@livekit/agents 1.0.21 → 1.0.23

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 (149) hide show
  1. package/dist/inference/api_protos.cjs +2 -2
  2. package/dist/inference/api_protos.cjs.map +1 -1
  3. package/dist/inference/api_protos.d.cts +16 -16
  4. package/dist/inference/api_protos.d.ts +16 -16
  5. package/dist/inference/api_protos.js +2 -2
  6. package/dist/inference/api_protos.js.map +1 -1
  7. package/dist/inference/stt.cjs +42 -30
  8. package/dist/inference/stt.cjs.map +1 -1
  9. package/dist/inference/stt.d.ts.map +1 -1
  10. package/dist/inference/stt.js +42 -30
  11. package/dist/inference/stt.js.map +1 -1
  12. package/dist/inference/tts.cjs +2 -3
  13. package/dist/inference/tts.cjs.map +1 -1
  14. package/dist/inference/tts.d.ts.map +1 -1
  15. package/dist/inference/tts.js +2 -3
  16. package/dist/inference/tts.js.map +1 -1
  17. package/dist/ipc/job_proc_lazy_main.cjs +35 -1
  18. package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
  19. package/dist/ipc/job_proc_lazy_main.js +13 -1
  20. package/dist/ipc/job_proc_lazy_main.js.map +1 -1
  21. package/dist/job.cjs +52 -6
  22. package/dist/job.cjs.map +1 -1
  23. package/dist/job.d.cts +2 -0
  24. package/dist/job.d.ts +2 -0
  25. package/dist/job.d.ts.map +1 -1
  26. package/dist/job.js +52 -6
  27. package/dist/job.js.map +1 -1
  28. package/dist/llm/llm.cjs +38 -3
  29. package/dist/llm/llm.cjs.map +1 -1
  30. package/dist/llm/llm.d.cts +1 -0
  31. package/dist/llm/llm.d.ts +1 -0
  32. package/dist/llm/llm.d.ts.map +1 -1
  33. package/dist/llm/llm.js +38 -3
  34. package/dist/llm/llm.js.map +1 -1
  35. package/dist/log.cjs +34 -10
  36. package/dist/log.cjs.map +1 -1
  37. package/dist/log.d.cts +7 -0
  38. package/dist/log.d.ts +7 -0
  39. package/dist/log.d.ts.map +1 -1
  40. package/dist/log.js +34 -11
  41. package/dist/log.js.map +1 -1
  42. package/dist/stt/stt.cjs +18 -5
  43. package/dist/stt/stt.cjs.map +1 -1
  44. package/dist/stt/stt.d.ts.map +1 -1
  45. package/dist/stt/stt.js +18 -5
  46. package/dist/stt/stt.js.map +1 -1
  47. package/dist/telemetry/index.cjs +23 -2
  48. package/dist/telemetry/index.cjs.map +1 -1
  49. package/dist/telemetry/index.d.cts +4 -1
  50. package/dist/telemetry/index.d.ts +4 -1
  51. package/dist/telemetry/index.d.ts.map +1 -1
  52. package/dist/telemetry/index.js +27 -2
  53. package/dist/telemetry/index.js.map +1 -1
  54. package/dist/telemetry/logging.cjs +65 -0
  55. package/dist/telemetry/logging.cjs.map +1 -0
  56. package/dist/telemetry/logging.d.cts +21 -0
  57. package/dist/telemetry/logging.d.ts +21 -0
  58. package/dist/telemetry/logging.d.ts.map +1 -0
  59. package/dist/telemetry/logging.js +40 -0
  60. package/dist/telemetry/logging.js.map +1 -0
  61. package/dist/telemetry/otel_http_exporter.cjs +144 -0
  62. package/dist/telemetry/otel_http_exporter.cjs.map +1 -0
  63. package/dist/telemetry/otel_http_exporter.d.cts +62 -0
  64. package/dist/telemetry/otel_http_exporter.d.ts +62 -0
  65. package/dist/telemetry/otel_http_exporter.d.ts.map +1 -0
  66. package/dist/telemetry/otel_http_exporter.js +120 -0
  67. package/dist/telemetry/otel_http_exporter.js.map +1 -0
  68. package/dist/telemetry/pino_otel_transport.cjs +217 -0
  69. package/dist/telemetry/pino_otel_transport.cjs.map +1 -0
  70. package/dist/telemetry/pino_otel_transport.d.cts +58 -0
  71. package/dist/telemetry/pino_otel_transport.d.ts +58 -0
  72. package/dist/telemetry/pino_otel_transport.d.ts.map +1 -0
  73. package/dist/telemetry/pino_otel_transport.js +189 -0
  74. package/dist/telemetry/pino_otel_transport.js.map +1 -0
  75. package/dist/telemetry/traces.cjs +225 -16
  76. package/dist/telemetry/traces.cjs.map +1 -1
  77. package/dist/telemetry/traces.d.cts +17 -0
  78. package/dist/telemetry/traces.d.ts +17 -0
  79. package/dist/telemetry/traces.d.ts.map +1 -1
  80. package/dist/telemetry/traces.js +211 -14
  81. package/dist/telemetry/traces.js.map +1 -1
  82. package/dist/tts/tts.cjs +68 -20
  83. package/dist/tts/tts.cjs.map +1 -1
  84. package/dist/tts/tts.d.cts +2 -0
  85. package/dist/tts/tts.d.ts +2 -0
  86. package/dist/tts/tts.d.ts.map +1 -1
  87. package/dist/tts/tts.js +68 -20
  88. package/dist/tts/tts.js.map +1 -1
  89. package/dist/utils.cjs +6 -0
  90. package/dist/utils.cjs.map +1 -1
  91. package/dist/utils.d.cts +1 -0
  92. package/dist/utils.d.ts +1 -0
  93. package/dist/utils.d.ts.map +1 -1
  94. package/dist/utils.js +5 -0
  95. package/dist/utils.js.map +1 -1
  96. package/dist/voice/agent_activity.cjs +93 -7
  97. package/dist/voice/agent_activity.cjs.map +1 -1
  98. package/dist/voice/agent_activity.d.cts +3 -0
  99. package/dist/voice/agent_activity.d.ts +3 -0
  100. package/dist/voice/agent_activity.d.ts.map +1 -1
  101. package/dist/voice/agent_activity.js +93 -7
  102. package/dist/voice/agent_activity.js.map +1 -1
  103. package/dist/voice/agent_session.cjs +122 -27
  104. package/dist/voice/agent_session.cjs.map +1 -1
  105. package/dist/voice/agent_session.d.cts +15 -0
  106. package/dist/voice/agent_session.d.ts +15 -0
  107. package/dist/voice/agent_session.d.ts.map +1 -1
  108. package/dist/voice/agent_session.js +122 -27
  109. package/dist/voice/agent_session.js.map +1 -1
  110. package/dist/voice/audio_recognition.cjs +69 -22
  111. package/dist/voice/audio_recognition.cjs.map +1 -1
  112. package/dist/voice/audio_recognition.d.cts +5 -0
  113. package/dist/voice/audio_recognition.d.ts +5 -0
  114. package/dist/voice/audio_recognition.d.ts.map +1 -1
  115. package/dist/voice/audio_recognition.js +69 -22
  116. package/dist/voice/audio_recognition.js.map +1 -1
  117. package/dist/voice/generation.cjs +43 -3
  118. package/dist/voice/generation.cjs.map +1 -1
  119. package/dist/voice/generation.d.ts.map +1 -1
  120. package/dist/voice/generation.js +43 -3
  121. package/dist/voice/generation.js.map +1 -1
  122. package/dist/voice/report.cjs +3 -2
