@livekit/agents 0.4.6 → 0.5.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (214) hide show
  1. package/README.md +17 -0
  2. package/dist/audio.cjs +77 -0
  3. package/dist/audio.cjs.map +1 -0
  4. package/dist/audio.js +48 -37
  5. package/dist/audio.js.map +1 -1
  6. package/dist/cli.cjs +131 -0
  7. package/dist/cli.cjs.map +1 -0
  8. package/dist/cli.js +96 -122
  9. package/dist/cli.js.map +1 -1
  10. package/dist/generator.cjs +36 -0
  11. package/dist/generator.cjs.map +1 -0
  12. package/dist/generator.js +8 -22
  13. package/dist/generator.js.map +1 -1
  14. package/dist/http_server.cjs +72 -0
  15. package/dist/http_server.cjs.map +1 -0
  16. package/dist/http_server.d.ts +1 -1
  17. package/dist/http_server.js +44 -47
  18. package/dist/http_server.js.map +1 -1
  19. package/dist/index.cjs +78 -0
  20. package/dist/index.cjs.map +1 -0
  21. package/dist/index.js +26 -28
  22. package/dist/index.js.map +1 -1
  23. package/dist/ipc/job_executor.cjs +33 -0
  24. package/dist/ipc/job_executor.cjs.map +1 -0
  25. package/dist/ipc/job_executor.js +7 -4
  26. package/dist/ipc/job_executor.js.map +1 -1
  27. package/dist/ipc/job_main.cjs +147 -0
  28. package/dist/ipc/job_main.cjs.map +1 -0
  29. package/dist/ipc/job_main.d.ts +1 -1
  30. package/dist/ipc/job_main.js +103 -103
  31. package/dist/ipc/job_main.js.map +1 -1
  32. package/dist/ipc/message.cjs +17 -0
  33. package/dist/ipc/message.cjs.map +1 -0
  34. package/dist/ipc/message.js +0 -1
  35. package/dist/ipc/message.js.map +1 -1
  36. package/dist/ipc/proc_job_executor.cjs +174 -0
  37. package/dist/ipc/proc_job_executor.cjs.map +1 -0
  38. package/dist/ipc/proc_job_executor.js +130 -126
  39. package/dist/ipc/proc_job_executor.js.map +1 -1
  40. package/dist/ipc/proc_pool.cjs +126 -0
  41. package/dist/ipc/proc_pool.cjs.map +1 -0
  42. package/dist/ipc/proc_pool.js +93 -96
  43. package/dist/ipc/proc_pool.js.map +1 -1
  44. package/dist/job.cjs +230 -0
  45. package/dist/job.cjs.map +1 -0
  46. package/dist/job.d.ts +6 -1
  47. package/dist/job.d.ts.map +1 -1
  48. package/dist/job.js +195 -198
  49. package/dist/job.js.map +1 -1
  50. package/dist/llm/chat_context.cjs +131 -0
  51. package/dist/llm/chat_context.cjs.map +1 -0
  52. package/dist/llm/chat_context.js +98 -86
  53. package/dist/llm/chat_context.js.map +1 -1
  54. package/dist/llm/function_context.cjs +103 -0
  55. package/dist/llm/function_context.cjs.map +1 -0
  56. package/dist/llm/function_context.js +72 -81
  57. package/dist/llm/function_context.js.map +1 -1
  58. package/dist/llm/function_context.test.cjs +218 -0
  59. package/dist/llm/function_context.test.cjs.map +1 -0
  60. package/dist/llm/function_context.test.js +209 -210
  61. package/dist/llm/function_context.test.js.map +1 -1
  62. package/dist/llm/index.cjs +43 -0
  63. package/dist/llm/index.cjs.map +1 -0
  64. package/dist/llm/index.js +22 -6
  65. package/dist/llm/index.js.map +1 -1
  66. package/dist/llm/llm.cjs +76 -0
  67. package/dist/llm/llm.cjs.map +1 -0
  68. package/dist/llm/llm.js +48 -42
  69. package/dist/llm/llm.js.map +1 -1
  70. package/dist/log.cjs +57 -0
  71. package/dist/log.cjs.map +1 -0
  72. package/dist/log.js +27 -26
  73. package/dist/log.js.map +1 -1
  74. package/dist/multimodal/agent_playout.cjs +228 -0
  75. package/dist/multimodal/agent_playout.cjs.map +1 -0
  76. package/dist/multimodal/agent_playout.d.ts +1 -1
  77. package/dist/multimodal/agent_playout.js +193 -180
  78. package/dist/multimodal/agent_playout.js.map +1 -1
  79. package/dist/multimodal/index.cjs +25 -0
  80. package/dist/multimodal/index.cjs.map +1 -0
  81. package/dist/multimodal/index.js +2 -5
  82. package/dist/multimodal/index.js.map +1 -1
  83. package/dist/multimodal/multimodal_agent.cjs +404 -0
  84. package/dist/multimodal/multimodal_agent.cjs.map +1 -0
  85. package/dist/multimodal/multimodal_agent.d.ts +1 -1
  86. package/dist/multimodal/multimodal_agent.js +351 -330
  87. package/dist/multimodal/multimodal_agent.js.map +1 -1
  88. package/dist/pipeline/agent_output.cjs +172 -0
  89. package/dist/pipeline/agent_output.cjs.map +1 -0
  90. package/dist/pipeline/agent_output.js +136 -138
  91. package/dist/pipeline/agent_output.js.map +1 -1
  92. package/dist/pipeline/agent_playout.cjs +169 -0
  93. package/dist/pipeline/agent_playout.cjs.map +1 -0
  94. package/dist/pipeline/agent_playout.js +126 -136
  95. package/dist/pipeline/agent_playout.js.map +1 -1
  96. package/dist/pipeline/human_input.cjs +158 -0
  97. package/dist/pipeline/human_input.cjs.map +1 -0
  98. package/dist/pipeline/human_input.js +124 -125
  99. package/dist/pipeline/human_input.js.map +1 -1
  100. package/dist/pipeline/index.cjs +31 -0
  101. package/dist/pipeline/index.cjs.map +1 -0
  102. package/dist/pipeline/index.js +8 -4
