}\n\n/**\n * Upload session report to LiveKit Cloud observability.\n * @param options - Configuration with agentName, cloudHostname, and report\n */\nexport async function uploadSessionReport(options: {\n agentName: string;\n cloudHostname: string;\n report: SessionReport;\n}): Promise<void> {\n const { agentName, cloudHostname, report } = options;\n\n // Create OTLP HTTP exporter for chat history logs\n // Uses raw HTTP JSON format which is required by LiveKit Cloud\n const logExporter = new SimpleOTLPHttpLogExporter({\n cloudHostname,\n resourceAttributes: {\n room_id: report.roomId,\n job_id: report.jobId,\n },\n scopeName: 'chat_history',\n scopeAttributes: {\n room_id: report.roomId,\n job_id: report.jobId,\n room: report.room,\n },\n });\n\n // Build log records for session report and chat items\n const logRecords: SimpleLogRecord[] = [];\n\n const commonAttrs = {\n room_id: report.roomId,\n job_id: report.jobId,\n 'logger.name': 'chat_history',\n };\n\n logRecords.push({\n body: 'session report',\n timestampMs: report.startedAt || report.timestamp || 0,\n attributes: {\n ...commonAttrs,\n 'session.options': report.options || {},\n 'session.report_timestamp': report.timestamp,\n agent_name: agentName,\n },\n });\n\n // Track last timestamp to ensure monotonic ordering when items have identical timestamps\n // This fixes the issue where function_call and function_call_output with same timestamp\n // get reordered by the dashboard\n let lastTimestamp = 0;\n for (const item of report.chatHistory.items) {\n // Ensure monotonically increasing timestamps for proper ordering\n // Add 0.001ms (1 microsecond) offset when timestamps collide\n let itemTimestamp = item.createdAt;\n if (itemTimestamp <= lastTimestamp) {\n itemTimestamp = lastTimestamp + 0.001; // Add 1 microsecond\n }\n lastTimestamp = itemTimestamp;\n\n const itemProto = chatItemToProto(item);\n let severityNumber = SeverityNumber.UNSPECIFIED;\n let severityText = 'unspecified';\n\n if (item.type === 'function_call_output' && item.isError) {\n severityNumber = SeverityNumber.ERROR;\n severityText = 'error';\n }\n\n logRecords.push({\n body: 'chat item',\n timestampMs: itemTimestamp, // Adjusted for monotonic ordering\n attributes: { 'chat.item': itemProto, ...commonAttrs },\n severityNumber,\n severityText,\n });\n }\n await logExporter.export(logRecords);\n\n const apiKey = process.env.LIVEKIT_API_KEY;\n const apiSecret = process.env.LIVEKIT_API_SECRET;\n\n if (!apiKey || !apiSecret) {\n throw new Error('LIVEKIT_API_KEY and LIVEKIT_API_SECRET must be set for session upload');\n }\n\n const token = new AccessToken(apiKey, apiSecret, { ttl: '6h' });\n token.addObservabilityGrant({ write: true });\n const jwt = await token.toJwt();\n\n const formData = new FormData();\n\n // Add header (protobuf MetricsRecordingHeader)\n const headerMsg = new MetricsRecordingHeader({\n roomId: report.roomId,\n duration: BigInt(0), // TODO: Calculate actual duration from report\n startTime: {\n seconds: BigInt(Math.floor(report.timestamp / 1000)),\n nanos: Math.floor((report.timestamp % 1000) * 1e6),\n },\n });\n\n const headerBytes = Buffer.from(headerMsg.toBinary());\n formData.append('header', headerBytes, {\n filename: 'header.binpb',\n contentType: 'application/protobuf',\n knownLength: headerBytes.length,\n header: {\n 'Content-Type': 'application/protobuf',\n 'Content-Length': headerBytes.length.toString(),\n },\n });\n\n // Add chat_history JSON\n const chatHistoryJson = JSON.stringify(report.chatHistory.toJSON({ excludeTimestamp: false }));\n const chatHistoryBuffer = Buffer.from(chatHistoryJson, 'utf-8');\n formData.append('chat_history', chatHistoryBuffer, {\n filename: 'chat_history.json',\n contentType: 'application/json',\n knownLength: chatHistoryBuffer.length,\n header: {\n 'Content-Type': 'application/json',\n 'Content-Length': chatHistoryBuffer.length.toString(),\n },\n });\n\n // TODO(brian): Add audio recording file when recorder IO is implemented\n\n // Upload to LiveKit Cloud using form-data's submit method\n // This properly streams the multipart form with all headers including Content-Length\n return new Promise<void>((resolve, reject) => {\n formData.submit(\n {\n protocol: 'https:',\n host: cloudHostname,\n path: '/observability/recordings/v0',\n method: 'POST',\n headers: {\n Authorization: `Bearer ${jwt}`,\n },\n },\n (err, res) => {\n if (err) {\n reject(new Error(`Failed to upload session report: ${err.message}`));\n return;\n }\n\n if (res.statusCode && res.statusCode >= 400) {\n reject(\n new Error(`Failed to upload session report: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.resume(); // Drain the response\n res.on('end', () => resolve());\n },\n );\n });\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAuC;AACvC,iBASO;AACP,sBAA+B;AAC/B,uCAAkC;AAClC,gCAAqC;AACrC,uBAAyB;AAEzB,4BAAuD;AACvD,kCAAkC;AAClC,uBAAqB;AACrB,gCAA4B;AAE5B,iBAAkC;AAElC,gCAAgE;AAChE,iCAAqD;AAgBrD,MAAM,cAAc;AAAA,EACV;AAAA,EACA;AAAA,EACS;AAAA,EAEjB,YAAY,yBAAiC;AAC3C,SAAK,0BAA0B;AAC/B,SAAK,iBAAiB,iBAAM,kBAAkB;AAC9C,SAAK,SAAS,iBAAM,UAAU,uBAAuB;AAAA,EACvD;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,YAAY,UAAgC;AAC1C,SAAK,iBAAiB;AACtB,SAAK,SAAS,KAAK,eAAe,UAAU,KAAK,uBAAuB;AAAA,EAC1E;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,YAAoB;AAClB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,UAAU,SAAiC;AACzC,UAAM,MAAM,QAAQ,WAAW,WAAAA,QAAY,OAAO;AAClD,UAAM,OAAO,KAAK,OAAO;AAAA,MACvB,QAAQ;AAAA,MACR;AAAA,QACE,YAAY,QAAQ;AAAA,MACtB;AAAA,MACA;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,gBAAmB,IAAgC,SAAuC;AAC9F,UAAM,MAAM,QAAQ,WAAW,WAAAA,QAAY,OAAO;AAClD,UAAM,YAAY,QAAQ,cAAc,SAAY,OAAO,QAAQ;AACnE,UAAM,OAAoB,EAAE,YAAY,QAAQ,WAAW;AAG3D,WAAO,MAAM,KAAK,OAAO,gBAAgB,QAAQ,MAAM,MAAM,KAAK,OAAO,SAAS;AAChF,UAAI;AACF,eAAO,MAAM,GAAG,IAAI;AAAA,MACtB,UAAE;AACA,YAAI,WAAW;AACb,eAAK,IAAI;AAAA,QACX;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,oBAAuB,IAAuB,SAA8B;AAC1E,UAAM,MAAM,QAAQ,WAAW,WAAAA,QAAY,OAAO;AAClD,UAAM,YAAY,QAAQ,cAAc,SAAY,OAAO,QAAQ;AACnE,UAAM,OAAoB,EAAE,YAAY,QAAQ,WAAW;AAE3D,WAAO,KAAK,OAAO,gBAAgB,QAAQ,MAAM,MAAM,KAAK,CAAC,SAAS;AACpE,UAAI;AACF,eAAO,GAAG,IAAI;AAAA,MAChB,UAAE;AACA,YAAI,WAAW;AACb,eAAK,IAAI;AAAA,QACX;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AACF;AAMO,MAAM,SAAS,IAAI,cAAc,gBAAgB;AAExD,MAAM,sBAA+C;AAAA,EAC3C;AAAA,EAER,YAAY,UAAsB;AAChC,SAAK,WAAW;AAAA,EAClB;AAAA,EAEA,QAAQ,MAAY,gBAA+B;AACjD,SAAK,cAAc,KAAK,QAAQ;AAAA,EAClC;AAAA,EAEA,MAAM,OAA2B;AAAA,EAAC;AAAA,EAElC,WAA0B;AACxB,WAAO,QAAQ,QAAQ;AAAA,EACzB;AAAA,EAEA,aAA4B;AAC1B,WAAO,QAAQ,QAAQ;AAAA,EACzB;AACF;AAoBO,SAAS,kBACd,UACA,SACM;AACN,MAAI,mCAAS,UAAU;AACrB,aAAS,iBAAiB,IAAI,sBAAsB,QAAQ,QAAQ,CAAC;AAAA,EACvE;AAEA,SAAO,YAAY,QAAQ;AAC7B;AAUA,eAAsB,iBAAiB,SAIrB;AAChB,QAAM,EAAE,QAAQ,OAAO,cAAc,IAAI;AAEzC,QAAM,SAAS,QAAQ,IAAI;AAC3B,QAAM,YAAY,QAAQ,IAAI;AAE9B,MAAI,CAAC,UAAU,CAAC,WAAW;AACzB,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAEA,QAAM,QAAQ,IAAI,sCAAY,QAAQ,WAAW;AAAA,IAC/C,UAAU;AAAA,IACV,KAAK;AAAA,EACP,CAAC;AACD,QAAM,sBAAsB,EAAE,OAAO,KAAK,CAAC;AAE3C,MAAI;AACF,UAAM,MAAM,MAAM,MAAM,MAAM;AAE9B,UAAM,UAAU;AAAA,MACd,eAAe,UAAU,GAAG;AAAA,IAC9B;AAEA,UAAM,WAAuB;AAAA,MAC3B,SAAS;AAAA,MACT,QAAQ;AAAA,IACV;AAEA,UAAM,WAAW,IAAI,0BAAS;AAAA,MAC5B,CAAC,6CAAiB,GAAG;AAAA,MACrB,SAAS;AAAA,MACT,QAAQ;AAAA,IACV,CAAC;AAGD,UAAM,eAAe,IAAI,mDAAkB;AAAA,MACzC,KAAK,WAAW,aAAa;AAAA,MAC7B;AAAA,MACA,aAAa,+CAAqB;AAAA,IACpC,CAAC;AAED,UAAM,iBAAiB,IAAI,yCAAmB;AAAA,MAC5C;AAAA,MACA,gBAAgB,CAAC,IAAI,sBAAsB,QAAQ,GAAG,IAAI,yCAAmB,YAAY,CAAC;AAAA,IAC5F,CAAC;AACD,mBAAe,SAAS;AAExB,sBAAkB,cAAc;AAGhC,0DAAsB;AAAA,MACpB;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAED,sCAAkB;AAAA,EACpB,SAAS,OAAO;AACd,YAAQ,MAAM,iCAAiC,KAAK;AACpD,UAAM;AAAA,EACR;AACF;AAQA,eAAsB,gBAA+B;AACnD,YAAM,0CAAc;AACtB;AAMA,SAAS,gBAAgB,MAAqC;AA9R9D;AA+RE,QAAM,WAAgC,CAAC;AAEvC,MAAI,KAAK,SAAS,WAAW;AAC3B,UAAM,UAAkC;AAAA,MACtC,WAAW;AAAA,MACX,QAAQ;AAAA,MACR,MAAM;AAAA,MACN,WAAW;AAAA,IACb;AAEA,UAAM,MAA2B;AAAA,MAC/B,IAAI,KAAK;AAAA,MACT,MAAM,QAAQ,KAAK,IAAI,KAAK,KAAK,KAAK,YAAY;AAAA,MAClD,SAAS,KAAK,QAAQ,IAAI,CAAC,OAAoB,EAAE,MAAM,EAAE,EAAE;AAAA,MAC3D,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AAEA,QAAI,KAAK,aAAa;AACpB,UAAI,cAAc,KAAK;AAAA,IACzB;AAyCA,aAAS,UAAU;AAAA,EACrB,WAAW,KAAK,SAAS,iBAAiB;AACxC,aAAS,eAAe;AAAA,MACtB,IAAI,KAAK;AAAA,MACT,QAAQ,KAAK;AAAA,MACb,WAAW,KAAK;AAAA,MAChB,MAAM,KAAK;AAAA,MACX,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AAAA,EACF,WAAW,KAAK,SAAS,wBAAwB;AAC/C,aAAS,qBAAqB;AAAA,MAC5B,IAAI,KAAK;AAAA,MACT,MAAM,KAAK;AAAA,MACX,QAAQ,KAAK;AAAA,MACb,QAAQ,KAAK;AAAA,MACb,SAAS,KAAK;AAAA,MACd,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AAAA,EACF,WAAW,KAAK,SAAS,iBAAiB;AACxC,UAAM,UAA+B;AAAA,MACnC,IAAI,KAAK;AAAA,MACT,YAAY,KAAK;AAAA,MACjB,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AACA,QAAI,KAAK,eAAe,UAAa,KAAK,eAAe,QAAQ,KAAK,eAAe,IAAI;AACvF,cAAQ,aAAa,KAAK;AAAA,IAC5B;AACA,aAAS,eAAe;AAAA,EAC1B;AAEA,MAAI;AACF,QAAI,KAAK,SAAS,mBAAmB,SAAO,cAAS,iBAAT,mBAAuB,eAAc,UAAU;AACzF,eAAS,aAAa,YAAY,KAAK,MAAM,SAAS,aAAa,SAAS;AAAA,IAC9E,WACE,KAAK,SAAS,0BACd,SAAO,cAAS,uBAAT,mBAA6B,YAAW,UAC/C;AACA,eAAS,mBAAmB,SAAS,KAAK,MAAM,SAAS,mBAAmB,MAAM;AAAA,IACpF;AAAA,EACF,QAAQ;AAAA,EAER;AAEA,SAAO;AACT;AAOA,SAAS,UAAU,SAAgC;AAEjD,QAAM,KAAK,mBAAmB,OAAO,UAAU,IAAI,KAAK,OAAO;AAE/D,QAAM,YAAY,IAAI,KAAK,KAAK,MAAM,GAAG,QAAQ,CAAC,CAAC;AACnD,SAAO,UAAU,YAAY;AAC/B;AAMA,eAAsB,oBAAoB,SAIxB;AAChB,QAAM,EAAE,WAAW,eAAe,OAAO,IAAI;AAI7C,QAAM,cAAc,IAAI,oDAA0B;AAAA,IAChD;AAAA,IACA,oBAAoB;AAAA,MAClB,SAAS,OAAO;AAAA,MAChB,QAAQ,OAAO;AAAA,IACjB;AAAA,IACA,WAAW;AAAA,IACX,iBAAiB;AAAA,MACf,SAAS,OAAO;AAAA,MAChB,QAAQ,OAAO;AAAA,MACf,MAAM,OAAO;AAAA,IACf;AAAA,EACF,CAAC;AAGD,QAAM,aAAgC,CAAC;AAEvC,QAAM,cAAc;AAAA,IAClB,SAAS,OAAO;AAAA,IAChB,QAAQ,OAAO;AAAA,IACf,eAAe;AAAA,EACjB;AAEA,aAAW,KAAK;AAAA,IACd,MAAM;AAAA,IACN,aAAa,OAAO,aAAa,OAAO,aAAa;AAAA,IACrD,YAAY;AAAA,MACV,GAAG;AAAA,MACH,mBAAmB,OAAO,WAAW,CAAC;AAAA,MACtC,4BAA4B,OAAO;AAAA,MACnC,YAAY;AAAA,IACd;AAAA,EACF,CAAC;AAKD,MAAI,gBAAgB;AACpB,aAAW,QAAQ,OAAO,YAAY,OAAO;AAG3C,QAAI,gBAAgB,KAAK;AACzB,QAAI,iBAAiB,eAAe;AAClC,sBAAgB,gBAAgB;AAAA,IAClC;AACA,oBAAgB;AAEhB,UAAM,YAAY,gBAAgB,IAAI;AACtC,QAAI,iBAAiB,+BAAe;AACpC,QAAI,eAAe;AAEnB,QAAI,KAAK,SAAS,0BAA0B,KAAK,SAAS;AACxD,uBAAiB,+BAAe;AAChC,qBAAe;AAAA,IACjB;AAEA,eAAW,KAAK;AAAA,MACd,MAAM;AAAA,MACN,aAAa;AAAA;AAAA,MACb,YAAY,EAAE,aAAa,WAAW,GAAG,YAAY;AAAA,MACrD;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AACA,QAAM,YAAY,OAAO,UAAU;AAEnC,QAAM,SAAS,QAAQ,IAAI;AAC3B,QAAM,YAAY,QAAQ,IAAI;AAE9B,MAAI,CAAC,UAAU,CAAC,WAAW;AACzB,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AAEA,QAAM,QAAQ,IAAI,sCAAY,QAAQ,WAAW,EAAE,KAAK,KAAK,CAAC;AAC9D,QAAM,sBAAsB,EAAE,OAAO,KAAK,CAAC;AAC3C,QAAM,MAAM,MAAM,MAAM,MAAM;AAE9B,QAAM,WAAW,IAAI,iBAAAC,QAAS;AAG9B,QAAM,YAAY,IAAI,uCAAuB;AAAA,IAC3C,QAAQ,OAAO;AAAA,IACf,UAAU,OAAO,CAAC;AAAA;AAAA,IAClB,WAAW;AAAA,MACT,SAAS,OAAO,KAAK,MAAM,OAAO,YAAY,GAAI,CAAC;AAAA,MACnD,OAAO,KAAK,MAAO,OAAO,YAAY,MAAQ,GAAG;AAAA,IACnD;AAAA,EACF,CAAC;AAED,QAAM,cAAc,OAAO,KAAK,UAAU,SAAS,CAAC;AACpD,WAAS,OAAO,UAAU,aAAa;AAAA,IACrC,UAAU;AAAA,IACV,aAAa;AAAA,IACb,aAAa,YAAY;AAAA,IACzB,QAAQ;AAAA,MACN,gBAAgB;AAAA,MAChB,kBAAkB,YAAY,OAAO,SAAS;AAAA,IAChD;AAAA,EACF,CAAC;AAGD,QAAM,kBAAkB,KAAK,UAAU,OAAO,YAAY,OAAO,EAAE,kBAAkB,MAAM,CAAC,CAAC;AAC7F,QAAM,oBAAoB,OAAO,KAAK,iBAAiB,OAAO;AAC9D,WAAS,OAAO,gBAAgB,mBAAmB;AAAA,IACjD,UAAU;AAAA,IACV,aAAa;AAAA,IACb,aAAa,kBAAkB;AAAA,IAC/B,QAAQ;AAAA,MACN,gBAAgB;AAAA,MAChB,kBAAkB,kBAAkB,OAAO,SAAS;AAAA,IACtD;AAAA,EACF,CAAC;AAMD,SAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,MAAM;AAAA,QACN,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,eAAe,UAAU,GAAG;AAAA,QAC9B;AAAA,MACF;AAAA,MACA,CAAC,KAAK,QAAQ;AACZ,YAAI,KAAK;AACP,iBAAO,IAAI,MAAM,oCAAoC,IAAI,OAAO,EAAE,CAAC;AACnE;AAAA,QACF;AAEA,YAAI,IAAI,cAAc,IAAI,cAAc,KAAK;AAC3C;AAAA,YACE,IAAI,MAAM,oCAAoC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,UACrF;AACA;AAAA,QACF;AAEA,YAAI,OAAO;AACX,YAAI,GAAG,OAAO,MAAM,QAAQ,CAAC;AAAA,MAC/B;AAAA,IACF;AAAA,EACF,CAAC;AACH;","names":["otelContext","FormData"]}
1
+ {"version":3,"sources":["../../src/telemetry/traces.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { MetricsRecordingHeader } from '@livekit/protocol';\nimport {\n type Attributes,\n type Context,\n type Span,\n type SpanOptions,\n type Tracer,\n type TracerProvider,\n context as otelContext,\n trace,\n} from '@opentelemetry/api';\nimport { SeverityNumber } from '@opentelemetry/api-logs';\nimport { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-proto';\nimport { CompressionAlgorithm } from '@opentelemetry/otlp-exporter-base';\nimport { Resource } from '@opentelemetry/resources';\nimport type { ReadableSpan, SpanProcessor } from '@opentelemetry/sdk-trace-base';\nimport { BatchSpanProcessor, NodeTracerProvider } from '@opentelemetry/sdk-trace-node';\nimport { ATTR_SERVICE_NAME } from '@opentelemetry/semantic-conventions';\nimport FormData from 'form-data';\nimport { AccessToken } from 'livekit-server-sdk';\nimport fs from 'node:fs/promises';\nimport type { ChatContent, ChatItem } from '../llm/index.js';\nimport { enableOtelLogging } from '../log.js';\nimport type { SessionReport } from '../voice/report.js';\nimport { type SimpleLogRecord, SimpleOTLPHttpLogExporter } from './otel_http_exporter.js';\nimport { flushPinoLogs, initPinoCloudExporter } from './pino_otel_transport.js';\n\nexport interface StartSpanOptions {\n /** Name of the span */\n name: string;\n /** Optional parent context to use for this span */\n context?: Context;\n /** Attributes to set on the span when it starts */\n attributes?: Attributes;\n /** Whether to end the span when the function exits (default: true) */\n endOnExit?: boolean;\n}\n\n/**\n * A dynamic tracer that allows the tracer provider to be changed at runtime.\n */\nclass DynamicTracer {\n private tracerProvider: TracerProvider;\n private tracer: Tracer;\n private readonly instrumentingModuleName: string;\n\n constructor(instrumentingModuleName: string) {\n this.instrumentingModuleName = instrumentingModuleName;\n this.tracerProvider = trace.getTracerProvider();\n this.tracer = trace.getTracer(instrumentingModuleName);\n }\n\n /**\n * Set a new tracer provider. This updates the underlying tracer instance.\n * @param provider - The new tracer provider to use\n */\n setProvider(provider: TracerProvider): void {\n this.tracerProvider = provider;\n this.tracer = this.tracerProvider.getTracer(this.instrumentingModuleName);\n }\n\n /**\n * Get the underlying OpenTelemetry tracer.\n * Use this to access the full Tracer API when needed.\n */\n getTracer(): Tracer {\n return this.tracer;\n }\n\n /**\n * Start a span manually (without making it active).\n * You must call span.end() when done.\n *\n * @param options - Span configuration including name\n * @returns The created span\n */\n startSpan(options: StartSpanOptions): Span {\n const ctx = options.context || otelContext.active();\n const span = this.tracer.startSpan(\n options.name,\n {\n attributes: options.attributes,\n },\n ctx,\n );\n\n return span;\n }\n\n /**\n * Start a new span and make it active in the current context.\n * The span will automatically be ended when the provided function completes (unless endOnExit=false).\n *\n * @param fn - The function to execute within the span context\n * @param options - Span configuration including name\n * @returns The result of the provided function\n */\n async startActiveSpan<T>(fn: (span: Span) => Promise<T>, options: StartSpanOptions): Promise<T> {\n const ctx = options.context || otelContext.active();\n const endOnExit = options.endOnExit === undefined ? true : options.endOnExit; // default true\n const opts: SpanOptions = { attributes: options.attributes };\n\n // Directly return the tracer's startActiveSpan result - it handles async correctly\n return await this.tracer.startActiveSpan(options.name, opts, ctx, async (span) => {\n try {\n return await fn(span);\n } finally {\n if (endOnExit) {\n span.end();\n }\n }\n });\n }\n\n /**\n * Synchronous version of startActiveSpan for non-async operations.