@livekit/agents 0.7.8 → 1.0.0-next.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (627) hide show
  1. package/dist/_exceptions.cjs +109 -0
  2. package/dist/_exceptions.cjs.map +1 -0
  3. package/dist/_exceptions.d.cts +64 -0
  4. package/dist/_exceptions.d.ts +64 -0
  5. package/dist/_exceptions.d.ts.map +1 -0
  6. package/dist/_exceptions.js +80 -0
  7. package/dist/_exceptions.js.map +1 -0
  8. package/dist/audio.cjs +10 -3
  9. package/dist/audio.cjs.map +1 -1
  10. package/dist/audio.d.cts +2 -0
  11. package/dist/audio.d.ts +2 -0
  12. package/dist/audio.d.ts.map +1 -1
  13. package/dist/audio.js +8 -2
  14. package/dist/audio.js.map +1 -1
  15. package/dist/cli.cjs +25 -0
  16. package/dist/cli.cjs.map +1 -1
  17. package/dist/cli.d.ts.map +1 -1
  18. package/dist/cli.js +25 -0
  19. package/dist/cli.js.map +1 -1
  20. package/dist/constants.cjs +6 -0
  21. package/dist/constants.cjs.map +1 -1
  22. package/dist/constants.d.cts +2 -0
  23. package/dist/constants.d.ts +2 -0
  24. package/dist/constants.d.ts.map +1 -1
  25. package/dist/constants.js +4 -0
  26. package/dist/constants.js.map +1 -1
  27. package/dist/http_server.cjs.map +1 -1
  28. package/dist/http_server.d.cts +1 -0
  29. package/dist/http_server.d.ts +1 -0
  30. package/dist/http_server.d.ts.map +1 -1
  31. package/dist/http_server.js.map +1 -1
  32. package/dist/index.cjs +27 -20
  33. package/dist/index.cjs.map +1 -1
  34. package/dist/index.d.cts +13 -10
  35. package/dist/index.d.ts +13 -10
  36. package/dist/index.d.ts.map +1 -1
  37. package/dist/index.js +15 -11
  38. package/dist/index.js.map +1 -1
  39. package/dist/inference_runner.cjs +0 -1
  40. package/dist/inference_runner.cjs.map +1 -1
  41. package/dist/inference_runner.d.cts +2 -3
  42. package/dist/inference_runner.d.ts +2 -3
  43. package/dist/inference_runner.d.ts.map +1 -1
  44. package/dist/inference_runner.js +0 -1
  45. package/dist/inference_runner.js.map +1 -1
  46. package/dist/ipc/inference_proc_executor.cjs +2 -2
  47. package/dist/ipc/inference_proc_executor.cjs.map +1 -1
  48. package/dist/ipc/inference_proc_executor.js +2 -2
  49. package/dist/ipc/inference_proc_executor.js.map +1 -1
  50. package/dist/ipc/job_executor.cjs.map +1 -1
  51. package/dist/ipc/job_executor.js.map +1 -1
  52. package/dist/ipc/job_proc_executor.cjs +1 -0
  53. package/dist/ipc/job_proc_executor.cjs.map +1 -1
  54. package/dist/ipc/job_proc_executor.js +1 -0
  55. package/dist/ipc/job_proc_executor.js.map +1 -1
  56. package/dist/ipc/job_proc_lazy_main.cjs +1 -1
  57. package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
  58. package/dist/ipc/job_proc_lazy_main.js +1 -1
  59. package/dist/ipc/job_proc_lazy_main.js.map +1 -1
  60. package/dist/ipc/supervised_proc.d.cts +1 -1
  61. package/dist/ipc/supervised_proc.d.ts +1 -1
  62. package/dist/ipc/supervised_proc.d.ts.map +1 -1
  63. package/dist/job.cjs +14 -2
  64. package/dist/job.cjs.map +1 -1
  65. package/dist/job.d.cts +8 -0
  66. package/dist/job.d.ts +8 -0
  67. package/dist/job.d.ts.map +1 -1
  68. package/dist/job.js +12 -1
  69. package/dist/job.js.map +1 -1
  70. package/dist/llm/chat_context.cjs +332 -82
  71. package/dist/llm/chat_context.cjs.map +1 -1
  72. package/dist/llm/chat_context.d.cts +152 -48
  73. package/dist/llm/chat_context.d.ts +152 -48
  74. package/dist/llm/chat_context.d.ts.map +1 -1
  75. package/dist/llm/chat_context.js +327 -81
  76. package/dist/llm/chat_context.js.map +1 -1
  77. package/dist/llm/chat_context.test.cjs +380 -0
  78. package/dist/llm/chat_context.test.cjs.map +1 -0
  79. package/dist/llm/chat_context.test.js +385 -0
  80. package/dist/llm/chat_context.test.js.map +1 -0
  81. package/dist/llm/index.cjs +37 -8
  82. package/dist/llm/index.cjs.map +1 -1
  83. package/dist/llm/index.d.cts +7 -3
  84. package/dist/llm/index.d.ts +7 -3
  85. package/dist/llm/index.d.ts.map +1 -1
  86. package/dist/llm/index.js +39 -9
  87. package/dist/llm/index.js.map +1 -1
  88. package/dist/llm/llm.cjs +98 -33
  89. package/dist/llm/llm.cjs.map +1 -1
  90. package/dist/llm/llm.d.cts +50 -24
  91. package/dist/llm/llm.d.ts +50 -24
  92. package/dist/llm/llm.d.ts.map +1 -1
  93. package/dist/llm/llm.js +99 -33
  94. package/dist/llm/llm.js.map +1 -1
  95. package/dist/llm/provider_format/google.cjs +128 -0
  96. package/dist/llm/provider_format/google.cjs.map +1 -0
  97. package/dist/llm/provider_format/google.d.cts +6 -0
  98. package/dist/llm/provider_format/google.d.ts +6 -0
  99. package/dist/llm/provider_format/google.d.ts.map +1 -0
  100. package/dist/llm/provider_format/google.js +104 -0
  101. package/dist/llm/provider_format/google.js.map +1 -0
  102. package/dist/llm/provider_format/google.test.cjs +676 -0
  103. package/dist/llm/provider_format/google.test.cjs.map +1 -0
  104. package/dist/llm/provider_format/google.test.js +675 -0
  105. package/dist/llm/provider_format/google.test.js.map +1 -0
  106. package/dist/llm/provider_format/index.cjs +40 -0
  107. package/dist/llm/provider_format/index.cjs.map +1 -0
  108. package/dist/llm/provider_format/index.d.cts +4 -0
  109. package/dist/llm/provider_format/index.d.ts +4 -0
  110. package/dist/llm/provider_format/index.d.ts.map +1 -0
  111. package/dist/llm/provider_format/index.js +16 -0
  112. package/dist/llm/provider_format/index.js.map +1 -0
  113. package/dist/llm/provider_format/openai.cjs +116 -0
  114. package/dist/llm/provider_format/openai.cjs.map +1 -0
  115. package/dist/llm/provider_format/openai.d.cts +3 -0
  116. package/dist/llm/provider_format/openai.d.ts +3 -0
  117. package/dist/llm/provider_format/openai.d.ts.map +1 -0
  118. package/dist/llm/provider_format/openai.js +92 -0
  119. package/dist/llm/provider_format/openai.js.map +1 -0
  120. package/dist/llm/provider_format/openai.test.cjs +490 -0
  121. package/dist/llm/provider_format/openai.test.cjs.map +1 -0
  122. package/dist/llm/provider_format/openai.test.js +489 -0
  123. package/dist/llm/provider_format/openai.test.js.map +1 -0
  124. package/dist/llm/provider_format/utils.cjs +146 -0
  125. package/dist/llm/provider_format/utils.cjs.map +1 -0
  126. package/dist/llm/provider_format/utils.d.cts +38 -0
  127. package/dist/llm/provider_format/utils.d.ts +38 -0
  128. package/dist/llm/provider_format/utils.d.ts.map +1 -0
  129. package/dist/llm/provider_format/utils.js +122 -0
  130. package/dist/llm/provider_format/utils.js.map +1 -0
  131. package/dist/llm/realtime.cjs +77 -0
  132. package/dist/llm/realtime.cjs.map +1 -0
  133. package/dist/llm/realtime.d.cts +98 -0
  134. package/dist/llm/realtime.d.ts +98 -0
  135. package/dist/llm/realtime.d.ts.map +1 -0
  136. package/dist/llm/realtime.js +52 -0
  137. package/dist/llm/realtime.js.map +1 -0
  138. package/dist/llm/remote_chat_context.cjs +112 -0
  139. package/dist/llm/remote_chat_context.cjs.map +1 -0
  140. package/dist/llm/remote_chat_context.d.cts +23 -0
  141. package/dist/llm/remote_chat_context.d.ts +23 -0
  142. package/dist/llm/remote_chat_context.d.ts.map +1 -0
  143. package/dist/llm/remote_chat_context.js +88 -0
  144. package/dist/llm/remote_chat_context.js.map +1 -0
  145. package/dist/llm/remote_chat_context.test.cjs +225 -0
  146. package/dist/llm/remote_chat_context.test.cjs.map +1 -0
  147. package/dist/llm/remote_chat_context.test.js +224 -0
  148. package/dist/llm/remote_chat_context.test.js.map +1 -0
  149. package/dist/llm/tool_context.cjs +111 -0
  150. package/dist/llm/tool_context.cjs.map +1 -0
  151. package/dist/llm/tool_context.d.cts +125 -0
  152. package/dist/llm/tool_context.d.ts +125 -0
  153. package/dist/llm/tool_context.d.ts.map +1 -0
  154. package/dist/llm/tool_context.js +80 -0
  155. package/dist/llm/tool_context.js.map +1 -0
  156. package/dist/llm/tool_context.test.cjs +162 -0
  157. package/dist/llm/tool_context.test.cjs.map +1 -0
  158. package/dist/llm/tool_context.test.js +161 -0
  159. package/dist/llm/tool_context.test.js.map +1 -0
  160. package/dist/llm/tool_context.type.test.cjs +92 -0
  161. package/dist/llm/tool_context.type.test.cjs.map +1 -0
  162. package/dist/llm/tool_context.type.test.js +91 -0
  163. package/dist/llm/tool_context.type.test.js.map +1 -0
  164. package/dist/llm/utils.cjs +260 -0
  165. package/dist/llm/utils.cjs.map +1 -0
  166. package/dist/llm/utils.d.cts +42 -0
  167. package/dist/llm/utils.d.ts +42 -0
  168. package/dist/llm/utils.d.ts.map +1 -0
  169. package/dist/llm/utils.js +223 -0
  170. package/dist/llm/utils.js.map +1 -0
  171. package/dist/llm/utils.test.cjs +513 -0
  172. package/dist/llm/utils.test.cjs.map +1 -0
  173. package/dist/llm/utils.test.js +490 -0
  174. package/dist/llm/utils.test.js.map +1 -0
  175. package/dist/metrics/base.cjs +0 -27
  176. package/dist/metrics/base.cjs.map +1 -1
  177. package/dist/metrics/base.d.cts +105 -63
  178. package/dist/metrics/base.d.ts +105 -63
  179. package/dist/metrics/base.d.ts.map +1 -1
  180. package/dist/metrics/base.js +0 -19
  181. package/dist/metrics/base.js.map +1 -1
  182. package/dist/metrics/index.cjs +0 -3
  183. package/dist/metrics/index.cjs.map +1 -1
  184. package/dist/metrics/index.d.cts +2 -3
  185. package/dist/metrics/index.d.ts +2 -3
  186. package/dist/metrics/index.d.ts.map +1 -1
  187. package/dist/metrics/index.js +0 -2
  188. package/dist/metrics/index.js.map +1 -1
  189. package/dist/metrics/usage_collector.cjs +17 -12
  190. package/dist/metrics/usage_collector.cjs.map +1 -1
  191. package/dist/metrics/usage_collector.d.cts +3 -2
  192. package/dist/metrics/usage_collector.d.ts +3 -2
  193. package/dist/metrics/usage_collector.d.ts.map +1 -1
  194. package/dist/metrics/usage_collector.js +17 -12
  195. package/dist/metrics/usage_collector.js.map +1 -1
  196. package/dist/metrics/utils.cjs +22 -59
  197. package/dist/metrics/utils.cjs.map +1 -1
  198. package/dist/metrics/utils.d.cts +1 -8
  199. package/dist/metrics/utils.d.ts +1 -8
  200. package/dist/metrics/utils.d.ts.map +1 -1
  201. package/dist/metrics/utils.js +22 -52
  202. package/dist/metrics/utils.js.map +1 -1
  203. package/dist/multimodal/index.cjs +0 -2
  204. package/dist/multimodal/index.cjs.map +1 -1
  205. package/dist/multimodal/index.d.cts +0 -1
  206. package/dist/multimodal/index.d.ts +0 -1
  207. package/dist/multimodal/index.d.ts.map +1 -1
  208. package/dist/multimodal/index.js +0 -1
  209. package/dist/multimodal/index.js.map +1 -1
  210. package/dist/plugin.cjs +24 -8
  211. package/dist/plugin.cjs.map +1 -1
  212. package/dist/plugin.d.cts +18 -4
  213. package/dist/plugin.d.ts +18 -4
  214. package/dist/plugin.d.ts.map +1 -1
  215. package/dist/plugin.js +22 -7
  216. package/dist/plugin.js.map +1 -1
  217. package/dist/stream/deferred_stream.cjs +98 -0
  218. package/dist/stream/deferred_stream.cjs.map +1 -0
  219. package/dist/stream/deferred_stream.d.cts +27 -0
  220. package/dist/stream/deferred_stream.d.ts +27 -0
  221. package/dist/stream/deferred_stream.d.ts.map +1 -0
  222. package/dist/stream/deferred_stream.js +73 -0
  223. package/dist/stream/deferred_stream.js.map +1 -0
  224. package/dist/stream/deferred_stream.test.cjs +527 -0
  225. package/dist/stream/deferred_stream.test.cjs.map +1 -0
  226. package/dist/stream/deferred_stream.test.js +526 -0