  103. package/dist/pipeline/index.js.map +1 -1
  104. package/dist/pipeline/pipeline_agent.cjs +642 -0
  105. package/dist/pipeline/pipeline_agent.cjs.map +1 -0
  106. package/dist/pipeline/pipeline_agent.js +595 -651
  107. package/dist/pipeline/pipeline_agent.js.map +1 -1
  108. package/dist/pipeline/speech_handle.cjs +128 -0
  109. package/dist/pipeline/speech_handle.cjs.map +1 -0
  110. package/dist/pipeline/speech_handle.js +102 -100
  111. package/dist/pipeline/speech_handle.js.map +1 -1
  112. package/dist/plugin.cjs +46 -0
  113. package/dist/plugin.cjs.map +1 -0
  114. package/dist/plugin.js +20 -20
  115. package/dist/plugin.js.map +1 -1
  116. package/dist/stt/index.cjs +38 -0
  117. package/dist/stt/index.cjs.map +1 -0
  118. package/dist/stt/index.js +13 -5
  119. package/dist/stt/index.js.map +1 -1
  120. package/dist/stt/stream_adapter.cjs +87 -0
  121. package/dist/stt/stream_adapter.cjs.map +1 -0
  122. package/dist/stt/stream_adapter.js +58 -55
  123. package/dist/stt/stream_adapter.js.map +1 -1
  124. package/dist/stt/stt.cjs +98 -0
  125. package/dist/stt/stt.cjs.map +1 -0
  126. package/dist/stt/stt.js +63 -98
  127. package/dist/stt/stt.js.map +1 -1
  128. package/dist/tokenize/basic/basic.cjs +98 -0
  129. package/dist/tokenize/basic/basic.cjs.map +1 -0
  130. package/dist/tokenize/basic/basic.d.ts +1 -1
  131. package/dist/tokenize/basic/basic.d.ts.map +1 -1
  132. package/dist/tokenize/basic/basic.js +56 -45
  133. package/dist/tokenize/basic/basic.js.map +1 -1
  134. package/dist/tokenize/basic/hyphenator.cjs +425 -0
  135. package/dist/tokenize/basic/hyphenator.cjs.map +1 -0
  136. package/dist/tokenize/basic/hyphenator.js +66 -82
  137. package/dist/tokenize/basic/hyphenator.js.map +1 -1
  138. package/dist/tokenize/basic/index.cjs +35 -0
  139. package/dist/tokenize/basic/index.cjs.map +1 -0
  140. package/dist/tokenize/basic/index.js +7 -4
  141. package/dist/tokenize/basic/index.js.map +1 -1
  142. package/dist/tokenize/basic/paragraph.cjs +57 -0
  143. package/dist/tokenize/basic/paragraph.cjs.map +1 -0
  144. package/dist/tokenize/basic/paragraph.js +30 -35
  145. package/dist/tokenize/basic/paragraph.js.map +1 -1
  146. package/dist/tokenize/basic/sentence.cjs +89 -0
  147. package/dist/tokenize/basic/sentence.cjs.map +1 -0
  148. package/dist/tokenize/basic/sentence.d.ts.map +1 -1
  149. package/dist/tokenize/basic/sentence.js +62 -57
  150. package/dist/tokenize/basic/sentence.js.map +1 -1
  151. package/dist/tokenize/basic/word.cjs +44 -0
  152. package/dist/tokenize/basic/word.cjs.map +1 -0
  153. package/dist/tokenize/basic/word.js +17 -20
  154. package/dist/tokenize/basic/word.js.map +1 -1
  155. package/dist/tokenize/index.cjs +55 -0
  156. package/dist/tokenize/index.cjs.map +1 -0
  157. package/dist/tokenize/index.js +18 -7
  158. package/dist/tokenize/index.js.map +1 -1
  159. package/dist/tokenize/token_stream.cjs +164 -0
  160. package/dist/tokenize/token_stream.cjs.map +1 -0
  161. package/dist/tokenize/token_stream.js +133 -139
  162. package/dist/tokenize/token_stream.js.map +1 -1
  163. package/dist/tokenize/tokenizer.cjs +184 -0
  164. package/dist/tokenize/tokenizer.cjs.map +1 -0
  165. package/dist/tokenize/tokenizer.js +138 -99
  166. package/dist/tokenize/tokenizer.js.map +1 -1
  167. package/dist/tokenize/tokenizer.test.cjs +220 -0
  168. package/dist/tokenize/tokenizer.test.cjs.map +1 -0
  169. package/dist/tokenize/tokenizer.test.d.ts +2 -0
  170. package/dist/tokenize/tokenizer.test.d.ts.map +1 -0
  171. package/dist/tokenize/tokenizer.test.js +219 -0
  172. package/dist/tokenize/tokenizer.test.js.map +1 -0
  173. package/dist/transcription.cjs +131 -0
  174. package/dist/transcription.cjs.map +1 -0
  175. package/dist/transcription.js +99 -96
  176. package/dist/transcription.js.map +1 -1
  177. package/dist/tts/index.cjs +38 -0
  178. package/dist/tts/index.cjs.map +1 -0
  179. package/dist/tts/index.js +13 -5
  180. package/dist/tts/index.js.map +1 -1
  181. package/dist/tts/stream_adapter.cjs +78 -0
  182. package/dist/tts/stream_adapter.cjs.map +1 -0
  183. package/dist/tts/stream_adapter.js +50 -47
  184. package/dist/tts/stream_adapter.js.map +1 -1
  185. package/dist/tts/tts.cjs +127 -0
  186. package/dist/tts/tts.cjs.map +1 -0
  187. package/dist/tts/tts.js +90 -120
  188. package/dist/tts/tts.js.map +1 -1
  189. package/dist/utils.cjs +284 -0
  190. package/dist/utils.cjs.map +1 -0
  191. package/dist/utils.js +242 -247
  192. package/dist/utils.js.map +1 -1
  193. package/dist/vad.cjs +92 -0
  194. package/dist/vad.cjs.map +1 -0
  195. package/dist/vad.js +57 -52
  196. package/dist/vad.js.map +1 -1
  197. package/dist/version.cjs +29 -0
  198. package/dist/version.cjs.map +1 -0
  199. package/dist/version.js +4 -4
  200. package/dist/version.js.map +1 -1