\n *\n * @param fn - The function to execute within the span context\n * @param options - Span configuration including name\n * @returns The result of the provided function\n */\n startActiveSpanSync<T>(fn: (span: Span) => T, options: StartSpanOptions): T {\n const ctx = options.context || otelContext.active();\n const endOnExit = options.endOnExit === undefined ? true : options.endOnExit; // default true\n const opts: SpanOptions = { attributes: options.attributes };\n\n return this.tracer.startActiveSpan(options.name, opts, ctx, (span) => {\n try {\n return fn(span);\n } finally {\n if (endOnExit) {\n span.end();\n }\n }\n });\n }\n}\n\n/**\n * The global tracer instance used throughout the agents framework.\n * This tracer can have its provider updated at runtime via setTracerProvider().\n */\nexport const tracer = new DynamicTracer('livekit-agents');\n\nclass MetadataSpanProcessor implements SpanProcessor {\n private metadata: Attributes;\n\n constructor(metadata: Attributes) {\n this.metadata = metadata;\n }\n\n onStart(span: Span, _parentContext: Context): void {\n span.setAttributes(this.metadata);\n }\n\n onEnd(_span: ReadableSpan): void {}\n\n shutdown(): Promise<void> {\n return Promise.resolve();\n }\n\n forceFlush(): Promise<void> {\n return Promise.resolve();\n }\n}\n\n/**\n * Set the tracer provider for the livekit-agents framework.\n * This should be called before agent session start if using custom tracer providers.\n *\n * @param provider - The tracer provider to use (must be a NodeTracerProvider)\n * @param options - Optional configuration with metadata property to inject into all spans\n *\n * @example\n * ```typescript\n * import { NodeTracerProvider } from '@opentelemetry/sdk-trace-node';\n * import { setTracerProvider } from '@livekit/agents/telemetry';\n *\n * const provider = new NodeTracerProvider();\n * setTracerProvider(provider, {\n * metadata: { room_id: 'room123', job_id: 'job456' }\n * });\n * ```\n */\nexport function setTracerProvider(\n provider: NodeTracerProvider,\n options?: { metadata?: Attributes },\n): void {\n if (options?.metadata) {\n provider.addSpanProcessor(new MetadataSpanProcessor(options.metadata));\n }\n\n tracer.setProvider(provider);\n}\n\n/**\n * Setup OpenTelemetry tracer for LiveKit Cloud observability.\n * This configures OTLP exporters to send traces to LiveKit Cloud.\n *\n * @param options - Configuration for cloud tracer with roomId, jobId, and cloudHostname properties\n *\n * @internal\n */\nexport async function setupCloudTracer(options: {\n roomId: string;\n jobId: string;\n cloudHostname: string;\n}): Promise<void> {\n const { roomId, jobId, cloudHostname } = options;\n\n const apiKey = process.env.LIVEKIT_API_KEY;\n const apiSecret = process.env.LIVEKIT_API_SECRET;\n\n if (!apiKey || !apiSecret) {\n throw new Error('LIVEKIT_API_KEY and LIVEKIT_API_SECRET must be set for cloud tracing');\n }\n\n const token = new AccessToken(apiKey, apiSecret, {\n identity: 'livekit-agents-telemetry',\n ttl: '6h',\n });\n token.addObservabilityGrant({ write: true });\n\n try {\n const jwt = await token.toJwt();\n\n const headers = {\n Authorization: `Bearer ${jwt}`,\n };\n\n const metadata: Attributes = {\n room_id: roomId,\n job_id: jobId,\n };\n\n const resource = new Resource({\n [ATTR_SERVICE_NAME]: 'livekit-agents',\n room_id: roomId,\n job_id: jobId,\n });\n\n // Configure OTLP exporter to send traces to LiveKit Cloud\n const spanExporter = new OTLPTraceExporter({\n url: `https://${cloudHostname}/observability/traces/otlp/v0`,\n headers,\n compression: CompressionAlgorithm.GZIP,\n });\n\n const tracerProvider = new NodeTracerProvider({\n resource,\n spanProcessors: [new MetadataSpanProcessor(metadata), new BatchSpanProcessor(spanExporter)],\n });\n tracerProvider.register();\n\n setTracerProvider(tracerProvider);\n\n // Initialize standalone Pino cloud exporter (no OTEL SDK dependency)\n initPinoCloudExporter({\n cloudHostname,\n roomId,\n jobId,\n });\n\n enableOtelLogging();\n } catch (error) {\n console.error('Failed to setup cloud tracer:', error);\n throw error;\n }\n}\n\n/**\n * Flush all pending Pino logs to ensure they are exported.\n * Call this before session/job ends to ensure all logs are sent.\n *\n * @internal\n */\nexport async function flushOtelLogs(): Promise<void> {\n await flushPinoLogs();\n}\n\n/**\n * Convert ChatItem to proto-compatible dictionary format.\n * TODO: Use actual agent_session proto types once @livekit/protocol v1.43.1+ is published\n */\nfunction chatItemToProto(item: ChatItem): Record<string, any> {\n const itemDict: Record<string, any> = {};\n\n if (item.type === 'message') {\n const roleMap: Record<string, string> = {\n developer: 'DEVELOPER',\n system: 'SYSTEM',\n user: 'USER',\n assistant: 'ASSISTANT',\n };\n\n const msg: Record<string, any> = {\n id: item.id,\n role: roleMap[item.role] || item.role.toUpperCase(),\n content: item.content.map((c: ChatContent) => ({ text: c })),\n createdAt: toRFC3339(item.createdAt),\n };\n\n if (item.interrupted) {\n msg.interrupted = item.interrupted;\n }\n\n // TODO(brian): Add extra and transcriptConfidence to ChatMessage\n // if (item.extra && Object.keys(item.extra).length > 0) {\n // msg.extra = item.extra;\n // }\n\n // if (item.transcriptConfidence !== undefined && item.transcriptConfidence !