  227. package/dist/stream/deferred_stream.test.js.map +1 -0
  228. package/dist/stream/identity_transform.cjs +42 -0
  229. package/dist/stream/identity_transform.cjs.map +1 -0
  230. package/dist/stream/identity_transform.d.cts +6 -0
  231. package/dist/stream/identity_transform.d.ts +6 -0
  232. package/dist/stream/identity_transform.d.ts.map +1 -0
  233. package/dist/stream/identity_transform.js +18 -0
  234. package/dist/stream/identity_transform.js.map +1 -0
  235. package/dist/stream/identity_transform.test.cjs +125 -0
  236. package/dist/stream/identity_transform.test.cjs.map +1 -0
  237. package/dist/stream/identity_transform.test.js +124 -0
  238. package/dist/stream/identity_transform.test.js.map +1 -0
  239. package/dist/stream/index.cjs +38 -0
  240. package/dist/stream/index.cjs.map +1 -0
  241. package/dist/stream/index.d.cts +5 -0
  242. package/dist/stream/index.d.ts +5 -0
  243. package/dist/stream/index.d.ts.map +1 -0
  244. package/dist/stream/index.js +11 -0
  245. package/dist/stream/index.js.map +1 -0
  246. package/dist/stream/merge_readable_streams.cjs +59 -0
  247. package/dist/stream/merge_readable_streams.cjs.map +1 -0
  248. package/dist/stream/merge_readable_streams.d.cts +4 -0
  249. package/dist/stream/merge_readable_streams.d.ts +4 -0
  250. package/dist/stream/merge_readable_streams.d.ts.map +1 -0
  251. package/dist/stream/merge_readable_streams.js +35 -0
  252. package/dist/stream/merge_readable_streams.js.map +1 -0
  253. package/dist/stream/stream_channel.cjs +47 -0
  254. package/dist/stream/stream_channel.cjs.map +1 -0
  255. package/dist/stream/stream_channel.d.cts +9 -0
  256. package/dist/stream/stream_channel.d.ts +9 -0
  257. package/dist/stream/stream_channel.d.ts.map +1 -0
  258. package/dist/stream/stream_channel.js +23 -0
  259. package/dist/stream/stream_channel.js.map +1 -0
  260. package/dist/stream/stream_channel.test.cjs +97 -0
  261. package/dist/stream/stream_channel.test.cjs.map +1 -0
  262. package/dist/stream/stream_channel.test.js +96 -0
  263. package/dist/stream/stream_channel.test.js.map +1 -0
  264. package/dist/stt/stream_adapter.cjs +3 -4
  265. package/dist/stt/stream_adapter.cjs.map +1 -1
  266. package/dist/stt/stream_adapter.d.cts +1 -0
  267. package/dist/stt/stream_adapter.d.ts +1 -0
  268. package/dist/stt/stream_adapter.d.ts.map +1 -1
  269. package/dist/stt/stream_adapter.js +3 -4
  270. package/dist/stt/stream_adapter.js.map +1 -1
  271. package/dist/stt/stt.cjs +101 -10
  272. package/dist/stt/stt.cjs.map +1 -1
  273. package/dist/stt/stt.d.cts +26 -5
  274. package/dist/stt/stt.d.ts +26 -5
  275. package/dist/stt/stt.d.ts.map +1 -1
  276. package/dist/stt/stt.js +102 -11
  277. package/dist/stt/stt.js.map +1 -1
  278. package/dist/tokenize/basic/basic.cjs +10 -5
  279. package/dist/tokenize/basic/basic.cjs.map +1 -1
  280. package/dist/tokenize/basic/basic.d.cts +7 -1
  281. package/dist/tokenize/basic/basic.d.ts +7 -1
  282. package/dist/tokenize/basic/basic.d.ts.map +1 -1
  283. package/dist/tokenize/basic/basic.js +10 -5
  284. package/dist/tokenize/basic/basic.js.map +1 -1
  285. package/dist/tokenize/basic/sentence.cjs +14 -6
  286. package/dist/tokenize/basic/sentence.cjs.map +1 -1
  287. package/dist/tokenize/basic/sentence.d.cts +1 -1
  288. package/dist/tokenize/basic/sentence.d.ts +1 -1
  289. package/dist/tokenize/basic/sentence.d.ts.map +1 -1
  290. package/dist/tokenize/basic/sentence.js +14 -6
  291. package/dist/tokenize/basic/sentence.js.map +1 -1
  292. package/dist/tokenize/token_stream.cjs +5 -3
  293. package/dist/tokenize/token_stream.cjs.map +1 -1
  294. package/dist/tokenize/token_stream.d.cts +1 -0
  295. package/dist/tokenize/token_stream.d.ts +1 -0
  296. package/dist/tokenize/token_stream.d.ts.map +1 -1
  297. package/dist/tokenize/token_stream.js +6 -4
  298. package/dist/tokenize/token_stream.js.map +1 -1
  299. package/dist/transcription.cjs +1 -2
  300. package/dist/transcription.cjs.map +1 -1
  301. package/dist/transcription.d.ts.map +1 -1
  302. package/dist/transcription.js +2 -3
  303. package/dist/transcription.js.map +1 -1
  304. package/dist/tts/index.cjs +2 -4
  305. package/dist/tts/index.cjs.map +1 -1
  306. package/dist/tts/index.d.cts +1 -1
  307. package/dist/tts/index.d.ts +1 -1
  308. package/dist/tts/index.d.ts.map +1 -1
  309. package/dist/tts/index.js +1 -3
  310. package/dist/tts/index.js.map +1 -1
  311. package/dist/tts/stream_adapter.cjs +26 -13
  312. package/dist/tts/stream_adapter.cjs.map +1 -1
  313. package/dist/tts/stream_adapter.d.cts +1 -1
  314. package/dist/tts/stream_adapter.d.ts +1 -1
  315. package/dist/tts/stream_adapter.d.ts.map +1 -1
  316. package/dist/tts/stream_adapter.js +27 -14
  317. package/dist/tts/stream_adapter.js.map +1 -1
  318. package/dist/tts/tts.cjs +157 -25
  319. package/dist/tts/tts.cjs.map +1 -1
  320. package/dist/tts/tts.d.cts +29 -5
  321. package/dist/tts/tts.d.ts +29 -5
  322. package/dist/tts/tts.d.ts.map +1 -1
  323. package/dist/tts/tts.js +157 -24
  324. package/dist/tts/tts.js.map +1 -1
  325. package/dist/types.cjs +60 -0
  326. package/dist/types.cjs.map +1 -0
  327. package/dist/types.d.cts +13 -0
  328. package/dist/types.d.ts +13 -0
  329. package/dist/types.d.ts.map +1 -0
  330. package/dist/types.js +35 -0
  331. package/dist/types.js.map +1 -0
  332. package/dist/utils.cjs +281 -27
  333. package/dist/utils.cjs.map +1 -1
  334. package/dist/utils.d.cts +134 -9
  335. package/dist/utils.d.ts +134 -9
  336. package/dist/utils.d.ts.map +1 -1
  337. package/dist/utils.js +265 -26
  338. package/dist/utils.js.map +1 -1
  339. package/dist/utils.test.cjs +492 -0
  340. package/dist/utils.test.cjs.map +1 -0
  341. package/dist/utils.test.js +498 -0
  342. package/dist/utils.test.js.map +1 -0
  343. package/dist/vad.cjs +76 -20
  344. package/dist/vad.cjs.map +1 -1
  345. package/dist/vad.d.cts +25 -5
  346. package/dist/vad.d.ts +25 -5
  347. package/dist/vad.d.ts.map +1 -1
  348. package/dist/vad.js +76 -20
  349. package/dist/vad.js.map +1 -1
  350. package/dist/voice/agent.cjs +245 -0
  351. package/dist/voice/agent.cjs.map +1 -0
  352. package/dist/voice/agent.d.cts +78 -0
  353. package/dist/voice/agent.d.ts +78 -0
  354. package/dist/voice/agent.d.ts.map +1 -0
  355. package/dist/voice/agent.js +220 -0
  356. package/dist/voice/agent.js.map +1 -0
  357. package/dist/voice/agent.test.cjs +61 -0
  358. package/dist/voice/agent.test.cjs.map +1 -0
  359. package/dist/voice/agent.test.js +60 -0
  360. package/dist/voice/agent.test.js.map +1 -0
  361. package/dist/voice/agent_activity.cjs +1453 -0
  362. package/dist/voice/agent_activity.cjs.map +1 -0
  363. package/dist/voice/agent_activity.d.cts +94 -0
  364. package/dist/voice/agent_activity.d.ts +94 -0
  365. package/dist/voice/agent_activity.d.ts.map +1 -0
  366. package/dist/voice/agent_activity.js +1449 -0
  367. package/dist/voice/agent_activity.js.map +1 -0
  368. package/dist/voice/agent_session.cjs +312 -0
  369. package/dist/voice/agent_session.cjs.map +1 -0
  370. package/dist/voice/agent_session.d.cts +121 -0
  371. package/dist/voice/agent_session.d.ts +121 -0
  372. package/dist/voice/agent_session.d.ts.map +1 -0
  373. package/dist/voice/agent_session.js +295 -0
  374. package/dist/voice/agent_session.js.map +1 -0
  375. package/dist/voice/audio_recognition.cjs +375 -0
  376. package/dist/voice/audio_recognition.cjs.map +1 -0
  377. package/dist/voice/audio_recognition.d.cts +80 -0
  378. package/dist/voice/audio_recognition.d.ts +80 -0
  379. package/dist/voice/audio_recognition.d.ts.map +1 -0
  380. package/dist/voice/audio_recognition.js +351 -0
  381. package/dist/voice/audio_recognition.js.map +1 -0
  382. package/dist/voice/events.cjs +145 -0
  383. package/dist/voice/events.cjs.map +1 -0
  384. package/dist/voice/events.d.cts +124 -0
  385. package/dist/voice/events.d.ts +124 -0
  386. package/dist/voice/events.d.ts.map +1 -0
  387. package/dist/voice/events.js +110 -0
  388. package/dist/voice/events.js.map +1 -0
  389. package/dist/voice/generation.cjs +700 -0
  390. package/dist/voice/generation.cjs.map +1 -0
  391. package/dist/voice/generation.d.cts +115 -0
  392. package/dist/voice/generation.d.ts +115 -0
  393. package/dist/voice/generation.d.ts.map +1 -0
  394. package/dist/voice/generation.js +672 -0
  395. package/dist/voice/generation.js.map +1 -0
  396. package/dist/voice/index.cjs +40 -0
  397. package/dist/voice/index.cjs.map +1 -0
  398. package/dist/voice/index.d.cts +5 -0
  399. package/dist/voice/index.d.ts +5 -0
  400. package/dist/voice/index.d.ts.map +1 -0
  401. package/dist/voice/index.js +11 -0
  402. package/dist/voice/index.js.map +1 -0
  403. package/dist/voice/io.cjs +245 -0
  404. package/dist/voice/io.cjs.map +1 -0
  405. package/dist/voice/io.d.cts +101 -0
  406. package/dist/voice/io.d.ts +101 -0
  407. package/dist/voice/io.d.ts.map +1 -0
  408. package/dist/voice/io.js +217 -0
  409. package/dist/voice/io.js.map +1 -0
  410. package/dist/voice/room_io/_input.cjs +121 -0
  411. package/dist/voice/room_io/_input.cjs.map +1 -0
  412. package/dist/voice/room_io/_input.d.cts +24 -0
  413. package/dist/voice/room_io/_input.d.ts +24 -0
  414. package/dist/voice/room_io/_input.d.ts.map +1 -0
  415. package/dist/voice/room_io/_input.js +102 -0
  416. package/dist/voice/room_io/_input.js.map +1 -0
  417. package/dist/voice/room_io/_output.cjs +358 -0
  418. package/dist/voice/room_io/_output.cjs.map +1 -0
  419. package/dist/voice/room_io/_output.d.cts +75 -0
  420. package/dist/voice/room_io/_output.d.ts +75 -0
  421. package/dist/voice/room_io/_output.d.ts.map +1 -0
  422. package/dist/voice/room_io/_output.js +342 -0
  423. package/dist/voice/room_io/_output.js.map +1 -0
  424. package/dist/voice/room_io/index.cjs +25 -0
  425. package/dist/voice/room_io/index.cjs.map +1 -0
  426. package/dist/voice/room_io/index.d.cts +3 -0
  427. package/dist/voice/room_io/index.d.ts +3 -0
  428. package/dist/voice/room_io/index.d.ts.map +1 -0
  429. package/dist/voice/room_io/index.js +3 -0
  430. package/dist/voice/room_io/index.js.map +1 -0
  431. package/dist/voice/room_io/room_io.cjs +370 -0
  432. package/dist/voice/room_io/room_io.cjs.map +1 -0
  433. package/dist/voice/room_io/room_io.d.cts +73 -0
  434. package/dist/voice/room_io/room_io.d.ts +73 -0
  435. package/dist/voice/room_io/room_io.d.ts.map +1 -0
  436. package/dist/voice/room_io/room_io.js +361 -0
  437. package/dist/voice/room_io/room_io.js.map +1 -0
  438. package/dist/{pipeline/index.cjs → voice/run_context.cjs} +16 -11
  439. package/dist/voice/run_context.cjs.map +1 -0
  440. package/dist/voice/run_context.d.cts +12 -0
  441. package/dist/voice/run_context.d.ts +12 -0
  442. package/dist/voice/run_context.d.ts.map +1 -0
  443. package/dist/voice/run_context.js +14 -0
  444. package/dist/voice/run_context.js.map +1 -0
  445. package/dist/voice/speech_handle.cjs +105 -0
  446. package/dist/voice/speech_handle.cjs.map +1 -0
  447. package/dist/voice/speech_handle.d.cts +46 -0
  448. package/dist/voice/speech_handle.d.ts +46 -0
  449. package/dist/voice/speech_handle.d.ts.map +1 -0
  450. package/dist/voice/speech_handle.js +81 -0
  451. package/dist/voice/speech_handle.js.map +1 -0