  201. package/dist/worker.cjs +577 -0
  202. package/dist/worker.cjs.map +1 -0
  203. package/dist/worker.d.ts +1 -1
  204. package/dist/worker.d.ts.map +1 -1
  205. package/dist/worker.js +512 -484
  206. package/dist/worker.js.map +1 -1
  207. package/package.json +18 -8
  208. package/src/ipc/job_main.ts +66 -64
  209. package/src/job.ts +3 -2
  210. package/src/pipeline/pipeline_agent.ts +23 -23
  211. package/src/tokenize/basic/basic.ts +1 -1
  212. package/src/tokenize/basic/sentence.ts +14 -8
  213. package/src/tokenize/tokenizer.test.ts +255 -0
  214. package/src/worker.ts +1 -0
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../../src/pipeline/human_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport { AudioStream, RoomEvent, TrackSource } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport { log } from '../log.js';\nimport type { STT, SpeechEvent } from '../stt/stt.js';\nimport { SpeechEventType } from '../stt/stt.js';\nimport { CancellablePromise, gracefullyCancel } from '../utils.js';\nimport type { VAD, VADEvent } from '../vad.js';\nimport { VADEventType } from '../vad.js';\n\nexport enum HumanInputEvent {\n START_OF_SPEECH,\n VAD_INFERENCE_DONE,\n END_OF_SPEECH,\n FINAL_TRANSCRIPT,\n INTERIM_TRANSCRIPT,\n}\n\nexport type HumanInputCallbacks = {\n [HumanInputEvent.START_OF_SPEECH]: (event: VADEvent) => void;\n [HumanInputEvent.VAD_INFERENCE_DONE]: (event: VADEvent) => void;\n [HumanInputEvent.END_OF_SPEECH]: (event: VADEvent) => void;\n [HumanInputEvent.FINAL_TRANSCRIPT]: (event: SpeechEvent) => void;\n [HumanInputEvent.INTERIM_TRANSCRIPT]: (event: SpeechEvent) => void;\n};\n\nexport class HumanInput extends (EventEmitter as new () => TypedEmitter<HumanInputCallbacks>) {\n #closed = false;\n #room: Room;\n #vad: VAD;\n #stt: STT;\n #participant: RemoteParticipant;\n #subscribedTrack?: RemoteAudioTrack;\n #recognizeTask?: CancellablePromise<void>;\n #speaking = false;\n #speechProbability = 0;\n #logger = log();\n\n constructor(room: Room, vad: VAD, stt: STT, participant: RemoteParticipant) {\n super();\n this.#room = room;\n this.#vad = vad;\n this.#stt = stt;\n this.#participant = participant;\n\n this.#room.on(RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));\n this.#room.on(RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));\n this.#subscribeToMicrophone();\n }\n\n #subscribeToMicrophone(): void {\n if (!this.#participant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.#participant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n\n const track = microphonePublication.track;\n if (track && track !== this.#subscribedTrack) {\n this.#subscribedTrack = track;\n if (this.#recognizeTask) {\n this.#recognizeTask.cancel();\n }\n\n const audioStream = new AudioStream(track, 16000);\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#recognizeTask = new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n const sttStream = this.#stt.stream();\n const vadStream = this.#vad.stream();\n\n const audioStreamCo = async () => {\n for await (const ev of audioStream) {\n if (cancelled) return;\n sttStream.pushFrame(ev);\n vadStream.pushFrame(ev);\n }\n };\n\n const vadStreamCo = async () => {\n for await (const ev of vadStream) {\n if (cancelled) return;\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.#speaking = true;\n this.emit(HumanInputEvent.START_OF_SPEECH, ev);\n break;\n case VADEventType.INFERENCE_DONE:\n this.#speechProbability = ev.probability;\n this.emit(HumanInputEvent.VAD_INFERENCE_DONE, ev);\n break;\n case VADEventType.END_OF_SPEECH:\n this.#speaking = false;\n this.emit(HumanInputEvent.END_OF_SPEECH, ev);\n break;\n }\n }\n };\n\n const sttStreamCo = async () => {\n for await (const ev of sttStream) {\n if (cancelled) return;\n if (ev.type === SpeechEventType.FINAL_TRANSCRIPT) {\n this.emit(HumanInputEvent.FINAL_TRANSCRIPT, ev);\n } else if (ev.type == SpeechEventType.INTERIM_TRANSCRIPT) {\n this.emit(HumanInputEvent.INTERIM_TRANSCRIPT, ev);\n }\n }\n };\n\n await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);\n sttStream.close();\n vadStream.close();\n resolve();\n });\n }\n }\n\n get speaking(): boolean {\n return this.#speaking;\n }\n\n get speakingProbability(): number {\n return this.#speechProbability;\n }\n\n async close() {\n if (this.#closed) {\n throw new Error('HumanInput already closed');\n }\n this.#closed = true;\n this.#room.removeAllListeners();\n this.#speaking = false;\n if (this.#recognizeTask) {\n await gracefullyCancel(this.