  123. package/dist/voice/report.cjs.map +1 -1
  124. package/dist/voice/report.d.cts +7 -1
  125. package/dist/voice/report.d.ts +7 -1
  126. package/dist/voice/report.d.ts.map +1 -1
  127. package/dist/voice/report.js +3 -2
  128. package/dist/voice/report.js.map +1 -1
  129. package/package.json +8 -2
  130. package/src/inference/api_protos.ts +2 -2
  131. package/src/inference/stt.ts +48 -33
  132. package/src/inference/tts.ts +4 -3
  133. package/src/ipc/job_proc_lazy_main.ts +12 -1
  134. package/src/job.ts +59 -10
  135. package/src/llm/llm.ts +48 -5
  136. package/src/log.ts +52 -15
  137. package/src/stt/stt.ts +18 -5
  138. package/src/telemetry/index.ts +22 -4
  139. package/src/telemetry/logging.ts +55 -0
  140. package/src/telemetry/otel_http_exporter.ts +191 -0
  141. package/src/telemetry/pino_otel_transport.ts +265 -0
  142. package/src/telemetry/traces.ts +320 -20
  143. package/src/tts/tts.ts +85 -24
  144. package/src/utils.ts +5 -0
  145. package/src/voice/agent_activity.ts +140 -22
  146. package/src/voice/agent_session.ts +174 -34
  147. package/src/voice/audio_recognition.ts +85 -26
  148. package/src/voice/generation.ts +59 -7
  149. package/src/voice/report.ts +10 -4
package/dist/llm/llm.js CHANGED
@@ -1,6 +1,7 @@
1
1
  import { EventEmitter } from "node:events";
2
2
  import { APIConnectionError, APIError } from "../_exceptions.js";
3
3
  import { log } from "../log.js";
4
+ import { recordException, traceTypes, tracer } from "../telemetry/index.js";
4
5
  import { AsyncIterableQueue, delay, startSoon, toError } from "../utils.js";
5
6
  import {} from "./chat_context.js";
6
7
  class LLM extends EventEmitter {
@@ -36,6 +37,7 @@ class LLMStream {
36
37
  #llm;
37
38
  #chatCtx;
38
39
  #toolCtx;
40
+ #llmRequestSpan;
39
41
  constructor(llm, {
40
42
  chatCtx,
41
43
  toolCtx,
@@ -52,10 +54,23 @@ class LLMStream {
52
54
  });
53
55
  startSoon(() => this.mainTask().then(() => this.queue.close()));
54
56
  }
55
- async mainTask() {
57
+ _mainTaskImpl = async (span) => {
58
+ this.#llmRequestSpan = span;
59
+ span.setAttribute(traceTypes.ATTR_GEN_AI_REQUEST_MODEL, this.#llm.model);
56
60
  for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {
57
61
  try {
58
- return await this.run();
62
+ return await tracer.startActiveSpan(
63
+ async (attemptSpan) => {
64
+ attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);
65
+ try {
66
+ return await this.run();
67
+ } catch (error) {
68
+ recordException(attemptSpan, toError(error));
69
+ throw error;
70
+ }
71
+ },
72
+ { name: "llm_request_run" }
73
+ );
59
74
  } catch (error) {
60
75
  if (error instanceof APIError) {
61
76
  const retryInterval = this._connOptions._intervalForRetry(i);
@@ -84,7 +99,11 @@ class LLMStream {
84
99
  }
85
100
  }
86
101
  }
87
- }
102
+ };
103
+ mainTask = async () => tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {
104
+ name: "llm_request",
105
+ endOnExit: false
106
+ });
88
107
  emitError({ error, recoverable }) {
89
108
  this.#llm.emit("error", {
90
109
  type: "llm_error",
@@ -99,6 +118,7 @@ class LLMStream {
99
118
  let ttft = BigInt(-1);
100
119
  let requestId = "";
101
120
  let usage;
121
+ let completionStartTime;
102
122
  for await (const ev of this.queue) {
103
123
  if (this.abortController.signal.aborted) {
104
124
  break;
@@ -107,6 +127,7 @@ class LLMStream {
107
127
  requestId = ev.id;
108
128
  if (ttft === BigInt(-1)) {
109
129
  ttft = process.hrtime.bigint() - startTime;
130
+ completionStartTime = (/* @__PURE__ */ new Date()).toISOString();
110
131
  }
111
132
  if (ev.usage) {
112
133
  usage = ev.usage;
@@ -134,6 +155,20 @@ class LLMStream {
134
155
  return ((usage == null ? void 0 : usage.completionTokens) || 0) / (durationMs / 1e3);
135
156
  })()
136
157
  };
158
+ if (this.#llmRequestSpan) {
159
+ this.#llmRequestSpan.setAttribute(traceTypes.ATTR_LLM_METRICS, JSON.stringify(metrics));
160
+ this.#llmRequestSpan.setAttributes({
161
+ [traceTypes.ATTR_GEN_AI_USAGE_INPUT_TOKENS]: metrics.promptTokens,
162
+ [traceTypes.ATTR_GEN_AI_USAGE_OUTPUT_TOKENS]: metrics.completionTokens
163
+ });
164
+ if (completionStartTime) {
165
+ this.#llmRequestSpan.setAttribute(
166
+ traceTypes.ATTR_LANGFUSE_COMPLETION_START_TIME,
167
+ completionStartTime
168
+ );
169
+ }
170
+ this.#llmRequestSpan.end();
171
+ }
137
172
  this.#llm.emit("metrics_collected", metrics);
138
173
  }
139
174
  /** The function context of this stream. */
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/llm/llm.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { LLMMetrics } from '../metrics/base.js';\nimport type { APIConnectOptions } from '../types.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\nimport { type ChatContext, type ChatRole, type FunctionCall } from './chat_context.js';\nimport type { ToolChoice, ToolContext } from './tool_context.js';\n\nexport interface ChoiceDelta {\n role: ChatRole;\n content?: string;\n toolCalls?: FunctionCall[];\n}\n\nexport interface CompletionUsage {\n completionTokens: number;\n promptTokens: number;\n promptCachedTokens: number;\n totalTokens: number;\n}\n\nexport interface ChatChunk {\n id: string;\n delta?: ChoiceDelta;\n usage?: CompletionUsage;\n}\n\nexport interface LLMError {\n type: 'llm_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type LLMCallbacks = {\n ['metrics_collected']: (metrics: LLMMetrics) => void;\n ['error']: (error: LLMError) => void;\n};\n\nexport abstract class LLM extends (EventEmitter as new () => TypedEmitter<LLMCallbacks>) {\n constructor() {\n super();\n }\n\n abstract label(): string;\n\n /**\n * Get the model name/identifier for this LLM instance.\n *\n * @returns The model name if available, \"unknown\" otherwise.\n *\n * @remarks\n * Plugins should override this property to provide their model information.\n */\n get model(): string {\n return 'unknown';\n }\n\n /**\n * Returns a {@link LLMStream} that can be used to push text and receive LLM responses.