== null) {\n // msg.transcriptConfidence = item.transcriptConfidence;\n // }\n\n // TODO(brian): Add metrics to ChatMessage\n // const metrics = item.metrics || {};\n // if (Object.keys(metrics).length > 0) {\n // msg.metrics = {};\n // if (metrics.started_speaking_at) {\n // msg.metrics.startedSpeakingAt = toRFC3339(metrics.started_speaking_at);\n // }\n // if (metrics.stopped_speaking_at) {\n // msg.metrics.stoppedSpeakingAt = toRFC3339(metrics.stopped_speaking_at);\n // }\n // if (metrics.transcription_delay !== undefined) {\n // msg.metrics.transcriptionDelay = metrics.transcription_delay;\n // }\n // if (metrics.end_of_turn_delay !== undefined) {\n // msg.metrics.endOfTurnDelay = metrics.end_of_turn_delay;\n // }\n // if (metrics.on_user_turn_completed_delay !== undefined) {\n // msg.metrics.onUserTurnCompletedDelay = metrics.on_user_turn_completed_delay;\n // }\n // if (metrics.llm_node_ttft !== undefined) {\n // msg.metrics.llmNodeTtft = metrics.llm_node_ttft;\n // }\n // if (metrics.tts_node_ttfb !== undefined) {\n // msg.metrics.ttsNodeTtfb = metrics.tts_node_ttfb;\n // }\n // if (metrics.e2e_latency !== undefined) {\n // msg.metrics.e2eLatency = metrics.e2e_latency;\n // }\n // }\n\n itemDict.message = msg;\n } else if (item.type === 'function_call') {\n itemDict.functionCall = {\n id: item.id,\n callId: item.callId,\n arguments: item.args,\n name: item.name,\n createdAt: toRFC3339(item.createdAt),\n };\n } else if (item.type === 'function_call_output') {\n itemDict.functionCallOutput = {\n id: item.id,\n name: item.name,\n callId: item.callId,\n output: item.output,\n isError: item.isError,\n createdAt: toRFC3339(item.createdAt),\n };\n } else if (item.type === 'agent_handoff') {\n const handoff: Record<string, any> = {\n id: item.id,\n newAgentId: item.newAgentId,\n createdAt: toRFC3339(item.createdAt),\n };\n if (item.oldAgentId !== undefined && item.oldAgentId !== null && item.oldAgentId !== '') {\n handoff.oldAgentId = item.oldAgentId;\n }\n itemDict.agentHandoff = handoff;\n }\n\n try {\n if (item.type === 'function_call' && typeof itemDict.functionCall?.arguments === 'string') {\n itemDict.functionCall.arguments = JSON.parse(itemDict.functionCall.arguments);\n } else if (\n item.type === 'function_call_output' &&\n typeof itemDict.functionCallOutput?.output === 'string'\n ) {\n itemDict.functionCallOutput.output = JSON.parse(itemDict.functionCallOutput.output);\n }\n } catch {\n // ignore parsing errors\n }\n\n return itemDict;\n}\n\n/**\n * Convert timestamp to RFC3339 format matching Python's _to_rfc3339.\n * Note: TypeScript createdAt is in milliseconds (Date.now()), not seconds like Python.\n * @internal\n */\nfunction toRFC3339(valueMs: number | Date): string {\n // valueMs is already in milliseconds (from Date.now())\n const dt = valueMs instanceof Date ? valueMs : new Date(valueMs);\n // Truncate sub-millisecond precision\n const truncated = new Date(Math.floor(dt.getTime()));\n return truncated.toISOString();\n}\n\n/**\n * Upload session report to LiveKit Cloud observability.\n * @param options - Configuration with agentName, cloudHostname, and report\n */\nexport async function uploadSessionReport(options: {\n agentName: string;\n cloudHostname: string;\n report: SessionReport;\n}): Promise<void> {\n const { agentName, cloudHostname, report } = options;\n\n // Create OTLP HTTP exporter for chat history logs\n // Uses raw HTTP JSON format which is required by LiveKit Cloud\n const logExporter = new SimpleOTLPHttpLogExporter({\n cloudHostname,\n resourceAttributes: {\n room_id: report.roomId,\n job_id: report.jobId,\n },\n scopeName: 'chat_history',\n scopeAttributes: {\n room_id: report.roomId,\n job_id: report.jobId,\n room: report.room,\n },\n });\n\n // Build log records for session report and chat items\n const logRecords: SimpleLogRecord[] = [];\n\n const commonAttrs = {\n room_id: report.roomId,\n job_id: report.jobId,\n 'logger.name': 'chat_history',\n };\n\n logRecords.push({\n body: 'session report',\n timestampMs: report.startedAt || report.timestamp || 0,\n attributes: {\n ...commonAttrs,\n 'session.options': report.options || {},\n 'session.report_timestamp': report.timestamp,\n agent_name: agentName,\n },\n });\n\n // Track last timestamp to ensure monotonic ordering when items have identical timestamps\n // This fixes the issue where function_call and function_call_output with same timestamp\n // get reordered by the dashboard\n let lastTimestamp = 0;\n for (const item of report.chatHistory.items) {\n // Ensure monotonically increasing timestamps for proper ordering\n // Add 0.001ms (1 microsecond) offset when timestamps collide\n let itemTimestamp = item.createdAt;\n if (itemTimestamp <= lastTimestamp) {\n itemTimestamp = lastTimestamp + 0.001; // Add 1 microsecond\n }\n lastTimestamp = itemTimestamp;\n\n const itemProto = chatItemToProto(item);\n let severityNumber = SeverityNumber.UNSPECIFIED;\n let severityText = 'unspecified';\n\n if (item.type === 'function_call_output' && item.isError) {\n severityNumber = SeverityNumber.ERROR;\n severityText = 'error';\n }\n\n logRecords.push({\n body: 'chat item',\n timestampMs: itemTimestamp, // Adjusted for monotonic ordering\n attributes: { 'chat.item': itemProto, ...commonAttrs },\n severityNumber,\n severityText,\n });\n }\n await logExporter.export(logRecords);\n\n const apiKey = process.env.LIVEKIT_API_KEY;\n const apiSecret = process.env.LIVEKIT_API_SECRET;\n\n if (!apiKey || !apiSecret) {\n throw new Error('LIVEKIT_API_KEY and LIVEKIT_API_SECRET must be set for session upload');\n }\n\n const token = new AccessToken(apiKey, apiSecret, { ttl: '6h' });\n token.addObservabilityGrant({ write: true });\n const jwt = await token.toJwt();\n\n const formData = new FormData();\n\n // Add header (protobuf MetricsRecordingHeader)\n const audioStartTime = report.audioRecordingStartedAt ?? 