#recognizeTask);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AASA,sBAAoD;AAEpD,yBAA6B;AAC7B,iBAAoB;AAEpB,iBAAgC;AAChC,mBAAqD;AAErD,iBAA6B;AAEtB,IAAK,kBAAL,kBAAKA,qBAAL;AACL,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AALU,SAAAA;AAAA,GAAA;AAgBL,MAAM,mBAAoB,gCAA6D;AAAA,EAC5F,UAAU;AAAA,EACV;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,cAAU,gBAAI;AAAA,EAEd,YAAY,MAAY,KAAU,KAAU,aAAgC;AAC1E,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,SAAK,MAAM,GAAG,0BAAU,gBAAgB,KAAK,uBAAuB,KAAK,IAAI,CAAC;AAC9E,SAAK,MAAM,GAAG,0BAAU,iBAAiB,KAAK,uBAAuB,KAAK,IAAI,CAAC;AAC/E,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,aAAa,kBAAkB,OAAO,GAAG;AACtE,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAEA,UAAM,QAAQ,sBAAsB;AACpC,QAAI,SAAS,UAAU,KAAK,kBAAkB;AAC5C,WAAK,mBAAmB;AACxB,UAAI,KAAK,gBAAgB;AACvB,aAAK,eAAe,OAAO;AAAA,MAC7B;AAEA,YAAM,cAAc,IAAI,4BAAY,OAAO,IAAK;AAGhD,WAAK,iBAAiB,IAAI,gCAAmB,OAAO,SAAS,GAAG,aAAa;AAC3E,YAAI,YAAY;AAChB,iBAAS,MAAM;AACb,sBAAY;AAAA,QACd,CAAC;AAED,cAAM,YAAY,KAAK,KAAK,OAAO;AACnC,cAAM,YAAY,KAAK,KAAK,OAAO;AAEnC,cAAM,gBAAgB,YAAY;AAChC,2BAAiB,MAAM,aAAa;AAClC,gBAAI,UAAW;AACf,sBAAU,UAAU,EAAE;AACtB,sBAAU,UAAU,EAAE;AAAA,UACxB;AAAA,QACF;AAEA,cAAM,cAAc,YAAY;AAC9B,2BAAiB,MAAM,WAAW;AAChC,gBAAI,UAAW;AACf,oBAAQ,GAAG,MAAM;AAAA,cACf,KAAK,wBAAa;AAChB,qBAAK,YAAY;AACjB,qBAAK,KAAK,yBAAiC,EAAE;AAC7C;AAAA,cACF,KAAK,wBAAa;AAChB,qBAAK,qBAAqB,GAAG;AAC7B,qBAAK,KAAK,4BAAoC,EAAE;AAChD;AAAA,cACF,KAAK,wBAAa;AAChB,qBAAK,YAAY;AACjB,qBAAK,KAAK,uBAA+B,EAAE;AAC3C;AAAA,YACJ;AAAA,UACF;AAAA,QACF;AAEA,cAAM,cAAc,YAAY;AAC9B,2BAAiB,MAAM,WAAW;AAChC,gBAAI,UAAW;AACf,gBAAI,GAAG,SAAS,2BAAgB,kBAAkB;AAChD,mBAAK,KAAK,0BAAkC,EAAE;AAAA,YAChD,WAAW,GAAG,QAAQ,2BAAgB,oBAAoB;AACxD,mBAAK,KAAK,4BAAoC,EAAE;AAAA,YAClD;AAAA,UACF;AAAA,QACF;AAEA,cAAM,QAAQ,IAAI,CAAC,cAAc,GAAG,YAAY,GAAG,YAAY,CAAC,CAAC;AACjE,kBAAU,MAAM;AAChB,kBAAU,MAAM;AAChB,gBAAQ;AAAA,MACV,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,2BAA2B;AAAA,IAC7C;AACA,SAAK,UAAU;AACf,SAAK,MAAM,mBAAmB;AAC9B,SAAK,YAAY;AACjB,QAAI,KAAK,gBAAgB;AACvB,gBAAM,+BAAiB,KAAK,cAAc;AAAA,IAC5C;AAAA,EACF;AACF;","names":["HumanInputEvent"]}
@@ -1,134 +1,133 @@
1
- import { AudioStream, RoomEvent, TrackSource } from '@livekit/rtc-node';
2
- import { EventEmitter } from 'node:events';
3
- import { log } from '../log.js';
4
- import { SpeechEventType } from '../stt/stt.js';
5
- import { CancellablePromise, gracefullyCancel } from '../utils.js';
6
- import { VADEventType } from '../vad.js';
7
- export var HumanInputEvent;
8
- (function (HumanInputEvent) {
9
- HumanInputEvent[HumanInputEvent["START_OF_SPEECH"] = 0] = "START_OF_SPEECH";
10
- HumanInputEvent[HumanInputEvent["VAD_INFERENCE_DONE"] = 1] = "VAD_INFERENCE_DONE";
11
- HumanInputEvent[HumanInputEvent["END_OF_SPEECH"] = 2] = "END_OF_SPEECH";
12
- HumanInputEvent[HumanInputEvent["FINAL_TRANSCRIPT"] = 3] = "FINAL_TRANSCRIPT";
13
- HumanInputEvent[HumanInputEvent["INTERIM_TRANSCRIPT"] = 4] = "INTERIM_TRANSCRIPT";
14
- })(HumanInputEvent || (HumanInputEvent = {}));
15
- export class HumanInput extends EventEmitter {
16
- #closed = false;
17
- #room;
18
- #vad;
19
- #stt;
20
- #participant;
21
- #subscribedTrack;
22
- #recognizeTask;
23
- #speaking = false;
24
- #speechProbability = 0;
25
- #logger = log();
26
- constructor(room, vad, stt, participant) {
27
- super();
28
- this.#room = room;
29
- this.#vad = vad;
30
- this.#stt = stt;
31
- this.#participant = participant;
32
- this.#room.on(RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));
33
- this.#room.on(RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));
34
- this.#subscribeToMicrophone();
1
+ import { AudioStream, RoomEvent, TrackSource } from "@livekit/rtc-node";
2
+ import { EventEmitter } from "node:events";
3
+ import { log } from "../log.js";
4
+ import { SpeechEventType } from "../stt/stt.js";
5
+ import { CancellablePromise, gracefullyCancel } from "../utils.js";
6
+ import { VADEventType } from "../vad.js";
7
+ var HumanInputEvent = /* @__PURE__ */ ((HumanInputEvent2) => {
8
+ HumanInputEvent2[HumanInputEvent2["START_OF_SPEECH"] = 0] = "START_OF_SPEECH";
9
+ HumanInputEvent2[HumanInputEvent2["VAD_INFERENCE_DONE"] = 1] = "VAD_INFERENCE_DONE";
10
+ HumanInputEvent2[HumanInputEvent2["END_OF_SPEECH"] = 2] = "END_OF_SPEECH";
11
+ HumanInputEvent2[HumanInputEvent2["FINAL_TRANSCRIPT"] = 3] = "FINAL_TRANSCRIPT";
12
+ HumanInputEvent2[HumanInputEvent2["INTERIM_TRANSCRIPT"] = 4] = "INTERIM_TRANSCRIPT";
13
+ return HumanInputEvent2;
14
+ })(HumanInputEvent || {});
15
+ class HumanInput extends EventEmitter {
16
+ #closed = false;
17
+ #room;
18
+ #vad;
19
+ #stt;
20
+ #participant;
21
+ #subscribedTrack;
22
+ #recognizeTask;
23
+ #speaking = false;
24
+ #speechProbability = 0;
25
+ #logger = log();
26
+ constructor(room, vad, stt, participant) {
27
+ super();
28
+ this.#room = room;
29
+ this.#vad = vad;
30
+ this.#stt = stt;
31
+ this.#participant = participant;
32
+ this.#room.on(RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));
33
+ this.#room.on(RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));
34
+ this.#subscribeToMicrophone();
35
+ }
36