  452. package/dist/voice/transcription/_utils.cjs +45 -0
  453. package/dist/voice/transcription/_utils.cjs.map +1 -0
  454. package/dist/voice/transcription/_utils.d.cts +3 -0
  455. package/dist/voice/transcription/_utils.d.ts +3 -0
  456. package/dist/voice/transcription/_utils.d.ts.map +1 -0
  457. package/dist/voice/transcription/_utils.js +21 -0
  458. package/dist/voice/transcription/_utils.js.map +1 -0
  459. package/dist/voice/transcription/index.cjs +23 -0
  460. package/dist/voice/transcription/index.cjs.map +1 -0
  461. package/dist/voice/transcription/index.d.cts +2 -0
  462. package/dist/voice/transcription/index.d.ts +2 -0
  463. package/dist/voice/transcription/index.d.ts.map +1 -0
  464. package/dist/voice/transcription/index.js +2 -0
  465. package/dist/voice/transcription/index.js.map +1 -0
  466. package/dist/voice/transcription/synchronizer.cjs +380 -0
  467. package/dist/voice/transcription/synchronizer.cjs.map +1 -0
  468. package/dist/voice/transcription/synchronizer.d.cts +86 -0
  469. package/dist/voice/transcription/synchronizer.d.ts +86 -0
  470. package/dist/voice/transcription/synchronizer.d.ts.map +1 -0
  471. package/dist/voice/transcription/synchronizer.js +355 -0
  472. package/dist/voice/transcription/synchronizer.js.map +1 -0
  473. package/dist/worker.cjs +22 -4
  474. package/dist/worker.cjs.map +1 -1
  475. package/dist/worker.d.cts +1 -1
  476. package/dist/worker.d.ts +1 -1
  477. package/dist/worker.d.ts.map +1 -1
  478. package/dist/worker.js +22 -4
  479. package/dist/worker.js.map +1 -1
  480. package/package.json +9 -2
  481. package/src/_exceptions.ts +137 -0
  482. package/src/audio.ts +12 -1
  483. package/src/cli.ts +37 -0
  484. package/src/constants.ts +2 -0
  485. package/src/http_server.ts +1 -0
  486. package/src/index.ts +13 -10
  487. package/src/inference_runner.ts +2 -3
  488. package/src/ipc/inference_proc_executor.ts +2 -2
  489. package/src/ipc/job_executor.ts +1 -1
  490. package/src/ipc/job_proc_executor.ts +1 -1
  491. package/src/ipc/job_proc_lazy_main.ts +1 -1
  492. package/src/job.ts +18 -0
  493. package/src/llm/__snapshots__/chat_context.test.ts.snap +527 -0
  494. package/src/llm/__snapshots__/tool_context.test.ts.snap +177 -0
  495. package/src/llm/__snapshots__/utils.test.ts.snap +65 -0
  496. package/src/llm/chat_context.test.ts +450 -0
  497. package/src/llm/chat_context.ts +501 -103
  498. package/src/llm/index.ts +53 -18
  499. package/src/llm/llm.ts +149 -50
  500. package/src/llm/provider_format/google.test.ts +772 -0
  501. package/src/llm/provider_format/google.ts +130 -0
  502. package/src/llm/provider_format/index.ts +23 -0
  503. package/src/llm/provider_format/openai.test.ts +581 -0
  504. package/src/llm/provider_format/openai.ts +118 -0
  505. package/src/llm/provider_format/utils.ts +183 -0
  506. package/src/llm/realtime.ts +151 -0
  507. package/src/llm/remote_chat_context.test.ts +290 -0
  508. package/src/llm/remote_chat_context.ts +114 -0
  509. package/src/llm/tool_context.test.ts +198 -0
  510. package/src/llm/tool_context.ts +259 -0
  511. package/src/llm/tool_context.type.test.ts +115 -0
  512. package/src/llm/utils.test.ts +670 -0
  513. package/src/llm/utils.ts +324 -0
  514. package/src/metrics/base.ts +110 -78
  515. package/src/metrics/index.ts +3 -9
  516. package/src/metrics/usage_collector.ts +19 -13
  517. package/src/metrics/utils.ts +24 -69
  518. package/src/multimodal/index.ts +0 -1
  519. package/src/plugin.ts +26 -8
  520. package/src/stream/deferred_stream.test.ts +755 -0
  521. package/src/stream/deferred_stream.ts +110 -0
  522. package/src/stream/identity_transform.test.ts +179 -0
  523. package/src/stream/identity_transform.ts +18 -0
  524. package/src/stream/index.ts +7 -0
  525. package/src/stream/merge_readable_streams.ts +40 -0
  526. package/src/stream/stream_channel.test.ts +129 -0
  527. package/src/stream/stream_channel.ts +32 -0
  528. package/src/stt/stream_adapter.ts +3 -5
  529. package/src/stt/stt.ts +135 -17
  530. package/src/tokenize/basic/basic.ts +13 -5
  531. package/src/tokenize/basic/sentence.ts +20 -6
  532. package/src/tokenize/token_stream.ts +7 -4
  533. package/src/transcription.ts +2 -3
  534. package/src/tts/index.ts +0 -1
  535. package/src/tts/stream_adapter.ts +42 -16
  536. package/src/tts/tts.ts +203 -21
  537. package/src/types.ts +42 -0
  538. package/src/utils.test.ts +658 -0
  539. package/src/utils.ts +375 -44
  540. package/src/vad.ts +90 -22
  541. package/src/voice/agent.test.ts +80 -0
  542. package/src/voice/agent.ts +332 -0
  543. package/src/voice/agent_activity.ts +1913 -0
  544. package/src/voice/agent_session.ts +460 -0
  545. package/src/voice/audio_recognition.ts +474 -0
  546. package/src/voice/events.ts +252 -0
  547. package/src/voice/generation.ts +881 -0
  548. package/src/voice/index.ts +7 -0
  549. package/src/voice/io.ts +304 -0
  550. package/src/voice/room_io/_input.ts +144 -0
  551. package/src/voice/room_io/_output.ts +436 -0
  552. package/src/voice/room_io/index.ts +5 -0
  553. package/src/voice/room_io/room_io.ts +495 -0
  554. package/src/voice/run_context.ts +20 -0
  555. package/src/voice/speech_handle.ts +104 -0
  556. package/src/voice/transcription/_utils.ts +25 -0
  557. package/src/voice/transcription/index.ts +4 -0
  558. package/src/voice/transcription/synchronizer.ts +478 -0
  559. package/src/worker.ts +22 -2
  560. package/dist/llm/function_context.cjs +0 -103
  561. package/dist/llm/function_context.cjs.map +0 -1
  562. package/dist/llm/function_context.d.cts +0 -47
  563. package/dist/llm/function_context.d.ts +0 -47
  564. package/dist/llm/function_context.d.ts.map +0 -1
  565. package/dist/llm/function_context.js +0 -78
  566. package/dist/llm/function_context.js.map +0 -1
  567. package/dist/llm/function_context.test.cjs +0 -218
  568. package/dist/llm/function_context.test.cjs.map +0 -1
  569. package/dist/llm/function_context.test.js +0 -217
  570. package/dist/llm/function_context.test.js.map +0 -1
  571. package/dist/multimodal/multimodal_agent.cjs +0 -451
  572. package/dist/multimodal/multimodal_agent.cjs.map +0 -1
  573. package/dist/multimodal/multimodal_agent.d.cts +0 -48
  574. package/dist/multimodal/multimodal_agent.d.ts +0 -48
  575. package/dist/multimodal/multimodal_agent.d.ts.map +0 -1
  576. package/dist/multimodal/multimodal_agent.js +0 -425
  577. package/dist/multimodal/multimodal_agent.js.map +0 -1
  578. package/dist/pipeline/agent_output.cjs +0 -197
  579. package/dist/pipeline/agent_output.cjs.map +0 -1
  580. package/dist/pipeline/agent_output.d.cts +0 -33
  581. package/dist/pipeline/agent_output.d.ts +0 -33
  582. package/dist/pipeline/agent_output.d.ts.map +0 -1
  583. package/dist/pipeline/agent_output.js +0 -172
  584. package/dist/pipeline/agent_output.js.map +0 -1
  585. package/dist/pipeline/agent_playout.cjs +0 -175
  586. package/dist/pipeline/agent_playout.cjs.map +0 -1
  587. package/dist/pipeline/agent_playout.d.cts +0 -40
  588. package/dist/pipeline/agent_playout.d.ts +0 -40
  589. package/dist/pipeline/agent_playout.d.ts.map +0 -1
  590. package/dist/pipeline/agent_playout.js +0 -139
  591. package/dist/pipeline/agent_playout.js.map +0 -1
  592. package/dist/pipeline/human_input.cjs +0 -171
  593. package/dist/pipeline/human_input.cjs.map +0 -1
  594. package/dist/pipeline/human_input.d.cts +0 -30
  595. package/dist/pipeline/human_input.d.ts +0 -30
  596. package/dist/pipeline/human_input.d.ts.map +0 -1
  597. package/dist/pipeline/human_input.js +0 -146
  598. package/dist/pipeline/human_input.js.map +0 -1
  599. package/dist/pipeline/index.cjs.map +0 -1
  600. package/dist/pipeline/index.d.cts +0 -2
  601. package/dist/pipeline/index.d.ts +0 -2
  602. package/dist/pipeline/index.d.ts.map +0 -1
  603. package/dist/pipeline/index.js +0 -11
  604. package/dist/pipeline/index.js.map +0 -1
  605. package/dist/pipeline/pipeline_agent.cjs +0 -849
  606. package/dist/pipeline/pipeline_agent.cjs.map +0 -1
  607. package/dist/pipeline/pipeline_agent.d.cts +0 -150
  608. package/dist/pipeline/pipeline_agent.d.ts +0 -150
  609. package/dist/pipeline/pipeline_agent.d.ts.map +0 -1
  610. package/dist/pipeline/pipeline_agent.js +0 -826
  611. package/dist/pipeline/pipeline_agent.js.map +0 -1
  612. package/dist/pipeline/speech_handle.cjs +0 -176
  613. package/dist/pipeline/speech_handle.cjs.map +0 -1
  614. package/dist/pipeline/speech_handle.d.cts +0 -37
  615. package/dist/pipeline/speech_handle.d.ts +0 -37
  616. package/dist/pipeline/speech_handle.d.ts.map +0 -1
  617. package/dist/pipeline/speech_handle.js +0 -152
  618. package/dist/pipeline/speech_handle.js.map +0 -1
  619. package/src/llm/function_context.test.ts +0 -248
  620. package/src/llm/function_context.ts +0 -142
  621. package/src/multimodal/multimodal_agent.ts +0 -555
  622. package/src/pipeline/agent_output.ts +0 -219
  623. package/src/pipeline/agent_playout.ts +0 -192
  624. package/src/pipeline/human_input.ts +0 -188
  625. package/src/pipeline/index.ts +0 -15
  626. package/src/pipeline/pipeline_agent.ts +0 -1185
  627. package/src/pipeline/speech_handle.ts +0 -201
@@ -1 +1 @@
1
- {"version":3,"sources":["../../../src/tokenize/basic/sentence.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Split the text into sentences.\n */\nexport const splitSentences = (text: string, minLength = 20): [string, number, number][] => {\n const alphabets = /([A-Za-z])/g;\n const prefixes = /(Mr|St|Mrs|Ms|Dr)[.]/g;\n const suffixes = /(Inc|Ltd|Jr|Sr|Co)/g;\n const starters =\n /(Mr|Mrs|Ms|Dr|Prof|Capt|Cpt|Lt|He\\s|She\\s|It\\s|They\\s|Their\\s|Our\\s|We\\s|But\\s|However\\s|That\\s|This\\s|Wherever)/g;\n const acronyms = /([A-Z][.][A-Z][.](?:[A-Z][.])?)/g;\n const websites = /[.](com|net|org|io|gov|edu|me)/g;\n const digits = /([0-9])/g;\n const dots = /\\.{2,}/g;\n\n text = text.replaceAll('\\n', ' ');\n text = text.replaceAll(prefixes, '$1<prd>');\n text = text.replaceAll(websites, '<prd>$2');\n text = text.replaceAll(new RegExp(`${digits.source}[.]${digits.source}`, 'g'), '$1<prd>$2');\n text = text.replaceAll(dots, (match) => '<prd>'.repeat(match.length));\n text = text.replaceAll('Ph.D.', 'Ph<prd>D<prd>');\n text = text.replaceAll(new RegExp(`\\\\s${alphabets.source}[.] `, 'g'), ' $1<prd> ');\n text = text.replaceAll(new RegExp(`${acronyms.source} ${starters.source}`, 'g'), '$1<stop> $2');\n text = text.replaceAll(\n new RegExp(`${alphabets.source}[.]${alphabets.source}[.]${alphabets.source}[.]`, 'g'),\n '$1<prd>$2<prd>$3<prd>',\n );\n text = text.replaceAll(\n new RegExp(`${alphabets.source}[.]${alphabets.source}[.]`, 'g'),\n '$1<prd>$2<prd>',\n );\n text = text.replaceAll(\n new RegExp(` ${suffixes.source}[.] ${starters.source}`, 'g'),\n '$1<stop> $2',\n );\n text = text.replaceAll(new RegExp(` ${suffixes.source}[.]`, 'g'), '$1<prd>');\n text = text.replaceAll(new RegExp(` ${alphabets.source}[.]`, 'g'), '$1<prd>');\n text = text.replaceAll('.”', '”.');\n text = text.replaceAll('.\"', '\".');\n text = text.replaceAll('!\"', '\"!');\n text = text.replaceAll('?\"', '\"?');\n text = text.replaceAll('.', '.<stop>');\n text = text.replaceAll('?', '?<stop>');\n text = text.replaceAll('!', '!