0;\n const headerMsg = new MetricsRecordingHeader({\n roomId: report.roomId,\n duration: BigInt(0), // TODO: Calculate actual duration from report\n startTime: {\n seconds: BigInt(Math.floor(audioStartTime / 1000)),\n nanos: Math.floor((audioStartTime % 1000) * 1e6),\n },\n });\n\n const headerBytes = Buffer.from(headerMsg.toBinary());\n formData.append('header', headerBytes, {\n filename: 'header.binpb',\n contentType: 'application/protobuf',\n knownLength: headerBytes.length,\n header: {\n 'Content-Type': 'application/protobuf',\n 'Content-Length': headerBytes.length.toString(),\n },\n });\n\n // Add chat_history JSON\n const chatHistoryJson = JSON.stringify(report.chatHistory.toJSON({ excludeTimestamp: false }));\n const chatHistoryBuffer = Buffer.from(chatHistoryJson, 'utf-8');\n formData.append('chat_history', chatHistoryBuffer, {\n filename: 'chat_history.json',\n contentType: 'application/json',\n knownLength: chatHistoryBuffer.length,\n header: {\n 'Content-Type': 'application/json',\n 'Content-Length': chatHistoryBuffer.length.toString(),\n },\n });\n\n // Add audio recording file if available\n if (report.audioRecordingPath && report.audioRecordingStartedAt) {\n let audioBytes: Buffer;\n try {\n audioBytes = await fs.readFile(report.audioRecordingPath);\n } catch {\n audioBytes = Buffer.alloc(0);\n }\n\n if (audioBytes.length > 0) {\n formData.append('audio', audioBytes, {\n filename: 'recording.ogg',\n contentType: 'audio/ogg',\n knownLength: audioBytes.length,\n header: {\n 'Content-Type': 'audio/ogg',\n 'Content-Length': audioBytes.length.toString(),\n },\n });\n }\n }\n\n // Upload to LiveKit Cloud using form-data's submit method\n // This properly streams the multipart form with all headers including Content-Length\n return new Promise<void>((resolve, reject) => {\n formData.submit(\n {\n protocol: 'https:',\n host: cloudHostname,\n path: '/observability/recordings/v0',\n method: 'POST',\n headers: {\n Authorization: `Bearer ${jwt}`,\n },\n },\n (err, res) => {\n if (err) {\n reject(new Error(`Failed to upload session report: ${err.message}`));\n return;\n }\n\n if (res.statusCode && res.statusCode >= 400) {\n reject(\n new Error(`Failed to upload session report: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.resume(); // Drain the response\n res.on('end', () => resolve());\n },\n );\n });\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAuC;AACvC,iBASO;AACP,sBAA+B;AAC/B,uCAAkC;AAClC,gCAAqC;AACrC,uBAAyB;AAEzB,4BAAuD;AACvD,kCAAkC;AAClC,uBAAqB;AACrB,gCAA4B;AAC5B,sBAAe;AAEf,iBAAkC;AAElC,gCAAgE;AAChE,iCAAqD;AAgBrD,MAAM,cAAc;AAAA,EACV;AAAA,EACA;AAAA,EACS;AAAA,EAEjB,YAAY,yBAAiC;AAC3C,SAAK,0BAA0B;AAC/B,SAAK,iBAAiB,iBAAM,kBAAkB;AAC9C,SAAK,SAAS,iBAAM,UAAU,uBAAuB;AAAA,EACvD;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,YAAY,UAAgC;AAC1C,SAAK,iBAAiB;AACtB,SAAK,SAAS,KAAK,eAAe,UAAU,KAAK,uBAAuB;AAAA,EAC1E;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,YAAoB;AAClB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,UAAU,SAAiC;AACzC,UAAM,MAAM,QAAQ,WAAW,WAAAA,QAAY,OAAO;AAClD,UAAM,OAAO,KAAK,OAAO;AAAA,MACvB,QAAQ;AAAA,MACR;AAAA,QACE,YAAY,QAAQ;AAAA,MACtB;AAAA,MACA;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,gBAAmB,IAAgC,SAAuC;AAC9F,UAAM,MAAM,QAAQ,WAAW,WAAAA,QAAY,OAAO;AAClD,UAAM,YAAY,QAAQ,cAAc,SAAY,OAAO,QAAQ;AACnE,UAAM,OAAoB,EAAE,YAAY,QAAQ,WAAW;AAG3D,WAAO,MAAM,KAAK,OAAO,gBAAgB,QAAQ,MAAM,MAAM,KAAK,OAAO,SAAS;AAChF,UAAI;AACF,eAAO,MAAM,GAAG,IAAI;AAAA,MACtB,UAAE;AACA,YAAI,WAAW;AACb,eAAK,IAAI;AAAA,QACX;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,oBAAuB,IAAuB,SAA8B;AAC1E,UAAM,MAAM,QAAQ,WAAW,WAAAA,QAAY,OAAO;AAClD,UAAM,YAAY,QAAQ,cAAc,SAAY,OAAO,QAAQ;AACnE,UAAM,OAAoB,EAAE,YAAY,QAAQ,WAAW;AAE3D,WAAO,KAAK,OAAO,gBAAgB,QAAQ,MAAM,MAAM,KAAK,CAAC,SAAS;AACpE,UAAI;AACF,eAAO,GAAG,IAAI;AAAA,MAChB,UAAE;AACA,YAAI,WAAW;AACb,eAAK,IAAI;AAAA,QACX;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AACF;AAMO,MAAM,SAAS,IAAI,cAAc,gBAAgB;AAExD,MAAM,sBAA+C;AAAA,EAC3C;AAAA,EAER,YAAY,UAAsB;AAChC,SAAK,WAAW;AAAA,EAClB;AAAA,EAEA,QAAQ,MAAY,gBAA+B;AACjD,SAAK,cAAc,KAAK,QAAQ;AAAA,EAClC;AAAA,EAEA,MAAM,OAA2B;AAAA,EAAC;AAAA,EAElC,WAA0B;AACxB,WAAO,QAAQ,QAAQ;AAAA,EACzB;AAAA,EAEA,aAA4B;AAC1B,WAAO,QAAQ,QAAQ;AAAA,EACzB;AACF;AAoBO,SAAS,kBACd,UACA,SACM;AACN,MAAI,mCAAS,UAAU;AACrB,aAAS,iBAAiB,IAAI,sBAAsB,QAAQ,QAAQ,CAAC;AAAA,EACvE;AAEA,SAAO,YAAY,QAAQ;AAC7B;AAUA,eAAsB,iBAAiB,SAIrB;AAChB,QAAM,EAAE,QAAQ,OAAO,cAAc,IAAI;AAEzC,QAAM,SAAS,QAAQ,IAAI;AAC3B,QAAM,YAAY,QAAQ,IAAI;AAE9B,MAAI,CAAC,UAAU,CAAC,WAAW;AACzB,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAEA,QAAM,QAAQ,IAAI,sCAAY,QAAQ,WAAW;AAAA,IAC/C,UAAU;AAAA,IACV,KAAK;AAAA,EACP,CAAC;AACD,QAAM,sBAAsB,EAAE,OAAO,KAAK,CAAC;AAE3C,MAAI;AACF,UAAM,MAAM,MAAM,MAAM,MAAM;AAE9B,UAAM,UAAU;AAAA,MACd,eAAe,UAAU,GAAG;AAAA,IAC9B;AAEA,UAAM,WAAuB;AAAA,MAC3B,SAAS;AAAA,MACT,QAAQ;AAAA,IACV;AAEA,UAAM,WAAW,IAAI,0BAAS;AAAA,MAC5B,CAAC,6CAAiB,GAAG;AAAA,MACrB,SAAS;AAAA,MACT,QAAQ;AAAA,IACV,CAAC;AAGD,UAAM,eAAe,IAAI,mDAAkB;AAAA,MACzC,KAAK,WAAW,aAAa;AAAA,MAC7B;AAAA,MACA,aAAa,+CAAqB;AAAA,IACpC,CAAC;AAED,UAAM,iBAAiB,IAAI,yCAAmB;AAAA,MAC5C;AAAA,MACA,gBAAgB,CAAC,IAAI,sBAAsB,QAAQ,GAAG,IAAI,yCAAmB,YAAY,CAAC;AAAA,IAC5F,CAAC;AACD,mBAAe,SAAS;AAExB,sBAAkB,cAAc;AAGhC,0DAAsB;AAAA,MACpB;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAED,sCAAkB;AAAA,EACpB,SAAS,OAAO;AACd,YAAQ,MAAM,iCAAiC,KAAK;AACpD,UAAM;AAAA,EACR;AACF;AAQA,eAAsB,gBAA+B;AACnD,YAAM,0CAAc;AACtB;AAMA,SAAS,gBAAgB,MAAqC;AA/R9D;AAgSE,QAAM,WAAgC,CAAC;AAEvC,MAAI,KAAK,SAAS,WAAW;AAC3B,UAAM,UAAkC;AAAA,MACtC,WAAW;AAAA,MACX,QAAQ;AAAA,MACR,MAAM;AAAA,MACN,WAAW;AAAA,IACb;AAEA,UAAM,MAA2B;AAAA,MAC/B,IAAI,KAAK;AAAA,MACT,MAAM,QAAQ,KAAK,IAAI,KAAK,KAAK,KAAK,YAAY;AAAA,MAClD,SAAS,KAAK,QAAQ,IAAI,CAAC,OAAoB,EAAE,MAAM,EAAE,EAAE;AAAA,MAC3D,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AAEA,QAAI,KAAK,aAAa;AACpB,UAAI,cAAc,KAAK;AAAA,IACzB;AAyCA,aAAS,UAAU;AAAA,EACrB,WAAW,KAAK,SAAS,iBAAiB;AACxC,aAAS,eAAe;AAAA,MACtB,IAAI,KAAK;AAAA,MACT,QAAQ,KAAK;AAAA,MACb,WAAW,KAAK;AAAA,MAChB,MAAM,KAAK;AAAA,MACX,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AAAA,EACF,WAAW,KAAK,SAAS,wBAAwB;AAC/C,aAAS,qBAAqB;AAAA,MAC5B,IAAI,KAAK;AAAA,MACT,MAAM,KAAK;AAAA,MACX,QAAQ,KAAK;AAAA,MACb,QAAQ,KAAK;AAAA,MACb,SAAS,KAAK;AAAA,MACd,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AAAA,EACF,WAAW,KAAK,SAAS,iBAAiB;AACxC,UAAM,UAA+B;AAAA,MACnC,IAAI,KAAK;AAAA,MACT,YAAY,KAAK;AAAA,MACjB,WAAW,UAAU,KAAK,SAAS;AAAA,IACrC;AACA,QAAI,KAAK,eAAe,UAAa,KAAK,eAAe,QAAQ,KAAK,eAAe,IAAI;AACvF,cAAQ,aAAa,KAAK;AAAA,IAC5B;AACA,aAAS,eAAe;AAAA,EAC1B;AAEA,MAAI;AACF,QAAI,KAAK,SAAS,mBAAmB,SAAO,cAAS,iBAAT,mBAAuB,eAAc,UAAU;AACzF,eAAS,aAAa,YAAY,KAAK,MAAM,SAAS,aAAa,SAAS;AAAA,IAC9E,WACE,KAAK,SAAS,0BACd,SAAO,cAAS,uBAAT,mBAA6B,YAAW,UAC/C;AACA,eAAS,mBAAmB,SAAS,KAAK,MAAM,SAAS,mBAAmB,MAAM;AAAA,IACpF;AAAA,EACF,QAAQ;AAAA,EAER;AAEA,SAAO;AACT;AAOA,SAAS,UAAU,SAAgC;AAEjD,QAAM,KAAK,mBAAmB,OAAO,UAAU,IAAI,KAAK,OAAO;AAE/D,QAAM,YAAY,IAAI,KAAK,KAAK,MAAM,GAAG,QAAQ,CAAC,CAAC;AACnD,SAAO,UAAU,YAAY;AAC/B;AAMA,eAAsB,oBAAoB,SAIxB;AAChB,QAAM,EAAE,WAAW,eAAe,OAAO,IAAI;AAI7C,QAAM,cAAc,IAAI,oDAA0B;AAAA,IAChD;AAAA,IACA,oBAAoB;AAAA,MAClB,SAAS,OAAO;AAAA,MAChB,QAAQ,OAAO;AAAA,IACjB;AAAA,IACA,WAAW;AAAA,IACX,iBAAiB;AAAA,MACf,SAAS,OAAO;AAAA,MAChB,QAAQ,OAAO;AAAA,MACf,MAAM,OAAO;AAAA,IACf;AAAA,EACF,CAAC;AAGD,QAAM,aAAgC,CAAC;AAEvC,QAAM,cAAc;AAAA,IAClB,SAAS,OAAO;AAAA,IAChB,QAAQ,OAAO;AAAA,IACf,eAAe;AAAA,EACjB;AAEA,aAAW,KAAK;AAAA,IACd,MAAM;AAAA,IACN,aAAa,OAAO,aAAa,OAAO,aAAa;AAAA,IACrD,YAAY;AAAA,MACV,GAAG;AAAA,MACH,mBAAmB,OAAO,WAAW,CAAC;AAAA,MACtC,4BAA4B,OAAO;AAAA,MACnC,YAAY;AAAA,IACd;AAAA,EACF,CAAC;AAKD,MAAI,gBAAgB;AACpB,aAAW,QAAQ,OAAO,YAAY,OAAO;AAG3C,QAAI,gBAAgB,KAAK;AACzB,QAAI,iBAAiB,eAAe;AAClC,sBAAgB,gBAAgB;AAAA,IAClC;AACA,oBAAgB;AAEhB,UAAM,YAAY,gBAAgB,IAAI;AACtC,QAAI,iBAAiB,+BAAe;AACpC,QAAI,eAAe;AAEnB,QAAI,KAAK,SAAS,0BAA0B,KAAK,SAAS;AACxD,uBAAiB,+BAAe;AAChC,qBAAe;AAAA,IACjB;AAEA,eAAW,KAAK;AAAA,MACd,MAAM;AAAA,MACN,aAAa;AAAA;AAAA,MACb,YAAY,EAAE,aAAa,WAAW,GAAG,YAAY;AAAA,MACrD;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AACA,QAAM,YAAY,OAAO,UAAU;AAEnC,QAAM,SAAS,QAAQ,IAAI;AAC3B,QAAM,YAAY,QAAQ,IAAI;AAE9B,MAAI,CAAC,UAAU,CAAC,WAAW;AACzB,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AAEA,QAAM,QAAQ,IAAI,sCAAY,QAAQ,WAAW,EAAE,KAAK,KAAK,CAAC;AAC9D,QAAM,sBAAsB,EAAE,OAAO,KAAK,CAAC;AAC3C,QAAM,MAAM,MAAM,MAAM,MAAM;AAE9B,QAAM,WAAW,IAAI,iBAAAC,QAAS;AAG9B,QAAM,iBAAiB,OAAO,2BAA2B;AACzD,QAAM,YAAY,IAAI,uCAAuB;AAAA,IAC3C,QAAQ,OAAO;AAAA,IACf,UAAU,OAAO,CAAC;AAAA;AAAA,IAClB,WAAW;AAAA,MACT,SAAS,OAAO,KAAK,MAAM,iBAAiB,GAAI,CAAC;AAAA,MACjD,OAAO,KAAK,MAAO,iBAAiB,MAAQ,GAAG;AAAA,IACjD;AAAA,EACF,CAAC;AAED,QAAM,cAAc,OAAO,KAAK,UAAU,SAAS,CAAC;AACpD,WAAS,OAAO,UAAU,aAAa;AAAA,IACrC,UAAU;AAAA,IACV,aAAa;AAAA,IACb,aAAa,YAAY;AAAA,IACzB,QAAQ;AAAA,MACN,gBAAgB;AAAA,MAChB,kBAAkB,YAAY,OAAO,SAAS;AAAA,IAChD;AAAA,EACF,CAAC;AAGD,QAAM,kBAAkB,KAAK,UAAU,OAAO,YAAY,OAAO,EAAE,kBAAkB,MAAM,CAAC,CAAC;AAC7F,QAAM,oBAAoB,OAAO,KAAK,iBAAiB,OAAO;AAC9D,WAAS,OAAO,gBAAgB,mBAAmB;AAAA,IACjD,UAAU;AAAA,IACV,aAAa;AAAA,IACb,aAAa,kBAAkB;AAAA,IAC/B,QAAQ;AAAA,MACN,gBAAgB;AAAA,MAChB,kBAAkB,kBAAkB,OAAO,SAAS;AAAA,IACtD;AAAA,EACF,CAAC;AAGD,MAAI,OAAO,sBAAsB,OAAO,yBAAyB;AAC/D,QAAI;AACJ,QAAI;AACF,mBAAa,MAAM,gBAAAC,QAAG,SAAS,OAAO,kBAAkB;AAAA,IAC1D,QAAQ;AACN,mBAAa,OAAO,MAAM,CAAC;AAAA,IAC7B;AAEA,QAAI,WAAW,SAAS,GAAG;AACzB,eAAS,OAAO,SAAS,YAAY;AAAA,QACnC,UAAU;AAAA,QACV,aAAa;AAAA,QACb,aAAa,WAAW;AAAA,QACxB,QAAQ;AAAA,UACN,gBAAgB;AAAA,UAChB,kBAAkB,WAAW,OAAO,SAAS;AAAA,QAC/C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAIA,SAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,MAAM;AAAA,QACN,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,eAAe,UAAU,GAAG;AAAA,QAC9B;AAAA,MACF;AAAA,MACA,CAAC,KAAK,QAAQ;AACZ,YAAI,KAAK;AACP,iBAAO,IAAI,MAAM,oCAAoC,IAAI,OAAO,EAAE,CAAC;AACnE;AAAA,QACF;AAEA,YAAI,IAAI,cAAc,IAAI,cAAc,KAAK;AAC3C;AAAA,YACE,IAAI,MAAM,oCAAoC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,UACrF;AACA;AAAA,QACF;AAEA,YAAI,OAAO;AACX,YAAI,GAAG,OAAO,MAAM,QAAQ,CAAC;AAAA,MAC/B;AAAA,IACF;AAAA,EACF,CAAC;AACH;","names":["otelContext","FormData","fs"]}
@@ -1 +1 @@
1