+ #subscribeToMicrophone() {
37
+ if (!this.#participant) {
38
+ this.#logger.error("Participant is not set");
39
+ return;
35
40
  }
36
- #subscribeToMicrophone() {
37
- if (!this.#participant) {
38
- this.#logger.error('Participant is not set');
39
- return;
40
- }
41
- let microphonePublication = undefined;
42
- for (const publication of this.#participant.trackPublications.values()) {
43
- if (publication.source === TrackSource.SOURCE_MICROPHONE) {
44
- microphonePublication = publication;
41
+ let microphonePublication = void 0;
42
+ for (const publication of this.#participant.trackPublications.values()) {
43
+ if (publication.source === TrackSource.SOURCE_MICROPHONE) {
44
+ microphonePublication = publication;
45
+ break;
46
+ }
47
+ }
48
+ if (!microphonePublication) {
49
+ return;
50
+ }
51
+ if (!microphonePublication.subscribed) {
52
+ microphonePublication.setSubscribed(true);
53
+ }
54
+ const track = microphonePublication.track;
55
+ if (track && track !== this.#subscribedTrack) {
56
+ this.#subscribedTrack = track;
57
+ if (this.#recognizeTask) {
58
+ this.#recognizeTask.cancel();
59
+ }
60
+ const audioStream = new AudioStream(track, 16e3);
61
+ this.#recognizeTask = new CancellablePromise(async (resolve, _, onCancel) => {
62
+ let cancelled = false;
63
+ onCancel(() => {
64
+ cancelled = true;
65
+ });
66
+ const sttStream = this.#stt.stream();
67
+ const vadStream = this.#vad.stream();
68
+ const audioStreamCo = async () => {
69
+ for await (const ev of audioStream) {
70
+ if (cancelled) return;
71
+ sttStream.pushFrame(ev);
72
+ vadStream.pushFrame(ev);
73
+ }
74
+ };
75
+ const vadStreamCo = async () => {
76
+ for await (const ev of vadStream) {
77
+ if (cancelled) return;
78
+ switch (ev.type) {
79
+ case VADEventType.START_OF_SPEECH:
80
+ this.#speaking = true;
81
+ this.emit(0 /* START_OF_SPEECH */, ev);
82
+ break;
83
+ case VADEventType.INFERENCE_DONE:
84
+ this.#speechProbability = ev.probability;
85
+ this.emit(1 /* VAD_INFERENCE_DONE */, ev);
86
+ break;
87
+ case VADEventType.END_OF_SPEECH:
88
+ this.#speaking = false;
89
+ this.emit(2 /* END_OF_SPEECH */, ev);
45
90
  break;
46
91
  }
47
- }
48
- if (!microphonePublication) {
49
- return;
50
- }
51
- if (!microphonePublication.subscribed) {
52
- microphonePublication.setSubscribed(true);
53
- }
54
- const track = microphonePublication.track;
55
- if (track && track !== this.#subscribedTrack) {
56
- this.#subscribedTrack = track;
57
- if (this.#recognizeTask) {
58
- this.#recognizeTask.cancel();
92
+ }
93
+ };
94
+ const sttStreamCo = async () => {
95
+ for await (const ev of sttStream) {
96
+ if (cancelled) return;
97
+ if (ev.type === SpeechEventType.FINAL_TRANSCRIPT) {
98
+ this.emit(3 /* FINAL_TRANSCRIPT */, ev);
99
+ } else if (ev.type == SpeechEventType.INTERIM_TRANSCRIPT) {
100
+ this.emit(4 /* INTERIM_TRANSCRIPT */, ev);
59
101
  }
60
- const audioStream = new AudioStream(track, 16000);
61
- // eslint-disable-next-line @typescript-eslint/no-unused-vars
62
- this.#recognizeTask = new CancellablePromise(async (resolve, _, onCancel) => {
63
- let cancelled = false;
64
- onCancel(() => {
65
- cancelled = true;
66
- });
67
- const sttStream = this.#stt.stream();
68
- const vadStream = this.#vad.stream();
69
- const audioStreamCo = async () => {
70
- for await (const ev of audioStream) {
71
- if (cancelled)
72
- return;
73
- sttStream.pushFrame(ev);
74
- vadStream.pushFrame(ev);
75
- }
76
- };
77
- const vadStreamCo = async () => {
78
- for await (const ev of vadStream) {
79
- if (cancelled)
80
- return;
81
- switch (ev.type) {
82
- case VADEventType.START_OF_SPEECH:
83
- this.#speaking = true;
84
- this.emit(HumanInputEvent.START_OF_SPEECH, ev);
85
- break;
86
- case VADEventType.INFERENCE_DONE:
87
- this.#speechProbability = ev.probability;
88
- this.emit(HumanInputEvent.VAD_INFERENCE_DONE, ev);
89
- break;
90
- case VADEventType.END_OF_SPEECH:
91
- this.#speaking = false;
92
- this.emit(HumanInputEvent.END_OF_SPEECH, ev);
93
- break;
94
- }
95
- }
96
- };
97
- const sttStreamCo = async () => {
98
- for await (const ev of sttStream) {
99
- if (cancelled)
100
- return;
101
- if (ev.type === SpeechEventType.FINAL_TRANSCRIPT) {
102
- this.emit(HumanInputEvent.FINAL_TRANSCRIPT, ev);
103
- }
104
- else if (ev.type == SpeechEventType.INTERIM_TRANSCRIPT) {
105
- this.emit(HumanInputEvent.INTERIM_TRANSCRIPT, ev);
106
- }
107
- }
108
- };
109
- await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);
110
- sttStream.close();
111
- vadStream.close();
112
- resolve();
113
- });
114
- }
115
- }
116
- get speaking() {
117
- return this.#speaking;
102
+ }
103
+ };
104
+ await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);
105
+ sttStream.close();
106
+ vadStream.close();
107
+ resolve();
108
+ });
118
109
  }
119
- get speakingProbability() {