<stop>');\n text = text.replaceAll('<prd>', '.');\n\n const split = text.split('<stop>');\n text = text.replaceAll('<stop>', '');\n\n const sentences: [string, number, number][] = [];\n let buf = '';\n let start = 0;\n let end = 0;\n for (const match of split) {\n const sentence = match.trim();\n if (!sentence) continue;\n\n buf += ' ' + sentence;\n end += match.length;\n if (buf.length > minLength) {\n sentences.push([buf.slice(1), start, end]);\n start = end;\n buf = '';\n }\n }\n\n if (buf) {\n sentences.push([buf.slice(1), start, text.length - 1]);\n }\n\n return sentences;\n};\n"],"mappings":"AAOO,MAAM,iBAAiB,CAAC,MAAc,YAAY,OAAmC;AAC1F,QAAM,YAAY;AAClB,QAAM,WAAW;AACjB,QAAM,WAAW;AACjB,QAAM,WACJ;AACF,QAAM,WAAW;AACjB,QAAM,WAAW;AACjB,QAAM,SAAS;AACf,QAAM,OAAO;AAEb,SAAO,KAAK,WAAW,MAAM,GAAG;AAChC,SAAO,KAAK,WAAW,UAAU,SAAS;AAC1C,SAAO,KAAK,WAAW,UAAU,SAAS;AAC1C,SAAO,KAAK,WAAW,IAAI,OAAO,GAAG,OAAO,MAAM,MAAM,OAAO,MAAM,IAAI,GAAG,GAAG,WAAW;AAC1F,SAAO,KAAK,WAAW,MAAM,CAAC,UAAU,QAAQ,OAAO,MAAM,MAAM,CAAC;AACpE,SAAO,KAAK,WAAW,SAAS,eAAe;AAC/C,SAAO,KAAK,WAAW,IAAI,OAAO,MAAM,UAAU,MAAM,QAAQ,GAAG,GAAG,WAAW;AACjF,SAAO,KAAK,WAAW,IAAI,OAAO,GAAG,SAAS,MAAM,IAAI,SAAS,MAAM,IAAI,GAAG,GAAG,aAAa;AAC9F,SAAO,KAAK;AAAA,IACV,IAAI,OAAO,GAAG,UAAU,MAAM,MAAM,UAAU,MAAM,MAAM,UAAU,MAAM,OAAO,GAAG;AAAA,IACpF;AAAA,EACF;AACA,SAAO,KAAK;AAAA,IACV,IAAI,OAAO,GAAG,UAAU,MAAM,MAAM,UAAU,MAAM,OAAO,GAAG;AAAA,IAC9D;AAAA,EACF;AACA,SAAO,KAAK;AAAA,IACV,IAAI,OAAO,IAAI,SAAS,MAAM,OAAO,SAAS,MAAM,IAAI,GAAG;AAAA,IAC3D;AAAA,EACF;AACA,SAAO,KAAK,WAAW,IAAI,OAAO,IAAI,SAAS,MAAM,OAAO,GAAG,GAAG,SAAS;AAC3E,SAAO,KAAK,WAAW,IAAI,OAAO,IAAI,UAAU,MAAM,OAAO,GAAG,GAAG,SAAS;AAC5E,SAAO,KAAK,WAAW,WAAM,SAAI;AACjC,SAAO,KAAK,WAAW,MAAM,IAAI;AACjC,SAAO,KAAK,WAAW,MAAM,IAAI;AACjC,SAAO,KAAK,WAAW,MAAM,IAAI;AACjC,SAAO,KAAK,WAAW,KAAK,SAAS;AACrC,SAAO,KAAK,WAAW,KAAK,SAAS;AACrC,SAAO,KAAK,WAAW,KAAK,SAAS;AACrC,SAAO,KAAK,WAAW,SAAS,GAAG;AAEnC,QAAM,QAAQ,KAAK,MAAM,QAAQ;AACjC,SAAO,KAAK,WAAW,UAAU,EAAE;AAEnC,QAAM,YAAwC,CAAC;AAC/C,MAAI,MAAM;AACV,MAAI,QAAQ;AACZ,MAAI,MAAM;AACV,aAAW,SAAS,OAAO;AACzB,UAAM,WAAW,MAAM,KAAK;AAC5B,QAAI,CAAC,SAAU;AAEf,WAAO,MAAM;AACb,WAAO,MAAM;AACb,QAAI,IAAI,SAAS,WAAW;AAC1B,gBAAU,KAAK,CAAC,IAAI,MAAM,CAAC,GAAG,OAAO,GAAG,CAAC;AACzC,cAAQ;AACR,YAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI,KAAK;AACP,cAAU,KAAK,CAAC,IAAI,MAAM,CAAC,GAAG,OAAO,KAAK,SAAS,CAAC,CAAC;AAAA,EACvD;AAEA,SAAO;AACT;","names":[]}
1
+ {"version":3,"sources":["../../../src/tokenize/basic/sentence.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Split the text into sentences.\n */\nexport const splitSentences = (\n text: string,\n minLength = 20,\n retainFormat: boolean = false,\n): [string, number, number][] => {\n const alphabets = /([A-Za-z])/g;\n const prefixes = /(Mr|St|Mrs|Ms|Dr)[.]/g;\n const suffixes = /(Inc|Ltd|Jr|Sr|Co)/g;\n const starters =\n /(Mr|Mrs|Ms|Dr|Prof|Capt|Cpt|Lt|He\\s|She\\s|It\\s|They\\s|Their\\s|Our\\s|We\\s|But\\s|However\\s|That\\s|This\\s|Wherever)/g;\n const acronyms = /([A-Z][.][A-Z][.](?:[A-Z][.])?)/g;\n const websites = /[.](com|net|org|io|gov|edu|me)/g;\n const digits = /([0-9])/g;\n const dots = /\\.{2,}/g;\n\n if (retainFormat) {\n text = text.replaceAll('\\n', '<nel><stop>');\n } else {\n text = text.replaceAll('\\n', ' ');\n }\n\n text = text.replaceAll(prefixes, '$1<prd>');\n text = text.replaceAll(websites, '<prd>$2');\n text = text.replaceAll(new RegExp(`${digits.source}[.]${digits.source}`, 'g'), '$1<prd>$2');\n text = text.replaceAll(dots, (match) => '<prd>'.repeat(match.length));\n text = text.replaceAll('Ph.D.', 'Ph<prd>D<prd>');\n text = text.replaceAll(new RegExp(`\\\\s${alphabets.source}[.] `, 'g'), ' $1<prd> ');\n text = text.replaceAll(new RegExp(`${acronyms.source} ${starters.source}`, 'g'), '$1<stop> $2');\n text = text.replaceAll(\n new RegExp(`${alphabets.source}[.]${alphabets.source}[.]${alphabets.source}[.]`, 'g'),\n '$1<prd>$2<prd>$3<prd>',\n );\n text = text.replaceAll(\n new RegExp(`${alphabets.source}[.]${alphabets.source}[.]`, 'g'),\n '$1<prd>$2<prd>',\n );\n text = text.replaceAll(\n new RegExp(` ${suffixes.source}[.] ${starters.source}`, 'g'),\n '$1<stop> $2',\n );\n text = text.replaceAll(new RegExp(` ${suffixes.source}[.]`, 'g'), '$1<prd>');\n text = text.replaceAll(new RegExp(` ${alphabets.source}[.]`, 'g'), '$1<prd>');\n text = text.replaceAll('.”', '”.');\n text = text.replaceAll('.\"', '\".');\n text = text.replaceAll('!\"', '\"!');\n text = text.replaceAll('?\"', '\"?');\n text = text.replaceAll('.', '.<stop>');\n text = text.replaceAll('?', '?<stop>');\n text = text.replaceAll('!', '!<stop>');\n text = text.replaceAll('<prd>', '.');\n\n if (retainFormat) {\n text = text.replaceAll('<nel>', '\\n');\n }\n\n const split = text.split('<stop>');\n text = text.replaceAll('<stop>', '');\n\n const sentences: [string, number, number][] = [];\n let buf = '';\n let start = 0;\n let end = 0;\n const prePad = retainFormat ? '' : ' ';\n for (const match of split) {\n const sentence = retainFormat ? match : match.trim();\n if (!sentence) continue;\n\n buf += prePad + sentence;\n end += match.length;\n if (buf.length > minLength) {\n sentences.push([buf.slice(prePad.length), start, end]);\n start = end;\n buf = '';\n }\n }\n\n if (buf) {\n sentences.push([buf.slice(prePad.length), start, text.length - 1]);\n }\n\n return sentences;\n};\n"],"mappings":"AAOO,MAAM,iBAAiB,CAC5B,MACA,YAAY,IACZ,eAAwB,UACO;AAC/B,QAAM,YAAY;AAClB,QAAM,WAAW;AACjB,QAAM,WAAW;AACjB,QAAM,WACJ;AACF,QAAM,WAAW;AACjB,QAAM,WAAW;AACjB,QAAM,SAAS;AACf,QAAM,OAAO;AAEb,MAAI,cAAc;AAChB,WAAO,KAAK,WAAW,MAAM,aAAa;AAAA,EAC5C,OAAO;AACL,WAAO,KAAK,WAAW,MAAM,GAAG;AAAA,EAClC;AAEA,SAAO,KAAK,WAAW,UAAU,SAAS;AAC1C,SAAO,KAAK,WAAW,UAAU,SAAS;AAC1C,SAAO,KAAK,WAAW,IAAI,OAAO,GAAG,OAAO,MAAM,MAAM,OAAO,MAAM,IAAI,GAAG,GAAG,WAAW;AAC1F,SAAO,KAAK,WAAW,MAAM,CAAC,UAAU,QAAQ,OAAO,MAAM,MAAM,CAAC;AACpE,SAAO,KAAK,WAAW,SAAS,eAAe;AAC/C,SAAO,KAAK,WAAW,IAAI,OAAO,MAAM,UAAU,MAAM,QAAQ,GAAG,GAAG,WAAW;AACjF,SAAO,KAAK,WAAW,IAAI,OAAO,GAAG,SAAS,MAAM,IAAI,SAAS,MAAM,IAAI,GAAG,GAAG,aAAa;AAC9F,SAAO,KAAK;AAAA,IACV,IAAI,OAAO,GAAG,UAAU,MAAM,MAAM,UAAU,MAAM,MAAM,UAAU,MAAM,OAAO,GAAG;AAAA,IACpF;AAAA,EACF;AACA,SAAO,KAAK;AAAA,IACV,IAAI,OAAO,GAAG,UAAU,MAAM,MAAM,UAAU,MAAM,OAAO,GAAG;AAAA,IAC9D;AAAA,EACF;AACA,SAAO,KAAK;AAAA,IACV,IAAI,OAAO,IAAI,SAAS,MAAM,OAAO,SAAS,MAAM,IAAI,GAAG;AAAA,IAC3D;AAAA,EACF;AACA,SAAO,KAAK,WAAW,IAAI,OAAO,IAAI,SAAS,MAAM,OAAO,GAAG,GAAG,SAAS;AAC3E,SAAO,KAAK,WAAW,IAAI,OAAO,IAAI,UAAU,MAAM,OAAO,GAAG,GAAG,SAAS;AAC5E,SAAO,KAAK,WAAW,WAAM,SAAI;AACjC,SAAO,KAAK,WAAW,MAAM,IAAI;AACjC,SAAO,KAAK,WAAW,MAAM,IAAI;AACjC,SAAO,KAAK,WAAW,MAAM,IAAI;AACjC,SAAO,KAAK,WAAW,KAAK,SAAS;AACrC,SAAO,KAAK,WAAW,KAAK,SAAS;AACrC,SAAO,KAAK,WAAW,KAAK,SAAS;AACrC,SAAO,KAAK,WAAW,SAAS,GAAG;AAEnC,MAAI,cAAc;AAChB,WAAO,KAAK,WAAW,SAAS,IAAI;AAAA,EACtC;AAEA,QAAM,QAAQ,KAAK,MAAM,QAAQ;AACjC,SAAO,KAAK,WAAW,UAAU,EAAE;AAEnC,QAAM,YAAwC,CAAC;AAC/C,MAAI,MAAM;AACV,MAAI,QAAQ;AACZ,MAAI,MAAM;AACV,QAAM,SAAS,eAAe,KAAK;AACnC,aAAW,SAAS,OAAO;AACzB,UAAM,WAAW,eAAe,QAAQ,MAAM,KAAK;AACnD,QAAI,CAAC,SAAU;AAEf,WAAO,SAAS;AAChB,WAAO,MAAM;AACb,QAAI,IAAI,SAAS,WAAW;AAC1B,gBAAU,KAAK,CAAC,IAAI,MAAM,OAAO,MAAM,GAAG,OAAO,GAAG,CAAC;AACrD,cAAQ;AACR,YAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI,KAAK;AACP,cAAU,KAAK,CAAC,IAAI,MAAM,OAAO,MAAM,GAAG,OAAO,KAAK,SAAS,CAAC,CAAC;AAAA,EACnE;AAEA,SAAO;AACT;","names":[]}
@@ -23,7 +23,6 @@ __export(token_stream_exports, {
23
23
  BufferedWordStream: () => BufferedWordStream
24
24
  });
25
25
  module.exports = __toCommonJS(token_stream_exports);
26
- var import_node_crypto = require("node:crypto");
27
26
  var import_utils = require("../utils.cjs");
28
27
  var import_tokenizer = require("./tokenizer.cjs");
29
28
  class BufferedTokenStream {
@@ -40,7 +39,7 @@ class BufferedTokenStream {
40
39
  this.#func = func;
41
40
  this.#minTokenLength = minTokenLength;
42
41
  this.#minContextLength = minContextLength;
43
- this.#currentSegmentId = (0, import_node_crypto.randomUUID)();
42
+ this.#currentSegmentId = (0, import_utils.shortuuid)();
44
43
  }
45
44
  /** Push a string of text into the token stream */
46
45
  pushText(text) {
@@ -90,7 +89,7 @@ class BufferedTokenStream {
90
89
  if (this.#outBuf) {
91
90
  this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });
92
91
  }
93
- this.#currentSegmentId = (0, import_node_crypto.randomUUID)();
92
+ this.#currentSegmentId = (0, import_utils.shortuuid)();
94
93
  }
95
94
  this.#inBuf = "";
96
95
  this.#outBuf = "";
@@ -131,6 +130,9 @@ class BufferedSentenceStream extends import_tokenizer.SentenceStream {
131
130
  super.close();
132
131
  this.#stream.close();
133
132
  }
133
+ endInput() {
134
+ this.#stream.endInput();
135
+ }
134
136
  next() {
135
137
  return this.#stream.next();
136
138
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/tokenize/token_stream.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { randomUUID } from 'node:crypto';\nimport { AsyncIterableQueue } from '../utils.js';\nimport type { TokenData } from './tokenizer.js';\nimport { SentenceStream, WordStream } from './tokenizer.js';\n\ntype TokenizeFunc = (x: string) => string[] | [string, number, number][];\n\nexport class BufferedTokenStream implements AsyncIterableIterator<TokenData> {\n protected queue = new AsyncIterableQueue<TokenData>();\n protected closed = false;\n\n #func: TokenizeFunc;\n #minTokenLength: number;\n #minContextLength: number;\n #bufTokens: string[] = [];\n #inBuf = '';\n #outBuf = '';\n #currentSegmentId: string;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n this.#func = func;\n this.#minTokenLength = minTokenLength;\n this.#minContextLength = minContextLength;\n\n this.#currentSegmentId = randomUUID();\n }\n\n /** Push a string of text into the token stream */\n pushText(text: string) {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n this.#inBuf += text;\n if (this.#inBuf.length < this.#minContextLength) return;\n\n while (true) {\n const tokens = this.#func(this.#inBuf);\n if (tokens.length <= 1) break;\n\n if (this.#outBuf) this.#outBuf += ' ';\n\n const tok = tokens.shift()!;\n let tokText: string;\n if (Array.isArray(tok)) {\n tokText = tok[0];\n } else {\n tokText = tok;\n }\n\n this.#outBuf += tokText;\n\n if (this.#outBuf.length >= this.#minTokenLength) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n this.#outBuf = '';\n }\n\n if (typeof tok! !== 'string') {\n this.#inBuf = this.#inBuf.slice(tok![2]);\n } else {\n this.#inBuf = this.#inBuf\n .slice(Math.max(0, this.#inBuf.indexOf(tok)) + tok.length)\n .trimStart();\n }\n }\n }\n\n /** Flush the stream, causing it to process all pending text */\n flush() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.#inBuf || this.#outBuf) {\n const tokens = this.#func(this.#inBuf);\n if (tokens) {\n if (this.#outBuf) this.#outBuf += ' ';\n\n if (Array.isArray(tokens[0])) {\n this.#outBuf += tokens.map((tok) => tok[0]).join(' ');\n } else {\n this.#outBuf += tokens.join(' ');\n }\n }\n\n if (this.#outBuf) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n }\n\n this.#currentSegmentId = randomUUID();\n }\n\n this.#inBuf = '';\n this.#outBuf = '';\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.flush();\n this.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.queue.next();\n }\n\n /** Close both the input and output of the token stream */\n close() {\n this.queue.