\n */\n abstract chat({\n chatCtx,\n toolCtx,\n connOptions,\n parallelToolCalls,\n toolChoice,\n extraKwargs,\n }: {\n chatCtx: ChatContext;\n toolCtx?: ToolContext;\n connOptions?: APIConnectOptions;\n parallelToolCalls?: boolean;\n toolChoice?: ToolChoice;\n extraKwargs?: Record<string, unknown>;\n }): LLMStream;\n\n /**\n * Pre-warm connection to the LLM service\n */\n prewarm(): void {\n // Default implementation - subclasses can override\n }\n\n async aclose(): Promise<void> {\n // Default implementation - subclasses can override\n }\n}\n\nexport abstract class LLMStream implements AsyncIterableIterator<ChatChunk> {\n protected output = new AsyncIterableQueue<ChatChunk>();\n protected queue = new AsyncIterableQueue<ChatChunk>();\n protected closed = false;\n protected abortController = new AbortController();\n protected _connOptions: APIConnectOptions;\n protected logger = log();\n\n #llm: LLM;\n #chatCtx: ChatContext;\n #toolCtx?: ToolContext;\n\n constructor(\n llm: LLM,\n {\n chatCtx,\n toolCtx,\n connOptions,\n }: {\n chatCtx: ChatContext;\n toolCtx?: ToolContext;\n connOptions: APIConnectOptions;\n },\n ) {\n this.#llm = llm;\n this.#chatCtx = chatCtx;\n this.#toolCtx = toolCtx;\n this._connOptions = connOptions;\n this.monitorMetrics();\n this.abortController.signal.addEventListener('abort', () => {\n // TODO (AJS-37) clean this up when we refactor with streams\n this.output.close();\n this.closed = true;\n });\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 // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('llm_request', ..., { endOnExit: false })\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('llm_request_run', ...)\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 generate LLM completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n this.emitError({ error, recoverable: true });\n this.logger.warn(\n { llm: this.#llm.label(), attempt: i + 1, error },\n `failed to generate LLM completion, 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.#llm.emit('error', {\n type: 'llm_error',\n timestamp: Date.now(),\n label: this.#llm.label(),\n error,\n recoverable,\n });\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let ttft: bigint = BigInt(-1);\n let requestId = '';\n let usage: CompletionUsage | undefined;\n\n for await (const ev of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(ev);\n requestId = ev.id;\n if (ttft === BigInt(-1)) {\n ttft = process.hrtime.bigint() - startTime;\n }\n if (ev.usage) {\n usage = ev.usage;\n }\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const durationMs = Math.trunc(Number(duration / BigInt(1000000)));\n const metrics: LLMMetrics = {\n type: 'llm_metrics',\n timestamp: Date.now(),\n requestId,\n ttftMs: ttft === BigInt(-1) ? -1 : Math.trunc(Number(ttft / BigInt(1000000))),\n durationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#llm.label(),\n completionTokens: usage?.completionTokens || 0,\n promptTokens: usage?.promptTokens || 0,\n promptCachedTokens: usage?.promptCachedTokens || 0,\n totalTokens: usage?.totalTokens || 0,\n tokensPerSecond: (() => {\n if (durationMs <= 0) {\n return 0;\n }\n return (usage?.completionTokens || 0) / (durationMs / 1000);\n })(),\n };\n this.#llm.emit('metrics_collected', metrics);\n }\n\n protected abstract run(): Promise<void>;\n\n /** The function context of this stream. */\n get toolCtx(): ToolContext | undefined {\n return this.#toolCtx;\n }\n\n /** The initial chat context of this stream. */\n get chatCtx(): ChatContext {\n return this.#chatCtx;\n }\n\n /** The connection options for this stream. */\n get connOptions(): APIConnectOptions {\n return this._connOptions;\n }\n\n next(): Promise<IteratorResult<ChatChunk>> {\n return this.output.next();\n }\n\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): LLMStream {\n return this;\n }\n}\n"],"mappings":"AAIA,SAAS,oBAAoB;AAC7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAGpB,SAAS,oBAAoB,OAAO,WAAW,eAAe;AAC9D,eAAmE;AAmC5D,MAAe,YAAa,aAAsD;AAAA,EACvF,cAAc;AACZ,UAAM;AAAA,EACR;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAwBA,UAAgB;AAAA,EAEhB;AAAA,EAEA,MAAM,SAAwB;AAAA,EAE9B;AACF;AAEO,MAAe,UAAsD;AAAA,EAChE,SAAS,IAAI,mBAA8B;AAAA,EAC3C,QAAQ,IAAI,mBAA8B;AAAA,EAC1C,SAAS;AAAA,EACT,kBAAkB,IAAI,gBAAgB;AAAA,EACtC;AAAA,EACA,SAAS,IAAI;AAAA,EAEvB;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YACE,KACA;AAAA,IACE;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKA;AACA,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,WAAW;AAChB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAE1D,WAAK,OAAO,MAAM;AAClB,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,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,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AACL,iBAAK,UAAU,EAAE,OAAO,aAAa,KAAK,CAAC;AAC3C,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,MAAM,GAAG,SAAS,IAAI,GAAG,MAAM;AAAA,cAChD,kDAAkD,aAAa;AAAA,YACjE;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,MAAM;AAAA,MACvB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAChB,QAAI;AAEJ,qBAAiB,MAAM,KAAK,OAAO;AACjC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,EAAE;AAClB,kBAAY,GAAG;AACf,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,UAAI,GAAG,OAAO;AACZ,gBAAQ,GAAG;AAAA,MACb;AAAA,IACF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,aAAa,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAChE,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E;AAAA,MACA,WAAW,KAAK,gBAAgB,OAAO;AAAA,MACvC,OAAO,KAAK,KAAK,MAAM;AAAA,MACvB,mBAAkB,+BAAO,qBAAoB;AAAA,MAC7C,eAAc,+BAAO,iBAAgB;AAAA,MACrC,qBAAoB,+BAAO,uBAAsB;AAAA,MACjD,cAAa,+BAAO,gBAAe;AAAA,MACnC,kBAAkB,MAAM;AACtB,YAAI,cAAc,GAAG;AACnB,iBAAO;AAAA,QACT;AACA,iBAAQ,+BAAO,qBAAoB,MAAM,aAAa;AAAA,MACxD,GAAG;AAAA,IACL;AACA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAKA,IAAI,UAAmC;AACrC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAiC;AACnC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA,EAEA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAe;AAClC,WAAO;AAAA,EACT;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/llm/llm.