120
- return this.#speechProbability;
110
+ }
111
+ get speaking() {
112
+ return this.#speaking;
113
+ }
114
+ get speakingProbability() {
115
+ return this.#speechProbability;
116
+ }
117
+ async close() {
118
+ if (this.#closed) {
119
+ throw new Error("HumanInput already closed");
121
120
  }
122
- async close() {
123
- if (this.#closed) {
124
- throw new Error('HumanInput already closed');
125
- }
126
- this.#closed = true;
127
- this.#room.removeAllListeners();
128
- this.#speaking = false;
129
- if (this.#recognizeTask) {
130
- await gracefullyCancel(this.#recognizeTask);
131
- }
121
+ this.#closed = true;
122
+ this.#room.removeAllListeners();
123
+ this.#speaking = false;
124
+ if (this.#recognizeTask) {
125
+ await gracefullyCancel(this.#recognizeTask);
132
126
  }
127
+ }
133
128
  }
129
+ export {
130
+ HumanInput,
131
+ HumanInputEvent
132
+ };
134
133
  //# sourceMappingURL=human_input.js.map
@@ -1 +1 @@
1
- {"version":3,"file":"human_input.js","sourceRoot":"","sources":["../../src/pipeline/human_input.ts"],"names":[],"mappings":"AASA,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAExE,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AAEhC,OAAO,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AAChD,OAAO,EAAE,kBAAkB,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAEnE,OAAO,EAAE,YAAY,EAAE,MAAM,WAAW,CAAC;AAEzC,MAAM,CAAN,IAAY,eAMX;AAND,WAAY,eAAe;IACzB,2EAAe,CAAA;IACf,iFAAkB,CAAA;IAClB,uEAAa,CAAA;IACb,6EAAgB,CAAA;IAChB,iFAAkB,CAAA;AACpB,CAAC,EANW,eAAe,KAAf,eAAe,QAM1B;AAUD,MAAM,OAAO,UAAW,SAAS,YAA4D;IAC3F,OAAO,GAAG,KAAK,CAAC;IAChB,KAAK,CAAO;IACZ,IAAI,CAAM;IACV,IAAI,CAAM;IACV,YAAY,CAAoB;IAChC,gBAAgB,CAAoB;IACpC,cAAc,CAA4B;IAC1C,SAAS,GAAG,KAAK,CAAC;IAClB,kBAAkB,GAAG,CAAC,CAAC;IACvB,OAAO,GAAG,GAAG,EAAE,CAAC;IAEhB,YAAY,IAAU,EAAE,GAAQ,EAAE,GAAQ,EAAE,WAA8B;QACxE,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QAClB,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC;QAChB,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC;QAChB,IAAI,CAAC,YAAY,GAAG,WAAW,CAAC;QAEhC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,SAAS,CAAC,cAAc,EAAE,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC;QAChF,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,SAAS,CAAC,eAAe,EAAE,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC;QACjF,IAAI,CAAC,sBAAsB,EAAE,CAAC;IAChC,CAAC;IAED,sBAAsB;QACpB,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC;YACvB,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,wBAAwB,CAAC,CAAC;YAC7C,OAAO;QACT,CAAC;QAED,IAAI,qBAAqB,GAAuC,SAAS,CAAC;QAC1E,KAAK,MAAM,WAAW,IAAI,IAAI,CAAC,YAAY,CAAC,iBAAiB,CAAC,MAAM,EAAE,EAAE,CAAC;YACvE,IAAI,WAAW,CAAC,MAAM,KAAK,WAAW,CAAC,iBAAiB,EAAE,CAAC;gBACzD,qBAAqB,GAAG,WAAW,CAAC;gBACpC,MAAM;YACR,CAAC;QACH,CAAC;QACD,IAAI,CAAC,qBAAqB,EAAE,CAAC;YAC3B,OAAO;QACT,CAAC;QAED,IAAI,CAAC,qBAAqB,CAAC,UAAU,EAAE,CAAC;YACtC,qBAAqB,CAAC,aAAa,CAAC,IAAI,CAAC,CAAC;QAC5C,CAAC;QAED,MAAM,KAAK,GAAG,qBAAqB,CAAC,KAAK,CAAC;QAC1C,IAAI,KAAK,IAAI,KAAK,KAAK,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC7C,IAAI,CAAC,gBAAgB,GAAG,KAAK,CAAC;YAC9B,IAAI,IAAI,CAAC,cAAc,EAAE,CAAC;gBACxB,IAAI,CAAC,cAAc,CAAC,MAAM,EAAE,CAAC;YAC/B,CAAC;YAED,MAAM,WAAW,GAAG,IAAI,WAAW,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;YAElD,6DAA6D;YAC7D,IAAI,CAAC,cAAc,GAAG,IAAI,kBAAkB,CAAC,KAAK,EAAE,OAAO,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE;gBAC1E,IAAI,SAAS,GAAG,KAAK,CAAC;gBACtB,QAAQ,CAAC,GAAG,EAAE;oBACZ,SAAS,GAAG,IAAI,CAAC;gBACnB,CAAC,CAAC,CAAC;gBAEH,MAAM,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;gBACrC,MAAM,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;gBAErC,MAAM,aAAa,GAAG,KAAK,IAAI,EAAE;oBAC/B,IAAI,KAAK,EAAE,MAAM,EAAE,IAAI,WAAW,EAAE,CAAC;wBACnC,IAAI,SAAS;4BAAE,OAAO;wBACtB,SAAS,CAAC,SAAS,CAAC,EAAE,CAAC,CAAC;wBACxB,SAAS,CAAC,SAAS,CAAC,EAAE,CAAC,CAAC;oBAC1B,CAAC;gBACH,CAAC,CAAC;gBAEF,MAAM,WAAW,GAAG,KAAK,IAAI,EAAE;oBAC7B,IAAI,KAAK,EAAE,MAAM,EAAE,IAAI,SAAS,EAAE,CAAC;wBACjC,IAAI,SAAS;4BAAE,OAAO;wBACtB,QAAQ,EAAE,CAAC,IAAI,EAAE,CAAC;4BAChB,KAAK,YAAY,CAAC,eAAe;gCAC/B,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;gCACtB,IAAI,CAAC,IAAI,CAAC,eAAe,CAAC,eAAe,EAAE,EAAE,CAAC,CAAC;gCAC/C,MAAM;4BACR,KAAK,YAAY,CAAC,cAAc;gCAC9B,IAAI,CAAC,kBAAkB,GAAG,EAAE,CAAC,WAAW,CAAC;gCACzC,IAAI,CAAC,IAAI,CAAC,eAAe,CAAC,kBAAkB,EAAE,EAAE,CAAC,CAAC;gCAClD,MAAM;4BACR,KAAK,YAAY,CAAC,aAAa;gCAC7B,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;gCACvB,IAAI,CAAC,IAAI,CAAC,eAAe,CAAC,aAAa,EAAE,EAAE,CAAC,CAAC;gCAC7C,MAAM;wBACV,CAAC;oBACH,CAAC;gBACH,CAAC,CAAC;gBAEF,MAAM,WAAW,GAAG,KAAK,IAAI,EAAE;oBAC7B,IAAI,KAAK,EAAE,MAAM,EAAE,IAAI,SAAS,EAAE,CAAC;wBACjC,IAAI,SAAS;4BAAE,OAAO;wBACtB,IAAI,EAAE,CAAC,IAAI,KAAK,eAAe,CAAC,gBAAgB,EAAE,CAAC;4BACjD,IAAI,CAAC,IAAI,CAAC,eAAe,CAAC,gBAAgB,EAAE,EAAE,CAAC,CAAC;wBAClD,CAAC;6BAAM,IAAI,EAAE,CAAC,IAAI,IAAI,eAAe,CAAC,kBAAkB,EAAE,CAAC;4BACzD,IAAI,CAAC,IAAI,CAAC,eAAe,CAAC,kBAAkB,EAAE,EAAE,CAAC,CAAC;wBACpD,CAAC;oBACH,CAAC;gBACH,CAAC,CAAC;gBAEF,MAAM,OAAO,CAAC,GAAG,CAAC,CAAC,aAAa,EAAE,EAAE,WAAW,EAAE,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC;gBACnE,SAAS,CAAC,KAAK,EAAE,CAAC;gBAClB,SAAS,CAAC,KAAK,EAAE,CAAC;gBAClB,OAAO,EAAE,CAAC;YACZ,CAAC,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IAED,IAAI,QAAQ;QACV,OAAO,IAAI,CAAC,SAAS,CAAC;IACxB,CAAC;IAED,IAAI,mBAAmB;QACrB,OAAO,IAAI,CAAC,kBAAkB,CAAC;IACjC,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC;YACjB,MAAM,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC;QAC/C,CAAC;QACD,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;QACpB,IAAI,CAAC,KAAK,CAAC,kBAAkB,EAAE,CAAC;QAChC,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;QACvB,IAAI,IAAI,CAAC,cAAc,EAAE,CAAC;YACxB,MAAM,gBAAgB,CAAC,IAAI,CAAC,cAAc,CAAC,CAAC;QAC9C,CAAC;IACH,CAAC;CACF"}
1
+ {"version":3,"sources":["../../src/pipeline/human_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport { AudioStream, RoomEvent, TrackSource } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport { log } from '../log.js';\nimport type { STT, SpeechEvent } from '../stt/stt.js';\nimport { SpeechEventType } from '../stt/stt.js';\nimport { CancellablePromise, gracefullyCancel } from '../utils.js';\nimport type { VAD, VADEvent } from '../vad.js';\nimport { VADEventType } from '../vad.js';\n\nexport enum HumanInputEvent {\n START_OF_SPEECH,\n VAD_INFERENCE_DONE,\n END_OF_SPEECH,\n FINAL_TRANSCRIPT,\n INTERIM_TRANSCRIPT,\n}\n\nexport type HumanInputCallbacks = {\n [HumanInputEvent.START_OF_SPEECH]: (event: VADEvent) => void;\n [HumanInputEvent.VAD_INFERENCE_DONE]: (event: VADEvent) => void;\n [HumanInputEvent.END_OF_SPEECH]: (event: VADEvent) => void;\n [HumanInputEvent.FINAL_TRANSCRIPT]: (event: SpeechEvent) => void;\n [HumanInputEvent.INTERIM_TRANSCRIPT]: (event: SpeechEvent) => void;\n};\n\nexport class HumanInput extends (EventEmitter as new () => TypedEmitter<HumanInputCallbacks>) {\n #closed = false;\n #room: Room;\n #vad: VAD;\n #stt: STT;\n #participant: RemoteParticipant;\n #subscribedTrack?: RemoteAudioTrack;\n #recognizeTask?: CancellablePromise<void>;\n #speaking = false;\n #speechProbability = 0;\n #logger = log();\n\n constructor(room: Room, vad: VAD, stt: STT, participant: RemoteParticipant) {\n super();\n this.#room = room;\n this.#vad = vad;\n this.#stt = stt;\n this.#participant = participant;\n\n this.#room.on(RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));\n this.#room.on(RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));\n this.#subscribeToMicrophone();\n }\n\n #subscribeToMicrophone(): void {\n if (!this.#participant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.#participant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n\n const track = microphonePublication.track;\n if (track && track !== this.#subscribedTrack) {\n this.#subscribedTrack = track;\n if (this.#recognizeTask) {\n this.#recognizeTask.cancel();\n }\n\n const audioStream = new AudioStream(track, 16000);\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#recognizeTask = new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n const sttStream = this.#stt.stream();\n const vadStream = this.#vad.stream();\n\n const audioStreamCo = async () => {\n for await (const ev of audioStream) {\n if (cancelled) return;\n sttStream.pushFrame(ev);\n vadStream.pushFrame(ev);\n }\n };\n\n const vadStreamCo = async () => {\n for await (const ev of vadStream) {\n if (cancelled) return;\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.#speaking = true;\n this.emit(HumanInputEvent.START_OF_SPEECH, ev);\n break;\n case VADEventType.INFERENCE_DONE:\n this.#speechProbability = ev.probability;\n this.emit(HumanInputEvent.VAD_INFERENCE_DONE, ev);\n break;\n case VADEventType.END_OF_SPEECH:\n this.#speaking = false;\n this.emit(HumanInputEvent.END_OF_SPEECH, ev);\n break;\n }\n }\n };\n\n const sttStreamCo = async () => {\n for await (const ev of sttStream) {\n if (cancelled) return;\n if (ev.type === SpeechEventType.FINAL_TRANSCRIPT) {\n this.emit(HumanInputEvent.FINAL_TRANSCRIPT, ev);\n } else if (ev.type == SpeechEventType.INTERIM_TRANSCRIPT) {\n this.emit(HumanInputEvent.INTERIM_TRANSCRIPT, ev);\n }\n }\n };\n\n await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);\n sttStream.close();\n vadStream.close();\n resolve();\n });\n }\n }\n\n get speaking(): boolean {\n return this.#speaking;\n }\n\n get speakingProbability(): number {\n return this.#speechProbability;\n }\n\n async close() {\n if (this.#closed) {\n throw new Error('HumanInput already closed');\n }\n this.#closed = true;\n this.#room.removeAllListeners();\n this.#speaking = false;\n if (this.#recognizeTask) {\n await gracefullyCancel(this.