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): BufferedTokenStream {\n return this;\n }\n}\n\nexport class BufferedSentenceStream extends SentenceStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n close() {\n super.close();\n this.#stream.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n\nexport class BufferedWordStream extends WordStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n endInput() {\n this.#stream.endInput();\n }\n\n close() {\n this.#stream.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,yBAA2B;AAC3B,mBAAmC;AAEnC,uBAA2C;AAIpC,MAAM,oBAAgE;AAAA,EACjE,QAAQ,IAAI,gCAA8B;AAAA,EAC1C,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACA,aAAuB,CAAC;AAAA,EACxB,SAAS;AAAA,EACT,UAAU;AAAA,EACV;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,SAAK,QAAQ;AACb,SAAK,kBAAkB;AACvB,SAAK,oBAAoB;AAEzB,SAAK,wBAAoB,+BAAW;AAAA,EACtC;AAAA;AAAA,EAGA,SAAS,MAAc;AACrB,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,SAAK,UAAU;AACf,QAAI,KAAK,OAAO,SAAS,KAAK,kBAAmB;AAEjD,WAAO,MAAM;AACX,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,OAAO,UAAU,EAAG;AAExB,UAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAM,MAAM,OAAO,MAAM;AACzB,UAAI;AACJ,UAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,kBAAU,IAAI,CAAC;AAAA,MACjB,OAAO;AACL,kBAAU;AAAA,MACZ;AAEA,WAAK,WAAW;AAEhB,UAAI,KAAK,QAAQ,UAAU,KAAK,iBAAiB;AAC/C,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AACzE,aAAK,UAAU;AAAA,MACjB;AAEA,UAAI,OAAO,QAAS,UAAU;AAC5B,aAAK,SAAS,KAAK,OAAO,MAAM,IAAK,CAAC,CAAC;AAAA,MACzC,OAAO;AACL,aAAK,SAAS,KAAK,OAChB,MAAM,KAAK,IAAI,GAAG,KAAK,OAAO,QAAQ,GAAG,CAAC,IAAI,IAAI,MAAM,EACxD,UAAU;AAAA,MACf;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,UAAU,KAAK,SAAS;AAC/B,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,QAAQ;AACV,YAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAI,MAAM,QAAQ,OAAO,CAAC,CAAC,GAAG;AAC5B,eAAK,WAAW,OAAO,IAAI,CAAC,QAAQ,IAAI,CAAC,CAAC,EAAE,KAAK,GAAG;AAAA,QACtD,OAAO;AACL,eAAK,WAAW,OAAO,KAAK,GAAG;AAAA,QACjC;AAAA,MACF;AAEA,UAAI,KAAK,SAAS;AAChB,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AAAA,MAC3E;AAEA,WAAK,wBAAoB,+BAAW;AAAA,IACtC;AAEA,SAAK,SAAS;AACd,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM;AACX,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAyB;AAC5C,WAAO;AAAA,EACT;AACF;AAEO,MAAM,+BAA+B,gCAAe;AAAA,EACzD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,QAAQ;AACN,UAAM,MAAM;AACZ,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;AAEO,MAAM,2BAA2B,4BAAW;AAAA,EACjD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,WAAW;AACT,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/tokenize/token_stream.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AsyncIterableQueue, shortuuid } from '../utils.js';\nimport type { TokenData } from './tokenizer.js';\nimport { SentenceStream, WordStream } from './tokenizer.js';\n\ntype TokenizeFunc = (x: string) => string[] | [string, number, number][];\n\nexport class BufferedTokenStream implements AsyncIterableIterator<TokenData> {\n protected queue = new AsyncIterableQueue<TokenData>();\n protected closed = false;\n\n #func: TokenizeFunc;\n #minTokenLength: number;\n #minContextLength: number;\n #bufTokens: string[] = [];\n #inBuf = '';\n #outBuf = '';\n #currentSegmentId: string;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n this.#func = func;\n this.#minTokenLength = minTokenLength;\n this.#minContextLength = minContextLength;\n\n this.#currentSegmentId = shortuuid();\n }\n\n /** Push a string of text into the token stream */\n pushText(text: string) {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n this.#inBuf += text;\n if (this.#inBuf.length < this.#minContextLength) return;\n\n while (true) {\n const tokens = this.#func(this.#inBuf);\n if (tokens.length <= 1) break;\n\n if (this.#outBuf) this.#outBuf += ' ';\n\n const tok = tokens.shift()!;\n let tokText: string;\n if (Array.isArray(tok)) {\n tokText = tok[0];\n } else {\n tokText = tok;\n }\n\n this.#outBuf += tokText;\n\n if (this.#outBuf.length >= this.#minTokenLength) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n this.#outBuf = '';\n }\n\n if (typeof tok! !== 'string') {\n this.#inBuf = this.#inBuf.slice(tok![2]);\n } else {\n this.#inBuf = this.#inBuf\n .slice(Math.max(0, this.#inBuf.indexOf(tok)) + tok.length)\n .trimStart();\n }\n }\n }\n\n /** Flush the stream, causing it to process all pending text */\n flush() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.#inBuf || this.#outBuf) {\n const tokens = this.#func(this.#inBuf);\n if (tokens) {\n if (this.#outBuf) this.#outBuf += ' ';\n\n if (Array.isArray(tokens[0])) {\n this.#outBuf += tokens.map((tok) => tok[0]).join(' ');\n } else {\n this.#outBuf += tokens.join(' ');\n }\n }\n\n if (this.#outBuf) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n }\n\n this.#currentSegmentId = shortuuid();\n }\n\n this.#inBuf = '';\n this.#outBuf = '';\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.flush();\n this.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.queue.next();\n }\n\n /** Close both the input and output of the token stream */\n close() {\n this.queue.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): BufferedTokenStream {\n return this;\n }\n}\n\nexport class BufferedSentenceStream extends SentenceStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n close() {\n super.close();\n this.#stream.close();\n }\n\n endInput() {\n this.#stream.endInput();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n\nexport class BufferedWordStream extends WordStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n endInput() {\n this.#stream.endInput();\n }\n\n close() {\n this.#stream.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,mBAA8C;AAE9C,uBAA2C;AAIpC,MAAM,oBAAgE;AAAA,EACjE,QAAQ,IAAI,gCAA8B;AAAA,EAC1C,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACA,aAAuB,CAAC;AAAA,EACxB,SAAS;AAAA,EACT,UAAU;AAAA,EACV;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,SAAK,QAAQ;AACb,SAAK,kBAAkB;AACvB,SAAK,oBAAoB;AAEzB,SAAK,wBAAoB,wBAAU;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,MAAc;AACrB,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,SAAK,UAAU;AACf,QAAI,KAAK,OAAO,SAAS,KAAK,kBAAmB;AAEjD,WAAO,MAAM;AACX,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,OAAO,UAAU,EAAG;AAExB,UAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAM,MAAM,OAAO,MAAM;AACzB,UAAI;AACJ,UAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,kBAAU,IAAI,CAAC;AAAA,MACjB,OAAO;AACL,kBAAU;AAAA,MACZ;AAEA,WAAK,WAAW;AAEhB,UAAI,KAAK,QAAQ,UAAU,KAAK,iBAAiB;AAC/C,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AACzE,aAAK,UAAU;AAAA,MACjB;AAEA,UAAI,OAAO,QAAS,UAAU;AAC5B,aAAK,SAAS,KAAK,OAAO,MAAM,IAAK,CAAC,CAAC;AAAA,MACzC,OAAO;AACL,aAAK,SAAS,KAAK,OAChB,MAAM,KAAK,IAAI,GAAG,KAAK,OAAO,QAAQ,GAAG,CAAC,IAAI,IAAI,MAAM,EACxD,UAAU;AAAA,MACf;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,UAAU,KAAK,SAAS;AAC/B,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,QAAQ;AACV,YAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAI,MAAM,QAAQ,OAAO,CAAC,CAAC,GAAG;AAC5B,eAAK,WAAW,OAAO,IAAI,CAAC,QAAQ,IAAI,CAAC,CAAC,EAAE,KAAK,GAAG;AAAA,QACtD,OAAO;AACL,eAAK,WAAW,OAAO,KAAK,GAAG;AAAA,QACjC;AAAA,MACF;AAEA,UAAI,KAAK,SAAS;AAChB,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AAAA,MAC3E;AAEA,WAAK,wBAAoB,wBAAU;AAAA,IACrC;AAEA,SAAK,SAAS;AACd,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM;AACX,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAyB;AAC5C,WAAO;AAAA,EACT;AACF;AAEO,MAAM,+BAA+B,gCAAe;AAAA,EACzD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,QAAQ;AACN,UAAM,MAAM;AACZ,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,WAAW;AACT,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;AAEO,MAAM,2BAA2B,4BAAW;AAAA,EACjD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,WAAW;AACT,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;","names":[]}
@@ -24,6 +24,7 @@ export declare class BufferedSentenceStream extends SentenceStream {
24
24
  pushText(text: string): void;
25
25
  flush(): void;
26
26
  close(): void;
27
+ endInput(): void;
27
28
  next(): Promise<IteratorResult<TokenData>>;
28
29
  }
29
30
  export declare class BufferedWordStream extends WordStream {
@@ -24,6 +24,7 @@ export declare class BufferedSentenceStream extends SentenceStream {
24
24
  pushText(text: string): void;
25
25
  flush(): void;
26
26
  close(): void;
27
+ endInput(): void;
27
28
  next(): Promise<IteratorResult<TokenData>>;
28
29
  }
29
30
  export declare class BufferedWordStream extends WordStream {
@@ -1 +1 @@
1
- {"version":3,"file":"token_stream.d.ts","sourceRoot":"","sources":["../../src/tokenize/token_stream.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,kBAAkB,EAAE,MAAM,aAAa,CAAC;AACjD,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,gBAAgB,CAAC;AAChD,OAAO,EAAE,cAAc,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAE5D,KAAK,YAAY,GAAG,CAAC,CAAC,EAAE,MAAM,KAAK,MAAM,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;AAEzE,qBAAa,mBAAoB,YAAW,qBAAqB,CAAC,SAAS,CAAC;;IAC1E,SAAS,CAAC,KAAK,gCAAuC;IACtD,SAAS,CAAC,MAAM,UAAS;gBAUb,IAAI,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM;IAQhF,kDAAkD;IAClD,QAAQ,CAAC,IAAI,EAAE,MAAM;IAuCrB,+DAA+D;IAC/D,KAAK;IA4BL,2DAA2D;IAC3D,QAAQ;IAQR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,SAAS,CAAC,CAAC;IAI1C,0DAA0D;IAC1D,KAAK;IAKL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,mBAAmB;CAG9C;AAED,qBAAa,sBAAuB,SAAQ,cAAc;;gBAG5C,IAAI,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM;IAKhF,QAAQ,CAAC,IAAI,EAAE,MAAM;IAIrB,KAAK;IAIL,KAAK;IAKL,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,SAAS,CAAC,CAAC;CAG3C;AAED,qBAAa,kBAAmB,SAAQ,UAAU;;gBAGpC,IAAI,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM;IAKhF,QAAQ,CAAC,IAAI,EAAE,MAAM;IAIrB,KAAK;IAIL,QAAQ;IAIR,KAAK;IAIL,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,SAAS,CAAC,CAAC;CAG3C"}
1
+ {"version":3,"file":"token_stream.d.ts","sourceRoot":"","sources":["../../src/tokenize/token_stream.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,kBAAkB,EAAa,MAAM,aAAa,CAAC;AAC5D,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,gBAAgB,CAAC;AAChD,OAAO,EAAE,cAAc,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAE5D,KAAK,YAAY,GAAG,CAAC,CAAC,EAAE,MAAM,KAAK,MAAM,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;AAEzE,qBAAa,mBAAoB,YAAW,qBAAqB,CAAC,SAAS,CAAC;;IAC1E,SAAS,CAAC,KAAK,gCAAuC;IACtD,SAAS,CAAC,MAAM,UAAS;gBAUb,IAAI,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM;IAQhF,kDAAkD;IAClD,QAAQ,CAAC,IAAI,EAAE,MAAM;IAuCrB,+DAA+D;IAC/D,KAAK;IA4BL,2DAA2D;IAC3D,QAAQ;IAQR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,SAAS,CAAC,CAAC;IAI1C,0DAA0D;IAC1D,KAAK;IAKL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,mBAAmB;CAG9C;AAED,qBAAa,sBAAuB,SAAQ,cAAc;;gBAG5C,IAAI,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM;IAKhF,QAAQ,CAAC,IAAI,EAAE,MAAM;IAIrB,KAAK;IAIL,KAAK;IAKL,QAAQ;IAIR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,SAAS,CAAC,CAAC;CAG3C;AAED,qBAAa,kBAAmB,SAAQ,UAAU;;gBAGpC,IAAI,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM;IAKhF,QAAQ,CAAC,IAAI,EAAE,MAAM;IAIrB,KAAK;IAIL,QAAQ;IAIR,KAAK;IAIL,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,SAAS,CAAC,CAAC;CAG3C"}
@@ -1,5 +1,4 @@
1
- import { randomUUID } from "node:crypto";
2
- import { AsyncIterableQueue } from "../utils.js";