- {"version":3,"file":"traces.d.ts","sourceRoot":"","sources":["../../src/telemetry/traces.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,UAAU,EACf,KAAK,OAAO,EACZ,KAAK,IAAI,EAET,KAAK,MAAM,EACX,KAAK,cAAc,EAGpB,MAAM,oBAAoB,CAAC;AAM5B,OAAO,EAAsB,kBAAkB,EAAE,MAAM,+BAA+B,CAAC;AAMvF,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAIxD,MAAM,WAAW,gBAAgB;IAC/B,uBAAuB;IACvB,IAAI,EAAE,MAAM,CAAC;IACb,mDAAmD;IACnD,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB,mDAAmD;IACnD,UAAU,CAAC,EAAE,UAAU,CAAC;IACxB,sEAAsE;IACtE,SAAS,CAAC,EAAE,OAAO,CAAC;CACrB;AAED;;GAEG;AACH,cAAM,aAAa;IACjB,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,QAAQ,CAAC,uBAAuB,CAAS;gBAErC,uBAAuB,EAAE,MAAM;IAM3C;;;OAGG;IACH,WAAW,CAAC,QAAQ,EAAE,cAAc,GAAG,IAAI;IAK3C;;;OAGG;IACH,SAAS,IAAI,MAAM;IAInB;;;;;;OAMG;IACH,SAAS,CAAC,OAAO,EAAE,gBAAgB,GAAG,IAAI;IAa1C;;;;;;;OAOG;IACG,eAAe,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,gBAAgB,GAAG,OAAO,CAAC,CAAC,CAAC;IAiB/F;;;;;;OAMG;IACH,mBAAmB,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,IAAI,KAAK,CAAC,EAAE,OAAO,EAAE,gBAAgB,GAAG,CAAC;CAe5E;AAED;;;GAGG;AACH,eAAO,MAAM,MAAM,eAAsC,CAAC;AAwB1D;;;;;;;;;;;;;;;;;GAiBG;AACH,wBAAgB,iBAAiB,CAC/B,QAAQ,EAAE,kBAAkB,EAC5B,OAAO,CAAC,EAAE;IAAE,QAAQ,CAAC,EAAE,UAAU,CAAA;CAAE,GAClC,IAAI,CAMN;AAED;;;;;;;GAOG;AACH,wBAAsB,gBAAgB,CAAC,OAAO,EAAE;IAC9C,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,EAAE,MAAM,CAAC;IACd,aAAa,EAAE,MAAM,CAAC;CACvB,GAAG,OAAO,CAAC,IAAI,CAAC,CA6DhB;AAED;;;;;GAKG;AACH,wBAAsB,aAAa,IAAI,OAAO,CAAC,IAAI,CAAC,CAEnD;AA8HD;;;GAGG;AACH,wBAAsB,mBAAmB,CAAC,OAAO,EAAE;IACjD,SAAS,EAAE,MAAM,CAAC;IAClB,aAAa,EAAE,MAAM,CAAC;IACtB,MAAM,EAAE,aAAa,CAAC;CACvB,GAAG,OAAO,CAAC,IAAI,CAAC,CAuJhB"}
1
+ {"version":3,"file":"traces.d.ts","sourceRoot":"","sources":["../../src/telemetry/traces.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,UAAU,EACf,KAAK,OAAO,EACZ,KAAK,IAAI,EAET,KAAK,MAAM,EACX,KAAK,cAAc,EAGpB,MAAM,oBAAoB,CAAC;AAM5B,OAAO,EAAsB,kBAAkB,EAAE,MAAM,+BAA+B,CAAC;AAOvF,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAIxD,MAAM,WAAW,gBAAgB;IAC/B,uBAAuB;IACvB,IAAI,EAAE,MAAM,CAAC;IACb,mDAAmD;IACnD,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB,mDAAmD;IACnD,UAAU,CAAC,EAAE,UAAU,CAAC;IACxB,sEAAsE;IACtE,SAAS,CAAC,EAAE,OAAO,CAAC;CACrB;AAED;;GAEG;AACH,cAAM,aAAa;IACjB,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,QAAQ,CAAC,uBAAuB,CAAS;gBAErC,uBAAuB,EAAE,MAAM;IAM3C;;;OAGG;IACH,WAAW,CAAC,QAAQ,EAAE,cAAc,GAAG,IAAI;IAK3C;;;OAGG;IACH,SAAS,IAAI,MAAM;IAInB;;;;;;OAMG;IACH,SAAS,CAAC,OAAO,EAAE,gBAAgB,GAAG,IAAI;IAa1C;;;;;;;OAOG;IACG,eAAe,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,gBAAgB,GAAG,OAAO,CAAC,CAAC,CAAC;IAiB/F;;;;;;OAMG;IACH,mBAAmB,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,IAAI,KAAK,CAAC,EAAE,OAAO,EAAE,gBAAgB,GAAG,CAAC;CAe5E;AAED;;;GAGG;AACH,eAAO,MAAM,MAAM,eAAsC,CAAC;AAwB1D;;;;;;;;;;;;;;;;;GAiBG;AACH,wBAAgB,iBAAiB,CAC/B,QAAQ,EAAE,kBAAkB,EAC5B,OAAO,CAAC,EAAE;IAAE,QAAQ,CAAC,EAAE,UAAU,CAAA;CAAE,GAClC,IAAI,CAMN;AAED;;;;;;;GAOG;AACH,wBAAsB,gBAAgB,CAAC,OAAO,EAAE;IAC9C,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,EAAE,MAAM,CAAC;IACd,aAAa,EAAE,MAAM,CAAC;CACvB,GAAG,OAAO,CAAC,IAAI,CAAC,CA6DhB;AAED;;;;;GAKG;AACH,wBAAsB,aAAa,IAAI,OAAO,CAAC,IAAI,CAAC,CAEnD;AA8HD;;;GAGG;AACH,wBAAsB,mBAAmB,CAAC,OAAO,EAAE;IACjD,SAAS,EAAE,MAAM,CAAC;IAClB,aAAa,EAAE,MAAM,CAAC;IACtB,MAAM,EAAE,aAAa,CAAC;CACvB,GAAG,OAAO,CAAC,IAAI,CAAC,CA4KhB"}
@@ -11,6 +11,7 @@ import { BatchSpanProcessor, NodeTracerProvider } from "@opentelemetry/sdk-trace
11
11
  import { ATTR_SERVICE_NAME } from "@opentelemetry/semantic-conventions";
12
12
  import FormData from "form-data";
13
13
  import { AccessToken } from "livekit-server-sdk";
14
+ import fs from "node:fs/promises";
14
15
  import { enableOtelLogging } from "../log.js";
15
16
  import { SimpleOTLPHttpLogExporter } from "./otel_http_exporter.js";
16
17
  import { flushPinoLogs, initPinoCloudExporter } from "./pino_otel_transport.js";
@@ -302,13 +303,14 @@ async function uploadSessionReport(options) {
302
303
  token.addObservabilityGrant({ write: true });
303
304
  const jwt = await token.toJwt();
304
305
  const formData = new FormData();
306
+ const audioStartTime = report.audioRecordingStartedAt ?? 0;
305
307
  const headerMsg = new MetricsRecordingHeader({
306
308
  roomId: report.roomId,
307
309
  duration: BigInt(0),
308
310
  // TODO: Calculate actual duration from report
309
311
  startTime: {
310
- seconds: BigInt(Math.floor(report.timestamp / 1e3)),
311
- nanos: Math.floor(report.timestamp % 1e3 * 1e6)
312
+ seconds: BigInt(Math.floor(audioStartTime / 1e3)),
313
+ nanos: Math.floor(audioStartTime % 1e3 * 1e6)
312
314
  }
313
315
  });
314
316
  const headerBytes = Buffer.from(headerMsg.toBinary());
@@ -332,6 +334,25 @@ async function uploadSessionReport(options) {
332
334
  "Content-Length": chatHistoryBuffer.length.toString()
333
335
  }
334
336
  });
337
+ if (report.audioRecordingPath && report.audioRecordingStartedAt) {
338
+ let audioBytes;
339
+ try {
340
+ audioBytes = await fs.readFile(report.audioRecordingPath);
341
+ } catch {
342
+ audioBytes = Buffer.alloc(0);
343
+ }
344
+ if (audioBytes.length > 0) {
345
+ formData.append("audio", audioBytes, {
346
+ filename: "recording.ogg",
347
+ contentType: "audio/ogg",
348
+ knownLength: audioBytes.length,
349
+ header: {
350
+ "Content-Type": "audio/ogg",
351
+ "Content-Length": audioBytes.length.toString()
352
+ }
353
+ });
354
+ }
355
+ }
335
356
  return new Promise((resolve, reject) => {
336
357
  formData.submit(
337
358
  {