#recognizeTask);\n }\n }\n}\n"],"mappings":"AASA,SAAS,aAAa,WAAW,mBAAmB;AAEpD,SAAS,oBAAoB;AAC7B,SAAS,WAAW;AAEpB,SAAS,uBAAuB;AAChC,SAAS,oBAAoB,wBAAwB;AAErD,SAAS,oBAAoB;AAEtB,IAAK,kBAAL,kBAAKA,qBAAL;AACL,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AALU,SAAAA;AAAA,GAAA;AAgBL,MAAM,mBAAoB,aAA6D;AAAA,EAC5F,UAAU;AAAA,EACV;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,UAAU,IAAI;AAAA,EAEd,YAAY,MAAY,KAAU,KAAU,aAAgC;AAC1E,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,SAAK,MAAM,GAAG,UAAU,gBAAgB,KAAK,uBAAuB,KAAK,IAAI,CAAC;AAC9E,SAAK,MAAM,GAAG,UAAU,iBAAiB,KAAK,uBAAuB,KAAK,IAAI,CAAC;AAC/E,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,aAAa,kBAAkB,OAAO,GAAG;AACtE,UAAI,YAAY,WAAW,YAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAEA,UAAM,QAAQ,sBAAsB;AACpC,QAAI,SAAS,UAAU,KAAK,kBAAkB;AAC5C,WAAK,mBAAmB;AACxB,UAAI,KAAK,gBAAgB;AACvB,aAAK,eAAe,OAAO;AAAA,MAC7B;AAEA,YAAM,cAAc,IAAI,YAAY,OAAO,IAAK;AAGhD,WAAK,iBAAiB,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC3E,YAAI,YAAY;AAChB,iBAAS,MAAM;AACb,sBAAY;AAAA,QACd,CAAC;AAED,cAAM,YAAY,KAAK,KAAK,OAAO;AACnC,cAAM,YAAY,KAAK,KAAK,OAAO;AAEnC,cAAM,gBAAgB,YAAY;AAChC,2BAAiB,MAAM,aAAa;AAClC,gBAAI,UAAW;AACf,sBAAU,UAAU,EAAE;AACtB,sBAAU,UAAU,EAAE;AAAA,UACxB;AAAA,QACF;AAEA,cAAM,cAAc,YAAY;AAC9B,2BAAiB,MAAM,WAAW;AAChC,gBAAI,UAAW;AACf,oBAAQ,GAAG,MAAM;AAAA,cACf,KAAK,aAAa;AAChB,qBAAK,YAAY;AACjB,qBAAK,KAAK,yBAAiC,EAAE;AAC7C;AAAA,cACF,KAAK,aAAa;AAChB,qBAAK,qBAAqB,GAAG;AAC7B,qBAAK,KAAK,4BAAoC,EAAE;AAChD;AAAA,cACF,KAAK,aAAa;AAChB,qBAAK,YAAY;AACjB,qBAAK,KAAK,uBAA+B,EAAE;AAC3C;AAAA,YACJ;AAAA,UACF;AAAA,QACF;AAEA,cAAM,cAAc,YAAY;AAC9B,2BAAiB,MAAM,WAAW;AAChC,gBAAI,UAAW;AACf,gBAAI,GAAG,SAAS,gBAAgB,kBAAkB;AAChD,mBAAK,KAAK,0BAAkC,EAAE;AAAA,YAChD,WAAW,GAAG,QAAQ,gBAAgB,oBAAoB;AACxD,mBAAK,KAAK,4BAAoC,EAAE;AAAA,YAClD;AAAA,UACF;AAAA,QACF;AAEA,cAAM,QAAQ,IAAI,CAAC,cAAc,GAAG,YAAY,GAAG,YAAY,CAAC,CAAC;AACjE,kBAAU,MAAM;AAChB,kBAAU,MAAM;AAChB,gBAAQ;AAAA,MACV,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,2BAA2B;AAAA,IAC7C;AACA,SAAK,UAAU;AACf,SAAK,MAAM,mBAAmB;AAC9B,SAAK,YAAY;AACjB,QAAI,KAAK,gBAAgB;AACvB,YAAM,iBAAiB,KAAK,cAAc;AAAA,IAC5C;AAAA,EACF;AACF;","names":["HumanInputEvent"]}
@@ -0,0 +1,31 @@
1
+ "use strict";
2
+ var __defProp = Object.defineProperty;
3
+ var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
4
+ var __getOwnPropNames = Object.getOwnPropertyNames;
5
+ var __hasOwnProp = Object.prototype.hasOwnProperty;
6
+ var __export = (target, all) => {
7
+ for (var name in all)
8
+ __defProp(target, name, { get: all[name], enumerable: true });
9
+ };
10
+ var __copyProps = (to, from, except, desc) => {
11
+ if (from && typeof from === "object" || typeof from === "function") {
12
+ for (let key of __getOwnPropNames(from))
13
+ if (!__hasOwnProp.call(to, key) && key !== except)
14
+ __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
15
+ }
16
+ return to;
17
+ };
18
+ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
19
+ var pipeline_exports = {};
20
+ __export(pipeline_exports, {
21
+ VPAEvent: () => import_pipeline_agent.VPAEvent,
22
+ VoicePipelineAgent: () => import_pipeline_agent.VoicePipelineAgent
23
+ });
24
+ module.exports = __toCommonJS(pipeline_exports);
25
+ var import_pipeline_agent = require("./pipeline_agent.cjs");
26
+ // Annotate the CommonJS export names for ESM import in node:
27
+ 0 && (module.exports = {
28
+ VPAEvent,
29
+ VoicePipelineAgent
30
+ });
31
+ //# sourceMappingURL=index.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../../src/pipeline/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport {\n type AgentState,\n type BeforeTTSCallback,\n type BeforeLLMCallback,\n type VPACallbacks,\n type AgentCallContext,\n type AgentTranscriptionOptions,\n type VPAOptions,\n VPAEvent,\n VoicePipelineAgent,\n} from './pipeline_agent.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,4BAUO;","names":[]}
@@ -1,5 +1,9 @@
1
- // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
2
- //
3
- // SPDX-License-Identifier: Apache-2.0
4
- export { VPAEvent, VoicePipelineAgent, } from './pipeline_agent.js';
1
+ import {
2
+ VPAEvent,
3
+ VoicePipelineAgent
4
+ } from "./pipeline_agent.js";
5
+ export {
6
+ VPAEvent,
7
+ VoicePipelineAgent
8
+ };
5
9
  //# sourceMappingURL=index.js.map
@@ -1 +1 @@
1
- {"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/pipeline/index.ts"],"names":[],"mappings":"AAAA,6CAA6C;AAC7C,EAAE;AACF,sCAAsC;AAEtC,OAAO,EAQL,QAAQ,EACR,kBAAkB,GACnB,MAAM,qBAAqB,CAAC"}
1
+ {"version":3,"sources":["../../src/pipeline/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport {\n type AgentState,\n type BeforeTTSCallback,\n type BeforeLLMCallback,\n type VPACallbacks,\n type AgentCallContext,\n type AgentTranscriptionOptions,\n type VPAOptions,\n VPAEvent,\n VoicePipelineAgent,\n} from './pipeline_agent.js';\n"],"mappings":"AAIA;AAAA,EAQE;AAAA,EACA;AAAA,OACK;","names":[]}