1
+ import { AsyncIterableQueue, shortuuid } from "../utils.js";
3
2
  import { SentenceStream, WordStream } from "./tokenizer.js";
4
3
  class BufferedTokenStream {
5
4
  queue = new AsyncIterableQueue();
@@ -15,7 +14,7 @@ class BufferedTokenStream {
15
14
  this.#func = func;
16
15
  this.#minTokenLength = minTokenLength;
17
16
  this.#minContextLength = minContextLength;
18
- this.#currentSegmentId = randomUUID();
17
+ this.#currentSegmentId = shortuuid();
19
18
  }
20
19
  /** Push a string of text into the token stream */
21
20
  pushText(text) {
@@ -65,7 +64,7 @@ class BufferedTokenStream {
65
64
  if (this.#outBuf) {
66
65
  this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });
67
66
  }
68
- this.#currentSegmentId = randomUUID();
67
+ this.#currentSegmentId = shortuuid();
69
68
  }
70
69
  this.#inBuf = "";
71
70
  this.#outBuf = "";
@@ -106,6 +105,9 @@ class BufferedSentenceStream extends SentenceStream {
106
105
  super.close();
107
106
  this.#stream.close();
108
107
  }
108
+ endInput() {
109
+ this.#stream.endInput();
110
+ }
109
111
  next() {
110
112
  return this.#stream.next();
111
113
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/tokenize/token_stream.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { randomUUID } from 'node:crypto';\nimport { AsyncIterableQueue } from '../utils.js';\nimport type { TokenData } from './tokenizer.js';\nimport { SentenceStream, WordStream } from './tokenizer.js';\n\ntype TokenizeFunc = (x: string) => string[] | [string, number, number][];\n\nexport class BufferedTokenStream implements AsyncIterableIterator<TokenData> {\n protected queue = new AsyncIterableQueue<TokenData>();\n protected closed = false;\n\n #func: TokenizeFunc;\n #minTokenLength: number;\n #minContextLength: number;\n #bufTokens: string[] = [];\n #inBuf = '';\n #outBuf = '';\n #currentSegmentId: string;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n this.#func = func;\n this.#minTokenLength = minTokenLength;\n this.#minContextLength = minContextLength;\n\n this.#currentSegmentId = randomUUID();\n }\n\n /** Push a string of text into the token stream */\n pushText(text: string) {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n this.#inBuf += text;\n if (this.#inBuf.length < this.#minContextLength) return;\n\n while (true) {\n const tokens = this.#func(this.#inBuf);\n if (tokens.length <= 1) break;\n\n if (this.#outBuf) this.#outBuf += ' ';\n\n const tok = tokens.shift()!;\n let tokText: string;\n if (Array.isArray(tok)) {\n tokText = tok[0];\n } else {\n tokText = tok;\n }\n\n this.#outBuf += tokText;\n\n if (this.#outBuf.length >= this.#minTokenLength) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n this.#outBuf = '';\n }\n\n if (typeof tok! !== 'string') {\n this.#inBuf = this.#inBuf.slice(tok![2]);\n } else {\n this.#inBuf = this.#inBuf\n .slice(Math.max(0, this.#inBuf.indexOf(tok)) + tok.length)\n .trimStart();\n }\n }\n }\n\n /** Flush the stream, causing it to process all pending text */\n flush() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.#inBuf || this.#outBuf) {\n const tokens = this.#func(this.#inBuf);\n if (tokens) {\n if (this.#outBuf) this.#outBuf += ' ';\n\n if (Array.isArray(tokens[0])) {\n this.#outBuf += tokens.map((tok) => tok[0]).join(' ');\n } else {\n this.#outBuf += tokens.join(' ');\n }\n }\n\n if (this.#outBuf) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n }\n\n this.#currentSegmentId = randomUUID();\n }\n\n this.#inBuf = '';\n this.#outBuf = '';\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.flush();\n this.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.queue.next();\n }\n\n /** Close both the input and output of the token stream */\n close() {\n this.queue.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): BufferedTokenStream {\n return this;\n }\n}\n\nexport class BufferedSentenceStream extends SentenceStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n close() {\n super.close();\n this.#stream.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n\nexport class BufferedWordStream extends WordStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n endInput() {\n this.#stream.endInput();\n }\n\n close() {\n this.#stream.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n"],"mappings":"AAGA,SAAS,kBAAkB;AAC3B,SAAS,0BAA0B;AAEnC,SAAS,gBAAgB,kBAAkB;AAIpC,MAAM,oBAAgE;AAAA,EACjE,QAAQ,IAAI,mBAA8B;AAAA,EAC1C,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACA,aAAuB,CAAC;AAAA,EACxB,SAAS;AAAA,EACT,UAAU;AAAA,EACV;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,SAAK,QAAQ;AACb,SAAK,kBAAkB;AACvB,SAAK,oBAAoB;AAEzB,SAAK,oBAAoB,WAAW;AAAA,EACtC;AAAA;AAAA,EAGA,SAAS,MAAc;AACrB,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,SAAK,UAAU;AACf,QAAI,KAAK,OAAO,SAAS,KAAK,kBAAmB;AAEjD,WAAO,MAAM;AACX,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,OAAO,UAAU,EAAG;AAExB,UAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAM,MAAM,OAAO,MAAM;AACzB,UAAI;AACJ,UAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,kBAAU,IAAI,CAAC;AAAA,MACjB,OAAO;AACL,kBAAU;AAAA,MACZ;AAEA,WAAK,WAAW;AAEhB,UAAI,KAAK,QAAQ,UAAU,KAAK,iBAAiB;AAC/C,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AACzE,aAAK,UAAU;AAAA,MACjB;AAEA,UAAI,OAAO,QAAS,UAAU;AAC5B,aAAK,SAAS,KAAK,OAAO,MAAM,IAAK,CAAC,CAAC;AAAA,MACzC,OAAO;AACL,aAAK,SAAS,KAAK,OAChB,MAAM,KAAK,IAAI,GAAG,KAAK,OAAO,QAAQ,GAAG,CAAC,IAAI,IAAI,MAAM,EACxD,UAAU;AAAA,MACf;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,UAAU,KAAK,SAAS;AAC/B,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,QAAQ;AACV,YAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAI,MAAM,QAAQ,OAAO,CAAC,CAAC,GAAG;AAC5B,eAAK,WAAW,OAAO,IAAI,CAAC,QAAQ,IAAI,CAAC,CAAC,EAAE,KAAK,GAAG;AAAA,QACtD,OAAO;AACL,eAAK,WAAW,OAAO,KAAK,GAAG;AAAA,QACjC;AAAA,MACF;AAEA,UAAI,KAAK,SAAS;AAChB,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AAAA,MAC3E;AAEA,WAAK,oBAAoB,WAAW;AAAA,IACtC;AAEA,SAAK,SAAS;AACd,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM;AACX,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAyB;AAC5C,WAAO;AAAA,EACT;AACF;AAEO,MAAM,+BAA+B,eAAe;AAAA,EACzD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,QAAQ;AACN,UAAM,MAAM;AACZ,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;AAEO,MAAM,2BAA2B,WAAW;AAAA,EACjD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,WAAW;AACT,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/tokenize/token_stream.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AsyncIterableQueue, shortuuid } from '../utils.js';\nimport type { TokenData } from './tokenizer.js';\nimport { SentenceStream, WordStream } from './tokenizer.js';\n\ntype TokenizeFunc = (x: string) => string[] | [string, number, number][];\n\nexport class BufferedTokenStream implements AsyncIterableIterator<TokenData> {\n protected queue = new AsyncIterableQueue<TokenData>();\n protected closed = false;\n\n #func: TokenizeFunc;\n #minTokenLength: number;\n #minContextLength: number;\n #bufTokens: string[] = [];\n #inBuf = '';\n #outBuf = '';\n #currentSegmentId: string;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n this.#func = func;\n this.#minTokenLength = minTokenLength;\n this.#minContextLength = minContextLength;\n\n this.#currentSegmentId = shortuuid();\n }\n\n /** Push a string of text into the token stream */\n pushText(text: string) {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n this.#inBuf += text;\n if (this.#inBuf.length < this.#minContextLength) return;\n\n while (true) {\n const tokens = this.#func(this.#inBuf);\n if (tokens.length <= 1) break;\n\n if (this.#outBuf) this.#outBuf += ' ';\n\n const tok = tokens.shift()!;\n let tokText: string;\n if (Array.isArray(tok)) {\n tokText = tok[0];\n } else {\n tokText = tok;\n }\n\n this.#outBuf += tokText;\n\n if (this.#outBuf.length >= this.#minTokenLength) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n this.#outBuf = '';\n }\n\n if (typeof tok! !== 'string') {\n this.#inBuf = this.#inBuf.slice(tok![2]);\n } else {\n this.#inBuf = this.#inBuf\n .slice(Math.max(0, this.#inBuf.indexOf(tok)) + tok.length)\n .trimStart();\n }\n }\n }\n\n /** Flush the stream, causing it to process all pending text */\n flush() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.#inBuf || this.#outBuf) {\n const tokens = this.#func(this.#inBuf);\n if (tokens) {\n if (this.#outBuf) this.#outBuf += ' ';\n\n if (Array.isArray(tokens[0])) {\n this.#outBuf += tokens.map((tok) => tok[0]).join(' ');\n } else {\n this.#outBuf += tokens.join(' ');\n }\n }\n\n if (this.#outBuf) {\n this.queue.put({ token: this.#outBuf, segmentId: this.#currentSegmentId });\n }\n\n this.#currentSegmentId = shortuuid();\n }\n\n this.#inBuf = '';\n this.#outBuf = '';\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.flush();\n this.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.queue.next();\n }\n\n /** Close both the input and output of the token stream */\n close() {\n this.queue.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): BufferedTokenStream {\n return this;\n }\n}\n\nexport class BufferedSentenceStream extends SentenceStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n close() {\n super.close();\n this.#stream.close();\n }\n\n endInput() {\n this.#stream.endInput();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n\nexport class BufferedWordStream extends WordStream {\n #stream: BufferedTokenStream;\n\n constructor(func: TokenizeFunc, minTokenLength: number, minContextLength: number) {\n super();\n this.#stream = new BufferedTokenStream(func, minTokenLength, minContextLength);\n }\n\n pushText(text: string) {\n this.#stream.pushText(text);\n }\n\n flush() {\n this.#stream.flush();\n }\n\n endInput() {\n this.#stream.endInput();\n }\n\n close() {\n this.#stream.close();\n }\n\n next(): Promise<IteratorResult<TokenData>> {\n return this.#stream.next();\n }\n}\n"],"mappings":"AAGA,SAAS,oBAAoB,iBAAiB;AAE9C,SAAS,gBAAgB,kBAAkB;AAIpC,MAAM,oBAAgE;AAAA,EACjE,QAAQ,IAAI,mBAA8B;AAAA,EAC1C,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACA,aAAuB,CAAC;AAAA,EACxB,SAAS;AAAA,EACT,UAAU;AAAA,EACV;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,SAAK,QAAQ;AACb,SAAK,kBAAkB;AACvB,SAAK,oBAAoB;AAEzB,SAAK,oBAAoB,UAAU;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,MAAc;AACrB,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,SAAK,UAAU;AACf,QAAI,KAAK,OAAO,SAAS,KAAK,kBAAmB;AAEjD,WAAO,MAAM;AACX,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,OAAO,UAAU,EAAG;AAExB,UAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAM,MAAM,OAAO,MAAM;AACzB,UAAI;AACJ,UAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,kBAAU,IAAI,CAAC;AAAA,MACjB,OAAO;AACL,kBAAU;AAAA,MACZ;AAEA,WAAK,WAAW;AAEhB,UAAI,KAAK,QAAQ,UAAU,KAAK,iBAAiB;AAC/C,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AACzE,aAAK,UAAU;AAAA,MACjB;AAEA,UAAI,OAAO,QAAS,UAAU;AAC5B,aAAK,SAAS,KAAK,OAAO,MAAM,IAAK,CAAC,CAAC;AAAA,MACzC,OAAO;AACL,aAAK,SAAS,KAAK,OAChB,MAAM,KAAK,IAAI,GAAG,KAAK,OAAO,QAAQ,GAAG,CAAC,IAAI,IAAI,MAAM,EACxD,UAAU;AAAA,MACf;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,UAAU,KAAK,SAAS;AAC/B,YAAM,SAAS,KAAK,MAAM,KAAK,MAAM;AACrC,UAAI,QAAQ;AACV,YAAI,KAAK,QAAS,MAAK,WAAW;AAElC,YAAI,MAAM,QAAQ,OAAO,CAAC,CAAC,GAAG;AAC5B,eAAK,WAAW,OAAO,IAAI,CAAC,QAAQ,IAAI,CAAC,CAAC,EAAE,KAAK,GAAG;AAAA,QACtD,OAAO;AACL,eAAK,WAAW,OAAO,KAAK,GAAG;AAAA,QACjC;AAAA,MACF;AAEA,UAAI,KAAK,SAAS;AAChB,aAAK,MAAM,IAAI,EAAE,OAAO,KAAK,SAAS,WAAW,KAAK,kBAAkB,CAAC;AAAA,MAC3E;AAEA,WAAK,oBAAoB,UAAU;AAAA,IACrC;AAEA,SAAK,SAAS;AACd,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM;AACX,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAyB;AAC5C,WAAO;AAAA,EACT;AACF;AAEO,MAAM,+BAA+B,eAAe;AAAA,EACzD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,QAAQ;AACN,UAAM,MAAM;AACZ,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,WAAW;AACT,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;AAEO,MAAM,2BAA2B,WAAW;AAAA,EACjD;AAAA,EAEA,YAAY,MAAoB,gBAAwB,kBAA0B;AAChF,UAAM;AACN,SAAK,UAAU,IAAI,oBAAoB,MAAM,gBAAgB,gBAAgB;AAAA,EAC/E;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,QAAQ,SAAS,IAAI;AAAA,EAC5B;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,WAAW;AACT,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EAEA,QAAQ;AACN,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,OAA2C;AACzC,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B;AACF;","names":[]}