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Span } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { LLMMetrics } from '../metrics/base.js';\nimport { recordException, traceTypes, tracer } from '../telemetry/index.js';\nimport type { APIConnectOptions } from '../types.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\nimport { type ChatContext, type ChatRole, type FunctionCall } from './chat_context.js';\nimport type { ToolChoice, ToolContext } from './tool_context.js';\n\nexport interface ChoiceDelta {\n role: ChatRole;\n content?: string;\n toolCalls?: FunctionCall[];\n}\n\nexport interface CompletionUsage {\n completionTokens: number;\n promptTokens: number;\n promptCachedTokens: number;\n totalTokens: number;\n}\n\nexport interface ChatChunk {\n id: string;\n delta?: ChoiceDelta;\n usage?: CompletionUsage;\n}\n\nexport interface LLMError {\n type: 'llm_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type LLMCallbacks = {\n ['metrics_collected']: (metrics: LLMMetrics) => void;\n ['error']: (error: LLMError) => void;\n};\n\nexport abstract class LLM extends (EventEmitter as new () => TypedEmitter<LLMCallbacks>) {\n constructor() {\n super();\n }\n\n abstract label(): string;\n\n /**\n * Get the model name/identifier for this LLM instance.\n *\n * @returns The model name if available, \"unknown\" otherwise.\n *\n * @remarks\n * Plugins should override this property to provide their model information.\n */\n get model(): string {\n return 'unknown';\n }\n\n /**\n * Returns a {@link LLMStream} that can be used to push text and receive LLM responses.\n */\n abstract chat({\n chatCtx,\n toolCtx,\n connOptions,\n parallelToolCalls,\n toolChoice,\n extraKwargs,\n }: {\n chatCtx: ChatContext;\n toolCtx?: ToolContext;\n connOptions?: APIConnectOptions;\n parallelToolCalls?: boolean;\n toolChoice?: ToolChoice;\n extraKwargs?: Record<string, unknown>;\n }): LLMStream;\n\n /**\n * Pre-warm connection to the LLM service\n */\n prewarm(): void {\n // Default implementation - subclasses can override\n }\n\n async aclose(): Promise<void> {\n // Default implementation - subclasses can override\n }\n}\n\nexport abstract class LLMStream implements AsyncIterableIterator<ChatChunk> {\n protected output = new AsyncIterableQueue<ChatChunk>();\n protected queue = new AsyncIterableQueue<ChatChunk>();\n protected closed = false;\n protected abortController = new AbortController();\n protected _connOptions: APIConnectOptions;\n protected logger = log();\n\n #llm: LLM;\n #chatCtx: ChatContext;\n #toolCtx?: ToolContext;\n #llmRequestSpan?: Span;\n\n constructor(\n llm: LLM,\n {\n chatCtx,\n toolCtx,\n connOptions,\n }: {\n chatCtx: ChatContext;\n toolCtx?: ToolContext;\n connOptions: APIConnectOptions;\n },\n ) {\n this.#llm = llm;\n this.#chatCtx = chatCtx;\n this.#toolCtx = toolCtx;\n this._connOptions = connOptions;\n this.monitorMetrics();\n this.abortController.signal.addEventListener('abort', () => {\n // TODO (AJS-37) clean this up when we refactor with streams\n this.output.close();\n this.closed = true;\n });\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 _mainTaskImpl = async (span: Span) => {\n this.#llmRequestSpan = span;\n span.setAttribute(traceTypes.ATTR_GEN_AI_REQUEST_MODEL, this.#llm.model);\n\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'llm_request_run' },\n );\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 generate LLM completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n this.emitError({ error, recoverable: true });\n this.logger.warn(\n { llm: this.#llm.label(), attempt: i + 1, error },\n `failed to generate LLM completion, 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 mainTask = async () =>\n tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'llm_request',\n endOnExit: false,\n });\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#llm.emit('error', {\n type: 'llm_error',\n timestamp: Date.now(),\n label: this.#llm.label(),\n error,\n recoverable,\n });\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let ttft: bigint = BigInt(-1);\n let requestId = '';\n let usage: CompletionUsage | undefined;\n let completionStartTime: string | undefined;\n\n for await (const ev of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(ev);\n requestId = ev.id;\n if (ttft === BigInt(-1)) {\n ttft = process.hrtime.bigint() - startTime;\n completionStartTime = new Date().toISOString();\n }\n if (ev.usage) {\n usage = ev.usage;\n }\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const durationMs = Math.trunc(Number(duration / BigInt(1000000)));\n const metrics: LLMMetrics = {\n type: 'llm_metrics',\n timestamp: Date.now(),\n requestId,\n ttftMs: ttft === BigInt(-1) ? -1 : Math.trunc(Number(ttft / BigInt(1000000))),\n durationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#llm.label(),\n completionTokens: usage?.completionTokens || 0,\n promptTokens: usage?.promptTokens || 0,\n promptCachedTokens: usage?.promptCachedTokens || 0,\n totalTokens: usage?.totalTokens || 0,\n tokensPerSecond: (() => {\n if (durationMs <= 0) {\n return 0;\n }\n return (usage?.completionTokens || 0) / (durationMs / 1000);\n })(),\n };\n\n if (this.#llmRequestSpan) {\n this.#llmRequestSpan.setAttribute(traceTypes.ATTR_LLM_METRICS, JSON.stringify(metrics));\n\n this.#llmRequestSpan.setAttributes({\n [traceTypes.ATTR_GEN_AI_USAGE_INPUT_TOKENS]: metrics.promptTokens,\n [traceTypes.ATTR_GEN_AI_USAGE_OUTPUT_TOKENS]: metrics.completionTokens,\n });\n\n if (completionStartTime) {\n this.#llmRequestSpan.setAttribute(\n traceTypes.ATTR_LANGFUSE_COMPLETION_START_TIME,\n completionStartTime,\n );\n }\n\n // End the span now that metrics are collected\n this.#llmRequestSpan.end();\n }\n\n this.#llm.emit('metrics_collected', metrics);\n }\n\n protected abstract run(): Promise<void>;\n\n /** The function context of this stream. */\n get toolCtx(): ToolContext | undefined {\n return this.#toolCtx;\n }\n\n /** The initial chat context of this stream. */\n get chatCtx(): ChatContext {\n return this.#chatCtx;\n }\n\n /** The connection options for this stream. */\n get connOptions(): APIConnectOptions {\n return this._connOptions;\n }\n\n next(): Promise<IteratorResult<ChatChunk>> {\n return this.output.next();\n }\n\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): LLMStream {\n return this;\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAC7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,iBAAiB,YAAY,cAAc;AAEpD,SAAS,oBAAoB,OAAO,WAAW,eAAe;AAC9D,eAAmE;AAmC5D,MAAe,YAAa,aAAsD;AAAA,EACvF,cAAc;AACZ,UAAM;AAAA,EACR;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAwBA,UAAgB;AAAA,EAEhB;AAAA,EAEA,MAAM,SAAwB;AAAA,EAE9B;AACF;AAEO,MAAe,UAAsD;AAAA,EAChE,SAAS,IAAI,mBAA8B;AAAA,EAC3C,QAAQ,IAAI,mBAA8B;AAAA,EAC1C,SAAS;AAAA,EACT,kBAAkB,IAAI,gBAAgB;AAAA,EACtC;AAAA,EACA,SAAS,IAAI;AAAA,EAEvB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YACE,KACA;AAAA,IACE;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKA;AACA,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,WAAW;AAChB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAE1D,WAAK,OAAO,MAAM;AAClB,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,aAAa,WAAW,2BAA2B,KAAK,KAAK,KAAK;AAEvE,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,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,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AACL,iBAAK,UAAU,EAAE,OAAO,aAAa,KAAK,CAAC;AAC3C,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,MAAM,GAAG,SAAS,IAAI,GAAG,MAAM;AAAA,cAChD,kDAAkD,aAAa;AAAA,YACjE;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,WAAW,YACjB,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,IAC/D,MAAM;AAAA,IACN,WAAW;AAAA,EACb,CAAC;AAAA,EAEK,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK,MAAM;AAAA,MACvB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAChB,QAAI;AACJ,QAAI;AAEJ,qBAAiB,MAAM,KAAK,OAAO;AACjC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,EAAE;AAClB,kBAAY,GAAG;AACf,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AACjC,+BAAsB,oBAAI,KAAK,GAAE,YAAY;AAAA,MAC/C;AACA,UAAI,GAAG,OAAO;AACZ,gBAAQ,GAAG;AAAA,MACb;AAAA,IACF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,aAAa,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAChE,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E;AAAA,MACA,WAAW,KAAK,gBAAgB,OAAO;AAAA,MACvC,OAAO,KAAK,KAAK,MAAM;AAAA,MACvB,mBAAkB,+BAAO,qBAAoB;AAAA,MAC7C,eAAc,+BAAO,iBAAgB;AAAA,MACrC,qBAAoB,+BAAO,uBAAsB;AAAA,MACjD,cAAa,+BAAO,gBAAe;AAAA,MACnC,kBAAkB,MAAM;AACtB,YAAI,cAAc,GAAG;AACnB,iBAAO;AAAA,QACT;AACA,iBAAQ,+BAAO,qBAAoB,MAAM,aAAa;AAAA,MACxD,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AAEtF,WAAK,gBAAgB,cAAc;AAAA,QACjC,CAAC,WAAW,8BAA8B,GAAG,QAAQ;AAAA,QACrD,CAAC,WAAW,+BAA+B,GAAG,QAAQ;AAAA,MACxD,CAAC;AAED,UAAI,qBAAqB;AACvB,aAAK,gBAAgB;AAAA,UACnB,WAAW;AAAA,UACX;AAAA,QACF;AAAA,MACF;AAGA,WAAK,gBAAgB,IAAI;AAAA,IAC3B;AAEA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAKA,IAAI,UAAmC;AACrC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAiC;AACnC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA,EAEA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAe;AAClC,WAAO;AAAA,EACT;AACF;","names":[]}
package/dist/log.cjs CHANGED
@@ -18,14 +18,19 @@ var __copyProps = (to, from, except, desc) => {
18
18
  var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
19
19
  var log_exports = {};
20
20
  __export(log_exports, {
21
+ enableOtelLogging: () => enableOtelLogging,
21
22
  initializeLogger: () => initializeLogger,
22
23
  log: () => log,
23
24
  loggerOptions: () => loggerOptions
24
25
  });
25
26
  module.exports = __toCommonJS(log_exports);
27
+ var import_node_stream = require("node:stream");
26
28
  var import_pino = require("pino");
29
+ var import_pino_pretty = require("pino-pretty");
30
+ var import_pino_otel_transport = require("./telemetry/pino_otel_transport.cjs");
27
31
  let loggerOptions;
28
32
  let logger = void 0;
33
+ let otelEnabled = false;
29
34
  const log = () => {
30
35
  if (!logger) {
31
36
  throw new TypeError("logger not initialized. did you forget to run initializeLogger()?");
@@ -35,21 +40,40 @@ const log = () => {
35
40
  const initializeLogger = ({ pretty, level }) => {
36
41
  loggerOptions = { pretty, level };
37
42
  logger = (0, import_pino.pino)(
38
- pretty ? {
39
- transport: {
40
- target: "pino-pretty",
41
- options: {
42
- colorize: true
43
- }
44
- }
45
- } : {}
43
+ { level: level || "info" },
44
+ pretty ? (0, import_pino_pretty.build)({ colorize: true }) : process.stdout
46
45
  );
47
- if (level) {
48
- logger.level = level;
46
+ };
47
+ class OtelDestination extends import_node_stream.Writable {
48
+ _write(chunk, _encoding, callback) {
49
+ try {
50
+ const line = chunk.toString().trim();
51
+ if (line) {
52
+ const logObj = JSON.parse(line);
53
+ (0, import_pino_otel_transport.emitToOtel)(logObj);
54
+ }
55
+ } catch {
56
+ }
57
+ callback();
58
+ }
59
+ }
60
+ const enableOtelLogging = () => {
61
+ if (otelEnabled || !logger) {
62
+ console.warn("OTEL logging already enabled or logger not initialized");
63
+ return;
49
64
  }
65
+ otelEnabled = true;
66
+ const { pretty, level } = loggerOptions;
67
+ const logLevel = level || "info";
68
+ const streams = [
69
+ { stream: pretty ? (0, import_pino_pretty.build)({ colorize: true }) : process.stdout, level: logLevel },
70
+ { stream: new OtelDestination(), level: "debug" }
71
+ ];
72
+ logger = (0, import_pino.pino)({ level: logLevel }, (0, import_pino.multistream)(streams));
50
73
  };
51
74
  // Annotate the CommonJS export names for ESM import in node:
52
75
  0 && (module.exports = {
76
+ enableOtelLogging,
53
77
  initializeLogger,
54
78
  log,
55
79
  loggerOptions
package/dist/log.cjs.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/log.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { Logger } from 'pino';\nimport { pino } from 'pino';\n\n/** @internal */\nexport type LoggerOptions = {\n pretty: boolean;\n level?: string;\n};\n\n/** @internal */\nexport let loggerOptions: LoggerOptions;\n\n/** @internal */\nlet logger: Logger | undefined = undefined;\n\n/** @internal */\nexport const log = () => {\n if (!logger) {\n throw new TypeError('logger not initialized. did you forget to run initializeLogger()?');\n }\n return logger;\n};\n\n/** @internal */\nexport const initializeLogger = ({ pretty, level }: LoggerOptions) => {\n loggerOptions = { pretty, level };\n logger = pino(\n pretty\n ? {\n transport: {\n target: 'pino-pretty',\n options: {\n colorize: true,\n },\n },\n }\n : {},\n );\n if (level) {\n logger.level = level;\n }\n // TODO(brian): PR4 - Add Pino bridge to OTEL LoggingHandler for structured logging integration\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,kBAAqB;AASd,IAAI;AAGX,IAAI,SAA6B;AAG1B,MAAM,MAAM,MAAM;AACvB,MAAI,CAAC,QAAQ;AACX,UAAM,IAAI,UAAU,mEAAmE;AAAA,EACzF;AACA,SAAO;AACT;AAGO,MAAM,mBAAmB,CAAC,EAAE,QAAQ,MAAM,MAAqB;AACpE,kBAAgB,EAAE,QAAQ,MAAM;AAChC,eAAS;AAAA,IACP,SACI;AAAA,MACE,WAAW;AAAA,QACT,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,IACA,CAAC;AAAA,EACP;AACA,MAAI,OAAO;AACT,WAAO,QAAQ;AAAA,EACjB;AAEF;","names":[]}
1
+ {"version":3,"sources":["../src/log.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { Writable } from 'node:stream';\nimport type { DestinationStream, Logger } from 'pino';\nimport { multistream, pino } from 'pino';\nimport { build as pinoPretty } from 'pino-pretty';\nimport { type PinoLogObject, emitToOtel } from './telemetry/pino_otel_transport.js';\n\n/** @internal */\nexport type LoggerOptions = {\n pretty: boolean;\n level?: string;\n};\n\n/** @internal */\nexport let loggerOptions: LoggerOptions;\n\n/** @internal */\nlet logger: Logger | undefined = undefined;\n\n/** @internal */\nlet otelEnabled = false;\n\n/** @internal */\nexport const log = () => {\n if (!logger) {\n throw new TypeError('logger not initialized. did you forget to run initializeLogger()?');\n }\n return logger;\n};\n\n/** @internal */\nexport const initializeLogger = ({ pretty, level }: LoggerOptions) => {\n loggerOptions = { pretty, level };\n logger = pino(\n { level: level || 'info' },\n pretty ? pinoPretty({ colorize: true }) : process.stdout,\n );\n};\n\n/**\n * Custom Pino destination that parses JSON logs and emits to OTEL.\n * This receives the FULL serialized log including msg, level, time, etc.\n */\nclass OtelDestination extends Writable {\n _write(chunk: Buffer, _encoding: string, callback: (error?: Error | null) => void): void {\n try {\n const line = chunk.toString().trim();\n if (line) {\n const logObj = JSON.parse(line) as PinoLogObject;\n emitToOtel(logObj);\n }\n } catch {\n // Ignore parse errors (e.g., non-JSON lines)\n }\n callback();\n }\n}\n\n/**\n * Enable OTEL logging by reconfiguring the logger with multistream.\n * Uses a custom destination that receives full JSON logs (with msg, level, time).\n *\n * @internal\n */\nexport const enableOtelLogging = () => {\n if (otelEnabled || !logger) {\n console.warn('OTEL logging already enabled or logger not initialized');\n return;\n }\n otelEnabled = true;\n\n const { pretty, level } = loggerOptions;\n\n const logLevel = level || 'info';\n const streams: { stream: DestinationStream; level: string }[] = [\n { stream: pretty ? pinoPretty({ colorize: true }) : process.stdout, level: logLevel },\n { stream: new OtelDestination(), level: 'debug' },\n ];\n\n logger = pino({ level: logLevel }, multistream(streams));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,yBAAyB;AAEzB,kBAAkC;AAClC,yBAAoC;AACpC,iCAA+C;AASxC,IAAI;AAGX,IAAI,SAA6B;AAGjC,IAAI,cAAc;AAGX,MAAM,MAAM,MAAM;AACvB,MAAI,CAAC,QAAQ;AACX,UAAM,IAAI,UAAU,mEAAmE;AAAA,EACzF;AACA,SAAO;AACT;AAGO,MAAM,mBAAmB,CAAC,EAAE,QAAQ,MAAM,MAAqB;AACpE,kBAAgB,EAAE,QAAQ,MAAM;AAChC,eAAS;AAAA,IACP,EAAE,OAAO,SAAS,OAAO;AAAA,IACzB,aAAS,mBAAAA,OAAW,EAAE,UAAU,KAAK,CAAC,IAAI,QAAQ;AAAA,EACpD;AACF;AAMA,MAAM,wBAAwB,4BAAS;AAAA,EACrC,OAAO,OAAe,WAAmB,UAAgD;AACvF,QAAI;AACF,YAAM,OAAO,MAAM,SAAS,EAAE,KAAK;AACnC,UAAI,MAAM;AACR,cAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,mDAAW,MAAM;AAAA,MACnB;AAAA,IACF,QAAQ;AAAA,IAER;AACA,aAAS;AAAA,EACX;AACF;AAQO,MAAM,oBAAoB,MAAM;AACrC,MAAI,eAAe,CAAC,QAAQ;AAC1B,YAAQ,KAAK,wDAAwD;AACrE;AAAA,EACF;AACA,gBAAc;AAEd,QAAM,EAAE,QAAQ,MAAM,IAAI;AAE1B,QAAM,WAAW,SAAS;AAC1B,QAAM,UAA0D;AAAA,IAC9D,EAAE,QAAQ,aAAS,mBAAAA,OAAW,EAAE,UAAU,KAAK,CAAC,IAAI,QAAQ,QAAQ,OAAO,SAAS;AAAA,IACpF,EAAE,QAAQ,IAAI,gBAAgB,GAAG,OAAO,QAAQ;AAAA,EAClD;AAEA,eAAS,kBAAK,EAAE,OAAO,SAAS,OAAG,yBAAY,OAAO,CAAC;AACzD;","names":["pinoPretty"]}
package/dist/log.d.cts CHANGED
@@ -10,4 +10,11 @@ export declare let loggerOptions: LoggerOptions;
10
10
  export declare const log: () => Logger;
11
11
  /** @internal */
12
12
  export declare const initializeLogger: ({ pretty, level }: LoggerOptions) => void;
13
+ /**
14
+ * Enable OTEL logging by reconfiguring the logger with multistream.
15
+ * Uses a custom destination that receives full JSON logs (with msg, level, time).
16
+ *
17
+ * @internal
18
+ */
19
+ export declare const enableOtelLogging: () => void;
13
20
  //# sourceMappingURL=log.d.ts.map
package/dist/log.d.ts CHANGED
@@ -10,4 +10,11 @@ export declare let loggerOptions: LoggerOptions;
10
10
  export declare const log: () => Logger;
11
11
  /** @internal */
12
12
  export declare const initializeLogger: ({ pretty, level }: LoggerOptions) => void;
13
+ /**
14
+ * Enable OTEL logging by reconfiguring the logger with multistream.
15
+ * Uses a custom destination that receives full JSON logs (with msg, level, time).
16
+ *
17
+ * @internal
18
+ */
19
+ export declare const enableOtelLogging: () => void;
13
20
  //# sourceMappingURL=log.d.ts.map
package/dist/log.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"log.d.ts","sourceRoot":"","sources":["../src/log.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,MAAM,CAAC;AAGnC,gBAAgB;AAChB,MAAM,MAAM,aAAa,GAAG;IAC1B,MAAM,EAAE,OAAO,CAAC;IAChB,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB,CAAC;AAEF,gBAAgB;AAChB,eAAO,IAAI,aAAa,EAAE,aAAa,CAAC;AAKxC,gBAAgB;AAChB,eAAO,MAAM,GAAG,cAKf,CAAC;AAEF,gBAAgB;AAChB,eAAO,MAAM,gBAAgB,sBAAuB,aAAa,SAkBhE,CAAC"}
1
+ {"version":3,"file":"log.d.ts","sourceRoot":"","sources":["../src/log.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAAqB,MAAM,EAAE,MAAM,MAAM,CAAC;AAKtD,gBAAgB;AAChB,MAAM,MAAM,aAAa,GAAG;IAC1B,MAAM,EAAE,OAAO,CAAC;IAChB,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB,CAAC;AAEF,gBAAgB;AAChB,eAAO,IAAI,aAAa,EAAE,aAAa,CAAC;AAQxC,gBAAgB;AAChB,eAAO,MAAM,GAAG,cAKf,CAAC;AAEF,gBAAgB;AAChB,eAAO,MAAM,gBAAgB,sBAAuB,aAAa,SAMhE,CAAC;AAqBF;;;;;GAKG;AACH,eAAO,MAAM,iBAAiB,YAgB7B,CAAC"}
package/dist/log.js CHANGED
@@ -1,6 +1,10 @@
1
- import { pino } from "pino";
1
+ import { Writable } from "node:stream";
2
+ import { multistream, pino } from "pino";
3
+ import { build as pinoPretty } from "pino-pretty";
4
+ import { emitToOtel } from "./telemetry/pino_otel_transport.js";
2
5
  let loggerOptions;
3
6
  let logger = void 0;
7
+ let otelEnabled = false;
4
8
  const log = () => {
5
9
  if (!logger) {
6
10
  throw new TypeError("logger not initialized. did you forget to run initializeLogger()?");
@@ -10,20 +14,39 @@ const log = () => {
10
14
  const initializeLogger = ({ pretty, level }) => {
11
15
  loggerOptions = { pretty, level };
12
16
  logger = pino(
13
- pretty ? {
14
- transport: {
15
- target: "pino-pretty",
16
- options: {
17
- colorize: true
18
- }
19
- }
20
- } : {}
17
+ { level: level || "info" },
18
+ pretty ? pinoPretty({ colorize: true }) : process.stdout
21
19
  );
22
- if (level) {
23
- logger.level = level;
20
+ };
21
+ class OtelDestination extends Writable {
22
+ _write(chunk, _encoding, callback) {
23
+ try {
24
+ const line = chunk.toString().trim();
25
+ if (line) {
26
+ const logObj = JSON.parse(line);
27
+ emitToOtel(logObj);
28
+ }
29
+ } catch {
30
+ }
31
+ callback();
32
+ }
33
+ }
34
+ const enableOtelLogging = () => {
35
+ if (otelEnabled || !logger) {
36
+ console.warn("OTEL logging already enabled or logger not initialized");
37
+ return;
24
38
  }
39
+ otelEnabled = true;
40
+ const { pretty, level } = loggerOptions;
41
+ const logLevel = level || "info";
42
+ const streams = [
43
+ { stream: pretty ? pinoPretty({ colorize: true }) : process.stdout, level: logLevel },
44
+ { stream: new OtelDestination(), level: "debug" }
45
+ ];
46
+ logger = pino({ level: logLevel }, multistream(streams));
25
47
  };
26
48
  export {
49
+ enableOtelLogging,
27
50
  initializeLogger,
28
51
  log,
29
52
  loggerOptions
package/dist/log.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/log.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { Logger } from 'pino';\nimport { pino } from 'pino';\n\n/** @internal */\nexport type LoggerOptions = {\n pretty: boolean;\n level?: string;\n};\n\n/** @internal */\nexport let loggerOptions: LoggerOptions;\n\n/** @internal */\nlet logger: Logger | undefined = undefined;\n\n/** @internal */\nexport const log = () => {\n if (!logger) {\n throw new TypeError('logger not initialized. did you forget to run initializeLogger()?');\n }\n return logger;\n};\n\n/** @internal */\nexport const initializeLogger = ({ pretty, level }: LoggerOptions) => {\n loggerOptions = { pretty, level };\n logger = pino(\n pretty\n ? {\n transport: {\n target: 'pino-pretty',\n options: {\n colorize: true,\n },\n },\n }\n : {},\n );\n if (level) {\n logger.level = level;\n }\n // TODO(brian): PR4 - Add Pino bridge to OTEL LoggingHandler for structured logging integration\n};\n"],"mappings":"AAIA,SAAS,YAAY;AASd,IAAI;AAGX,IAAI,SAA6B;AAG1B,MAAM,MAAM,MAAM;AACvB,MAAI,CAAC,QAAQ;AACX,UAAM,IAAI,UAAU,mEAAmE;AAAA,EACzF;AACA,SAAO;AACT;AAGO,MAAM,mBAAmB,CAAC,EAAE,QAAQ,MAAM,MAAqB;AACpE,kBAAgB,EAAE,QAAQ,MAAM;AAChC,WAAS;AAAA,IACP,SACI;AAAA,MACE,WAAW;AAAA,QACT,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,IACA,CAAC;AAAA,EACP;AACA,MAAI,OAAO;AACT,WAAO,QAAQ;AAAA,EACjB;AAEF;","names":[]}
1
+ {"version":3,"sources":["../src/log.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { Writable } from 'node:stream';\nimport type { DestinationStream, Logger } from 'pino';\nimport { multistream, pino } from 'pino';\nimport { build as pinoPretty } from 'pino-pretty';\nimport { type PinoLogObject, emitToOtel } from './telemetry/pino_otel_transport.js';\n\n/** @internal */\nexport type LoggerOptions = {\n pretty: boolean;\n level?: string;\n};\n\n/** @internal */\nexport let loggerOptions: LoggerOptions;\n\n/** @internal */\nlet logger: Logger | undefined = undefined;\n\n/** @internal */\nlet otelEnabled = false;\n\n/** @internal */\nexport const log = () => {\n if (!logger) {\n throw new TypeError('logger not initialized. did you forget to run initializeLogger()?');\n }\n return logger;\n};\n\n/** @internal */\nexport const initializeLogger = ({ pretty, level }: LoggerOptions) => {\n loggerOptions = { pretty, level };\n logger = pino(\n { level: level || 'info' },\n pretty ? pinoPretty({ colorize: true }) : process.stdout,\n );\n};\n\n/**\n * Custom Pino destination that parses JSON logs and emits to OTEL.\n * This receives the FULL serialized log including msg, level, time, etc.\n */\nclass OtelDestination extends Writable {\n _write(chunk: Buffer, _encoding: string, callback: (error?: Error | null) => void): void {\n try {\n const line = chunk.toString().trim();\n if (line) {\n const logObj = JSON.parse(line) as PinoLogObject;\n emitToOtel(logObj);\n }\n } catch {\n // Ignore parse errors (e.g., non-JSON lines)\n }\n callback();\n }\n}\n\n/**\n * Enable OTEL logging by reconfiguring the logger with multistream.\n * Uses a custom destination that receives full JSON logs (with msg, level, time).\n *\n * @internal\n */\nexport const enableOtelLogging = () => {\n if (otelEnabled || !logger) {\n console.warn('OTEL logging already enabled or logger not initialized');\n return;\n }\n otelEnabled = true;\n\n const { pretty, level } = loggerOptions;\n\n const logLevel = level || 'info';\n const streams: { stream: DestinationStream; level: string }[] = [\n { stream: pretty ? pinoPretty({ colorize: true }) : process.stdout, level: logLevel },\n { stream: new OtelDestination(), level: 'debug' },\n ];\n\n logger = pino({ level: logLevel }, multistream(streams));\n};\n"],"mappings":"AAGA,SAAS,gBAAgB;AAEzB,SAAS,aAAa,YAAY;AAClC,SAAS,SAAS,kBAAkB;AACpC,SAA6B,kBAAkB;AASxC,IAAI;AAGX,IAAI,SAA6B;AAGjC,IAAI,cAAc;AAGX,MAAM,MAAM,MAAM;AACvB,MAAI,CAAC,QAAQ;AACX,UAAM,IAAI,UAAU,mEAAmE;AAAA,EACzF;AACA,SAAO;AACT;AAGO,MAAM,mBAAmB,CAAC,EAAE,QAAQ,MAAM,MAAqB;AACpE,kBAAgB,EAAE,QAAQ,MAAM;AAChC,WAAS;AAAA,IACP,EAAE,OAAO,SAAS,OAAO;AAAA,IACzB,SAAS,WAAW,EAAE,UAAU,KAAK,CAAC,IAAI,QAAQ;AAAA,EACpD;AACF;AAMA,MAAM,wBAAwB,SAAS;AAAA,EACrC,OAAO,OAAe,WAAmB,UAAgD;AACvF,QAAI;AACF,YAAM,OAAO,MAAM,SAAS,EAAE,KAAK;AACnC,UAAI,MAAM;AACR,cAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,mBAAW,MAAM;AAAA,MACnB;AAAA,IACF,QAAQ;AAAA,IAER;AACA,aAAS;AAAA,EACX;AACF;AAQO,MAAM,oBAAoB,MAAM;AACrC,MAAI,eAAe,CAAC,QAAQ;AAC1B,YAAQ,KAAK,wDAAwD;AACrE;AAAA,EACF;AACA,gBAAc;AAEd,QAAM,EAAE,QAAQ,MAAM,IAAI;AAE1B,QAAM,WAAW,SAAS;AAC1B,QAAM,UAA0D;AAAA,IAC9D,EAAE,QAAQ,SAAS,WAAW,EAAE,UAAU,KAAK,CAAC,IAAI,QAAQ,QAAQ,OAAO,SAAS;AAAA,IACpF,EAAE,QAAQ,IAAI,gBAAgB,GAAG,OAAO,QAAQ;AAAA,EAClD;AAEA,WAAS,KAAK,EAAE,OAAO,SAAS,GAAG,YAAY,OAAO,CAAC;AACzD;","names":[]}
package/dist/stt/stt.cjs CHANGED
@@ -149,7 +149,18 @@ class SpeechStream {
149
149
  }
150
150
  async monitorMetrics() {
151
151
  for await (const event of this.queue) {
152
- this.output.put(event);
152
+ if (!this.output.closed) {
153
+ try {
154
+ this.output.put(event);
155
+ } catch (e) {
156
+ if (e instanceof Error && e.message.includes("Queue is closed")) {
157
+ this.logger.warn(
158
+ { err: e },
159
+ "Queue closed during transcript processing (expected during disconnect)"
160
+ );
161
+ }
162
+ }
163
+ }
153
164
  if (event.type !== 4 /* RECOGNITION_USAGE */) continue;
154
165
  const metrics = {
155
166
  type: "stt_metrics",
@@ -162,7 +173,9 @@ class SpeechStream {
162
173
  };
163
174
  this.#stt.emit("metrics_collected", metrics);
164
175
  }
165
- this.output.close();
176
+ if (!this.output.closed) {
177
+ this.output.close();
178
+ }
166
179
  }
167
180
  updateInputStream(audioStream) {
168
181
  this.deferredInputStream.setSource(audioStream);
@@ -217,9 +230,9 @@ class SpeechStream {
217
230
  }
218
231
  /** Close both the input and output of the STT stream */
219
232
  close() {
220
- this.input.close();
221
- this.queue.close();
222
- this.output.close();
233
+ if (!this.input.closed) this.input.close();
234
+ if (!this.queue.closed) this.queue.close();
235
+ if (!this.output.closed) this.output.close();
223
236
  this.closed = true;
224
237
  }
225
238
  [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 this.output.put(event);\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 this.output.close();\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 this.input.close();\n this.queue.close();\n 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,WAAK,OAAO,IAAI,KAAK;AACrB,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,SAAK,OAAO,MAAM;AAAA,EACpB;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,SAAK,MAAM,MAAM;AACjB,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,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 } 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 +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;IAkB9B,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,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"}
package/dist/stt/stt.js CHANGED
@@ -124,7 +124,18 @@ class SpeechStream {
124
124
  }
125
125
  async monitorMetrics() {
126
126
  for await (const event of this.queue) {
127
- this.output.put(event);
127
+ if (!this.output.closed) {
128
+ try {
129
+ this.output.put(event);
130
+ } catch (e) {
131
+ if (e instanceof Error && e.message.includes("Queue is closed")) {
132
+ this.logger.warn(
133
+ { err: e },
134
+ "Queue closed during transcript processing (expected during disconnect)"
135
+ );
136
+ }
137
+ }
138
+ }
128
139
  if (event.type !== 4 /* RECOGNITION_USAGE */) continue;
129
140
  const metrics = {
130
141
  type: "stt_metrics",
@@ -137,7 +148,9 @@ class SpeechStream {
137
148
  };
138
149
  this.#stt.emit("metrics_collected", metrics);
139
150
  }
140
- this.output.close();
151
+ if (!this.output.closed) {
152
+ this.output.close();
153
+ }
141
154
  }
142
155
  updateInputStream(audioStream) {
143
156
  this.deferredInputStream.setSource(audioStream);
@@ -192,9 +205,9 @@ class SpeechStream {
192
205
  }
193
206
  /** Close both the input and output of the STT stream */
194
207
  close() {
195
- this.input.close();
196
- this.queue.close();
197
- this.output.close();
208
+ if (!this.input.closed) this.input.close();
209
+ if (!this.queue.closed) this.queue.close();
210
+ if (!this.output.closed) this.output.close();
198
211
  this.closed = true;
199
212
  }
200
213
  [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 this.output.put(event);\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 this.output.close();\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 this.input.close();\n this.queue.close();\n 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,WAAK,OAAO,IAAI,KAAK;AACrB,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,SAAK,OAAO,MAAM;AAAA,EACpB;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,SAAK,MAAM,MAAM;AACjB,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,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 } 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"]}