@@ -24,7 +24,6 @@ __export(transcription_exports, {
24
24
  module.exports = __toCommonJS(transcription_exports);
25
25
  var import_protocol = require("@livekit/protocol");
26
26
  var import_rtc_node = require("@livekit/rtc-node");
27
- var import_node_crypto = require("node:crypto");
28
27
  var import_node_events = require("node:events");
29
28
  var import_tokenize = require("./tokenize/index.cjs");
30
29
  var import_utils = require("./utils.cjs");
@@ -163,7 +162,7 @@ class TextAudioSynchronizer extends import_node_events.EventEmitter {
163
162
  if (audioData.pushedDuration > 0 && audioData.done) {
164
163
  realSpeed = this.#calcHyphens(textData.pushedText).length / audioData.pushedDuration;
165
164
  }
166
- const segId = "SG_" + (0, import_node_crypto.randomUUID)();
165
+ const segId = (0, import_utils.shortuuid)("SG_");
167
166
  const words = this.#opts.splitWords(sentence);
168
167
  const processedWords = [];
169
168
  const ogText = this.#playedText;
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/transcription.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { TranscriptionSegment } from '@livekit/protocol';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { randomUUID } from 'node:crypto';\nimport { EventEmitter } from 'node:events';\nimport { basic } from './tokenize/index.js';\nimport type { SentenceStream, SentenceTokenizer } from './tokenize/tokenizer.js';\nimport { AsyncIterableQueue, Future } from './utils.js';\n\n// standard speech rate in hyphens/ms\nconst STANDARD_SPEECH_RATE = 3830;\n\nexport interface TextSyncOptions {\n language: string;\n speed: number;\n newSentenceDelay: number;\n sentenceTokenizer: SentenceTokenizer;\n hyphenateWord: (word: string) => string[];\n splitWords: (words: string) => [string, number, number][];\n}\n\nexport const defaultTextSyncOptions: TextSyncOptions = {\n language: '',\n speed: 1,\n newSentenceDelay: 400,\n sentenceTokenizer: new basic.SentenceTokenizer(),\n hyphenateWord: basic.hyphenateWord,\n splitWords: basic.splitWords,\n};\n\ninterface AudioData {\n pushedDuration: number;\n done: boolean;\n}\n\ninterface TextData {\n sentenceStream: SentenceStream;\n pushedText: string;\n done: boolean;\n forwardedHyphens: number;\n forwardedSentences: number;\n}\n\ntype SyncCallbacks = {\n textUpdated: (text: TranscriptionSegment) => void;\n};\n\nexport class TextAudioSynchronizer extends (EventEmitter as new () => TypedEmitter<SyncCallbacks>) {\n #opts: TextSyncOptions;\n #speed: number;\n\n #closed = false;\n #interrupted = false;\n #closeFut = new Future();\n\n #playingSegIndex = -1;\n #finishedSegIndex = -1;\n\n #textQChanged = new AsyncIterableQueue<number>();\n #textQ: (TextData | undefined)[] = [];\n #audioQChanged = new AsyncIterableQueue<number>();\n #audioQ: (AudioData | undefined)[] = [];\n\n #playedText = '';\n #task?: Promise<void>;\n\n #audioData?: AudioData;\n #textData?: TextData;\n\n constructor(opts: TextSyncOptions) {\n super();\n\n this.#opts = opts;\n this.#speed = opts.speed * STANDARD_SPEECH_RATE;\n }\n\n pushAudio(frame: AudioFrame) {\n this.#checkNotClosed();\n if (!this.#audioData) {\n this.#audioData = { pushedDuration: 0, done: false };\n this.#audioQ.push(this.#audioData);\n this.#audioQChanged.put(1);\n }\n this.#audioData.pushedDuration += frame.samplesPerChannel / frame.sampleRate;\n }\n\n pushText(text: string) {\n this.#checkNotClosed();\n if (!this.#textData) {\n this.#textData = {\n sentenceStream: this.#opts.sentenceTokenizer.stream(),\n pushedText: '',\n done: false,\n forwardedHyphens: 0,\n forwardedSentences: 0,\n };\n this.#textQ.push(this.#textData);\n this.#textQChanged.put(1);\n }\n\n this.#textData.pushedText += text;\n this.#textData.sentenceStream.pushText(text);\n }\n\n markAudioSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#audioData) {\n // create empty audio data if none exists\n this.pushAudio(new AudioFrame(new Int16Array(), 24000, 1, 0));\n }\n\n this.#audioData!.done = true;\n this.#audioData = undefined;\n }\n\n markTextSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#textData) {\n this.pushText('');\n }\n\n this.#textData!.done = true;\n this.#textData?.sentenceStream.flush();\n this.#textData?.sentenceStream.close();\n this.#textData = undefined;\n }\n\n segmentPlayoutStarted() {\n this.#checkNotClosed();\n this.#playingSegIndex++;\n\n if (!this.#task) {\n this.#task = this.#mainLoop();\n }\n }\n\n segmentPlayoutFinished() {\n this.#checkNotClosed();\n this.#finishedSegIndex++;\n }\n\n get playedText(): string {\n return this.#playedText;\n }\n\n async close(interrupt: boolean) {\n if (this.#closed) {\n return;\n }\n this.#closed = true;\n this.#interrupted = interrupt;\n this.#closeFut.resolve();\n\n for (const textData of this.#textQ) {\n textData?.sentenceStream.close();\n }\n\n this.#textQ.push(undefined);\n this.#audioQ.push(undefined);\n this.#textQChanged.put(1);\n this.#audioQChanged.put(1);\n\n await this.#task;\n }\n\n async #mainLoop() {\n let segIndex = 0;\n let qDone = false;\n\n while (!qDone) {\n await this.#textQChanged.next();\n await this.#audioQChanged.next();\n\n while (this.#textQ.length && this.#audioQ.length) {\n const textData = this.#textQ.pop();\n const audioData = this.#audioQ.pop();\n\n if (!(textData && audioData)) {\n qDone = true;\n break;\n }\n\n // wait for segment to start playing\n while (!this.#closed) {\n if (this.#playingSegIndex >= segIndex) break;\n await this.#sleepIfNotClosed(125);\n }\n\n const sentenceStream = textData.sentenceStream;\n const forwardStartTime = Date.now();\n\n for await (const ev of sentenceStream) {\n await this.#syncSentence(segIndex, forwardStartTime, textData, audioData, ev.token);\n }\n\n segIndex++;\n }\n }\n }\n\n async #syncSentence(\n segIndex: number,\n segStartTime: number,\n textData: TextData,\n audioData: AudioData,\n sentence: string,\n ) {\n let realSpeed: number | undefined;\n if (audioData.pushedDuration > 0 && audioData.done) {\n realSpeed = this.#calcHyphens(textData.pushedText).length / audioData.pushedDuration;\n }\n\n const segId = 'SG_' + randomUUID();\n const words = this.#opts.splitWords(sentence);\n const processedWords: string[] = [];\n\n const ogText = this.#playedText;\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n for (const [word, _, end] of words) {\n if (segIndex <= this.#finishedSegIndex) break;\n if (this.#interrupted) return;\n\n const wordHyphens = this.#opts.hyphenateWord(word).length;\n processedWords.push(word);\n\n const elapsed = Date.now() - segStartTime;\n const text = sentence.slice(0, end); // TODO: rstrip punctuations\n\n let speed = this.#speed;\n let delay: number;\n if (realSpeed) {\n speed = realSpeed;\n const estimatedPausesMs = textData.forwardedSentences * this.#opts.newSentenceDelay;\n const hyphPauses = estimatedPausesMs * speed;\n const targetHyphens = Math.round(speed * elapsed);\n const dt = targetHyphens - textData.forwardedHyphens - hyphPauses;\n const toWaitHyphens = Math.max(0, wordHyphens - dt);\n delay = toWaitHyphens / speed;\n } else {\n delay = wordHyphens / speed;\n }\n\n const firstDelay = Math.min(delay / 2, 2 / speed);\n await this.#sleepIfNotClosed(firstDelay * 1000000);\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: text,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: false,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${text}`;\n await this.#sleepIfNotClosed((delay - firstDelay) * 1000000);\n textData.forwardedHyphens += wordHyphens;\n }\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: sentence,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: true,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${sentence}`;\n\n await this.#sleepIfNotClosed(this.#opts.newSentenceDelay);\n textData.forwardedSentences++;\n }\n\n async #sleepIfNotClosed(delay: number) {\n await Promise.race([\n this.#closeFut.await,\n new Promise((resolve) => setTimeout(resolve, delay)),\n ]);\n }\n\n #calcHyphens(text: string): string[] {\n const hyphens: string[] = [];\n const words = this.#opts.splitWords(text);\n for (const word of words) {\n const n = this.#opts.hyphenateWord(word[0]);\n hyphens.push(...n);\n }\n return hyphens;\n }\n\n #checkNotClosed() {\n if (this.#closed) {\n throw new Error('TextAudioSynchronizer is closed');\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAqC;AACrC,sBAA2B;AAE3B,yBAA2B;AAC3B,yBAA6B;AAC7B,sBAAsB;AAEtB,mBAA2C;AAG3C,MAAM,uBAAuB;AAWtB,MAAM,yBAA0C;AAAA,EACrD,UAAU;AAAA,EACV,OAAO;AAAA,EACP,kBAAkB;AAAA,EAClB,mBAAmB,IAAI,sBAAM,kBAAkB;AAAA,EAC/C,eAAe,sBAAM;AAAA,EACrB,YAAY,sBAAM;AACpB;AAmBO,MAAM,8BAA+B,gCAAuD;AAAA,EACjG;AAAA,EACA;AAAA,EAEA,UAAU;AAAA,EACV,eAAe;AAAA,EACf,YAAY,IAAI,oBAAO;AAAA,EAEvB,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EAEpB,gBAAgB,IAAI,gCAA2B;AAAA,EAC/C,SAAmC,CAAC;AAAA,EACpC,iBAAiB,IAAI,gCAA2B;AAAA,EAChD,UAAqC,CAAC;AAAA,EAEtC,cAAc;AAAA,EACd;AAAA,EAEA;AAAA,EACA;AAAA,EAEA,YAAY,MAAuB;AACjC,UAAM;AAEN,SAAK,QAAQ;AACb,SAAK,SAAS,KAAK,QAAQ;AAAA,EAC7B;AAAA,EAEA,UAAU,OAAmB;AAC3B,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa,EAAE,gBAAgB,GAAG,MAAM,MAAM;AACnD,WAAK,QAAQ,KAAK,KAAK,UAAU;AACjC,WAAK,eAAe,IAAI,CAAC;AAAA,IAC3B;AACA,SAAK,WAAW,kBAAkB,MAAM,oBAAoB,MAAM;AAAA,EACpE;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,YAAY;AAAA,QACf,gBAAgB,KAAK,MAAM,kBAAkB,OAAO;AAAA,QACpD,YAAY;AAAA,QACZ,MAAM;AAAA,QACN,kBAAkB;AAAA,QAClB,oBAAoB;AAAA,MACtB;AACA,WAAK,OAAO,KAAK,KAAK,SAAS;AAC/B,WAAK,cAAc,IAAI,CAAC;AAAA,IAC1B;AAEA,SAAK,UAAU,cAAc;AAC7B,SAAK,UAAU,eAAe,SAAS,IAAI;AAAA,EAC7C;AAAA,EAEA,sBAAsB;AACpB,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AAEpB,WAAK,UAAU,IAAI,2BAAW,IAAI,WAAW,GAAG,MAAO,GAAG,CAAC,CAAC;AAAA,IAC9D;AAEA,SAAK,WAAY,OAAO;AACxB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,qBAAqB;AAvHvB;AAwHI,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,SAAS,EAAE;AAAA,IAClB;AAEA,SAAK,UAAW,OAAO;AACvB,eAAK,cAAL,mBAAgB,eAAe;AAC/B,eAAK,cAAL,mBAAgB,eAAe;AAC/B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,wBAAwB;AACtB,SAAK,gBAAgB;AACrB,SAAK;AAEL,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,QAAQ,KAAK,UAAU;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,yBAAyB;AACvB,SAAK,gBAAgB;AACrB,SAAK;AAAA,EACP;AAAA,EAEA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,MAAM,WAAoB;AAC9B,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AACA,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,UAAU,QAAQ;AAEvB,eAAW,YAAY,KAAK,QAAQ;AAClC,2CAAU,eAAe;AAAA,IAC3B;AAEA,SAAK,OAAO,KAAK,MAAS;AAC1B,SAAK,QAAQ,KAAK,MAAS;AAC3B,SAAK,cAAc,IAAI,CAAC;AACxB,SAAK,eAAe,IAAI,CAAC;AAEzB,UAAM,KAAK;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,WAAW;AACf,QAAI,QAAQ;AAEZ,WAAO,CAAC,OAAO;AACb,YAAM,KAAK,cAAc,KAAK;AAC9B,YAAM,KAAK,eAAe,KAAK;AAE/B,aAAO,KAAK,OAAO,UAAU,KAAK,QAAQ,QAAQ;AAChD,cAAM,WAAW,KAAK,OAAO,IAAI;AACjC,cAAM,YAAY,KAAK,QAAQ,IAAI;AAEnC,YAAI,EAAE,YAAY,YAAY;AAC5B,kBAAQ;AACR;AAAA,QACF;AAGA,eAAO,CAAC,KAAK,SAAS;AACpB,cAAI,KAAK,oBAAoB,SAAU;AACvC,gBAAM,KAAK,kBAAkB,GAAG;AAAA,QAClC;AAEA,cAAM,iBAAiB,SAAS;AAChC,cAAM,mBAAmB,KAAK,IAAI;AAElC,yBAAiB,MAAM,gBAAgB;AACrC,gBAAM,KAAK,cAAc,UAAU,kBAAkB,UAAU,WAAW,GAAG,KAAK;AAAA,QACpF;AAEA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,cACJ,UACA,cACA,UACA,WACA,UACA;AACA,QAAI;AACJ,QAAI,UAAU,iBAAiB,KAAK,UAAU,MAAM;AAClD,kBAAY,KAAK,aAAa,SAAS,UAAU,EAAE,SAAS,UAAU;AAAA,IACxE;AAEA,UAAM,QAAQ,YAAQ,+BAAW;AACjC,UAAM,QAAQ,KAAK,MAAM,WAAW,QAAQ;AAC5C,UAAM,iBAA2B,CAAC;AAElC,UAAM,SAAS,KAAK;AAEpB,eAAW,CAAC,MAAM,GAAG,GAAG,KAAK,OAAO;AAClC,UAAI,YAAY,KAAK,kBAAmB;AACxC,UAAI,KAAK,aAAc;AAEvB,YAAM,cAAc,KAAK,MAAM,cAAc,IAAI,EAAE;AACnD,qBAAe,KAAK,IAAI;AAExB,YAAM,UAAU,KAAK,IAAI,IAAI;AAC7B,YAAM,OAAO,SAAS,MAAM,GAAG,GAAG;AAElC,UAAI,QAAQ,KAAK;AACjB,UAAI;AACJ,UAAI,WAAW;AACb,gBAAQ;AACR,cAAM,oBAAoB,SAAS,qBAAqB,KAAK,MAAM;AACnE,cAAM,aAAa,oBAAoB;AACvC,cAAM,gBAAgB,KAAK,MAAM,QAAQ,OAAO;AAChD,cAAM,KAAK,gBAAgB,SAAS,mBAAmB;AACvD,cAAM,gBAAgB,KAAK,IAAI,GAAG,cAAc,EAAE;AAClD,gBAAQ,gBAAgB;AAAA,MAC1B,OAAO;AACL,gBAAQ,cAAc;AAAA,MACxB;AAEA,YAAM,aAAa,KAAK,IAAI,QAAQ,GAAG,IAAI,KAAK;AAChD,YAAM,KAAK,kBAAkB,aAAa,GAAO;AAEjD,WAAK;AAAA,QACH;AAAA,QACA,IAAI,qCAAqB;AAAA,UACvB,IAAI;AAAA,UACJ;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,OAAO;AAAA,UACP,UAAU,KAAK,MAAM;AAAA,QACvB,CAAC;AAAA,MACH;AAEA,WAAK,cAAc,GAAG,MAAM,IAAI,IAAI;AACpC,YAAM,KAAK,mBAAmB,QAAQ,cAAc,GAAO;AAC3D,eAAS,oBAAoB;AAAA,IAC/B;AAEA,SAAK;AAAA,MACH;AAAA,MACA,IAAI,qCAAqB;AAAA,QACvB,IAAI;AAAA,QACJ,MAAM;AAAA,QACN,WAAW,OAAO,CAAC;AAAA,QACnB,SAAS,OAAO,CAAC;AAAA,QACjB,OAAO;AAAA,QACP,UAAU,KAAK,MAAM;AAAA,MACvB,CAAC;AAAA,IACH;AAEA,SAAK,cAAc,GAAG,MAAM,IAAI,QAAQ;AAExC,UAAM,KAAK,kBAAkB,KAAK,MAAM,gBAAgB;AACxD,aAAS;AAAA,EACX;AAAA,EAEA,MAAM,kBAAkB,OAAe;AACrC,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,UAAU;AAAA,MACf,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,KAAK,CAAC;AAAA,IACrD,CAAC;AAAA,EACH;AAAA,EAEA,aAAa,MAAwB;AACnC,UAAM,UAAoB,CAAC;AAC3B,UAAM,QAAQ,KAAK,MAAM,WAAW,IAAI;AACxC,eAAW,QAAQ,OAAO;AACxB,YAAM,IAAI,KAAK,MAAM,cAAc,KAAK,CAAC,CAAC;AAC1C,cAAQ,KAAK,GAAG,CAAC;AAAA,IACnB;AACA,WAAO;AAAA,EACT;AAAA,EAEA,kBAAkB;AAChB,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAAA,EACF;AACF;","names":[]}
1
+ {"version":3,"sources":["../src/transcription.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { TranscriptionSegment } from '@livekit/protocol';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport { basic } from './tokenize/index.js';\nimport type { SentenceStream, SentenceTokenizer } from './tokenize/tokenizer.js';\nimport { AsyncIterableQueue, Future, shortuuid } from './utils.js';\n\n// standard speech rate in hyphens/ms\nconst STANDARD_SPEECH_RATE = 3830;\n\nexport interface TextSyncOptions {\n language: string;\n speed: number;\n newSentenceDelay: number;\n sentenceTokenizer: SentenceTokenizer;\n hyphenateWord: (word: string) => string[];\n splitWords: (words: string) => [string, number, number][];\n}\n\nexport const defaultTextSyncOptions: TextSyncOptions = {\n language: '',\n speed: 1,\n newSentenceDelay: 400,\n sentenceTokenizer: new basic.SentenceTokenizer(),\n hyphenateWord: basic.hyphenateWord,\n splitWords: basic.splitWords,\n};\n\ninterface AudioData {\n pushedDuration: number;\n done: boolean;\n}\n\ninterface TextData {\n sentenceStream: SentenceStream;\n pushedText: string;\n done: boolean;\n forwardedHyphens: number;\n forwardedSentences: number;\n}\n\ntype SyncCallbacks = {\n textUpdated: (text: TranscriptionSegment) => void;\n};\n\nexport class TextAudioSynchronizer extends (EventEmitter as new () => TypedEmitter<SyncCallbacks>) {\n #opts: TextSyncOptions;\n #speed: number;\n\n #closed = false;\n #interrupted = false;\n #closeFut = new Future();\n\n #playingSegIndex = -1;\n #finishedSegIndex = -1;\n\n #textQChanged = new AsyncIterableQueue<number>();\n #textQ: (TextData | undefined)[] = [];\n #audioQChanged = new AsyncIterableQueue<number>();\n #audioQ: (AudioData | undefined)[] = [];\n\n #playedText = '';\n #task?: Promise<void>;\n\n #audioData?: AudioData;\n #textData?: TextData;\n\n constructor(opts: TextSyncOptions) {\n super();\n\n this.#opts = opts;\n this.#speed = opts.speed * STANDARD_SPEECH_RATE;\n }\n\n pushAudio(frame: AudioFrame) {\n this.#checkNotClosed();\n if (!this.#audioData) {\n this.#audioData = { pushedDuration: 0, done: false };\n this.#audioQ.push(this.#audioData);\n this.#audioQChanged.put(1);\n }\n this.#audioData.pushedDuration += frame.samplesPerChannel / frame.sampleRate;\n }\n\n pushText(text: string) {\n this.#checkNotClosed();\n if (!this.#textData) {\n this.#textData = {\n sentenceStream: this.#opts.sentenceTokenizer.stream(),\n pushedText: '',\n done: false,\n forwardedHyphens: 0,\n forwardedSentences: 0,\n };\n this.#textQ.push(this.#textData);\n this.#textQChanged.put(1);\n }\n\n this.#textData.pushedText += text;\n this.#textData.sentenceStream.pushText(text);\n }\n\n markAudioSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#audioData) {\n // create empty audio data if none exists\n this.pushAudio(new AudioFrame(new Int16Array(), 24000, 1, 0));\n }\n\n this.#audioData!.done = true;\n this.#audioData = undefined;\n }\n\n markTextSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#textData) {\n this.pushText('');\n }\n\n this.#textData!.done = true;\n this.#textData?.sentenceStream.flush();\n this.#textData?.sentenceStream.close();\n this.#textData = undefined;\n }\n\n segmentPlayoutStarted() {\n this.#checkNotClosed();\n this.#playingSegIndex++;\n\n if (!this.#task) {\n this.#task = this.#mainLoop();\n }\n }\n\n segmentPlayoutFinished() {\n this.#checkNotClosed();\n this.#finishedSegIndex++;\n }\n\n get playedText(): string {\n return this.#playedText;\n }\n\n async close(interrupt: boolean) {\n if (this.#closed) {\n return;\n }\n this.#closed = true;\n this.#interrupted = interrupt;\n this.#closeFut.resolve();\n\n for (const textData of this.#textQ) {\n textData?.sentenceStream.close();\n }\n\n this.#textQ.push(undefined);\n this.#audioQ.push(undefined);\n this.#textQChanged.put(1);\n this.#audioQChanged.put(1);\n\n await this.#task;\n }\n\n async #mainLoop() {\n let segIndex = 0;\n let qDone = false;\n\n while (!qDone) {\n await this.#textQChanged.next();\n await this.#audioQChanged.next();\n\n while (this.#textQ.length && this.#audioQ.length) {\n const textData = this.#textQ.pop();\n const audioData = this.#audioQ.pop();\n\n if (!(textData && audioData)) {\n qDone = true;\n break;\n }\n\n // wait for segment to start playing\n while (!this.#closed) {\n if (this.#playingSegIndex >= segIndex) break;\n await this.#sleepIfNotClosed(125);\n }\n\n const sentenceStream = textData.sentenceStream;\n const forwardStartTime = Date.now();\n\n for await (const ev of sentenceStream) {\n await this.#syncSentence(segIndex, forwardStartTime, textData, audioData, ev.token);\n }\n\n segIndex++;\n }\n }\n }\n\n async #syncSentence(\n segIndex: number,\n segStartTime: number,\n textData: TextData,\n audioData: AudioData,\n sentence: string,\n ) {\n let realSpeed: number | undefined;\n if (audioData.pushedDuration > 0 && audioData.done) {\n realSpeed = this.#calcHyphens(textData.pushedText).length / audioData.pushedDuration;\n }\n\n const segId = shortuuid('SG_');\n const words = this.#opts.splitWords(sentence);\n const processedWords: string[] = [];\n\n const ogText = this.#playedText;\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n for (const [word, _, end] of words) {\n if (segIndex <= this.#finishedSegIndex) break;\n if (this.#interrupted) return;\n\n const wordHyphens = this.#opts.hyphenateWord(word).length;\n processedWords.push(word);\n\n const elapsed = Date.now() - segStartTime;\n const text = sentence.slice(0, end); // TODO: rstrip punctuations\n\n let speed = this.#speed;\n let delay: number;\n if (realSpeed) {\n speed = realSpeed;\n const estimatedPausesMs = textData.forwardedSentences * this.#opts.newSentenceDelay;\n const hyphPauses = estimatedPausesMs * speed;\n const targetHyphens = Math.round(speed * elapsed);\n const dt = targetHyphens - textData.forwardedHyphens - hyphPauses;\n const toWaitHyphens = Math.max(0, wordHyphens - dt);\n delay = toWaitHyphens / speed;\n } else {\n delay = wordHyphens / speed;\n }\n\n const firstDelay = Math.min(delay / 2, 2 / speed);\n await this.#sleepIfNotClosed(firstDelay * 1000000);\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: text,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: false,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${text}`;\n await this.#sleepIfNotClosed((delay - firstDelay) * 1000000);\n textData.forwardedHyphens += wordHyphens;\n }\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: sentence,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: true,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${sentence}`;\n\n await this.#sleepIfNotClosed(this.#opts.newSentenceDelay);\n textData.forwardedSentences++;\n }\n\n async #sleepIfNotClosed(delay: number) {\n await Promise.race([\n this.#closeFut.await,\n new Promise((resolve) => setTimeout(resolve, delay)),\n ]);\n }\n\n #calcHyphens(text: string): string[] {\n const hyphens: string[] = [];\n const words = this.#opts.splitWords(text);\n for (const word of words) {\n const n = this.#opts.hyphenateWord(word[0]);\n hyphens.push(...n);\n }\n return hyphens;\n }\n\n #checkNotClosed() {\n if (this.#closed) {\n throw new Error('TextAudioSynchronizer is closed');\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAqC;AACrC,sBAA2B;AAE3B,yBAA6B;AAC7B,sBAAsB;AAEtB,mBAAsD;AAGtD,MAAM,uBAAuB;AAWtB,MAAM,yBAA0C;AAAA,EACrD,UAAU;AAAA,EACV,OAAO;AAAA,EACP,kBAAkB;AAAA,EAClB,mBAAmB,IAAI,sBAAM,kBAAkB;AAAA,EAC/C,eAAe,sBAAM;AAAA,EACrB,YAAY,sBAAM;AACpB;AAmBO,MAAM,8BAA+B,gCAAuD;AAAA,EACjG;AAAA,EACA;AAAA,EAEA,UAAU;AAAA,EACV,eAAe;AAAA,EACf,YAAY,IAAI,oBAAO;AAAA,EAEvB,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EAEpB,gBAAgB,IAAI,gCAA2B;AAAA,EAC/C,SAAmC,CAAC;AAAA,EACpC,iBAAiB,IAAI,gCAA2B;AAAA,EAChD,UAAqC,CAAC;AAAA,EAEtC,cAAc;AAAA,EACd;AAAA,EAEA;AAAA,EACA;AAAA,EAEA,YAAY,MAAuB;AACjC,UAAM;AAEN,SAAK,QAAQ;AACb,SAAK,SAAS,KAAK,QAAQ;AAAA,EAC7B;AAAA,EAEA,UAAU,OAAmB;AAC3B,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa,EAAE,gBAAgB,GAAG,MAAM,MAAM;AACnD,WAAK,QAAQ,KAAK,KAAK,UAAU;AACjC,WAAK,eAAe,IAAI,CAAC;AAAA,IAC3B;AACA,SAAK,WAAW,kBAAkB,MAAM,oBAAoB,MAAM;AAAA,EACpE;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,YAAY;AAAA,QACf,gBAAgB,KAAK,MAAM,kBAAkB,OAAO;AAAA,QACpD,YAAY;AAAA,QACZ,MAAM;AAAA,QACN,kBAAkB;AAAA,QAClB,oBAAoB;AAAA,MACtB;AACA,WAAK,OAAO,KAAK,KAAK,SAAS;AAC/B,WAAK,cAAc,IAAI,CAAC;AAAA,IAC1B;AAEA,SAAK,UAAU,cAAc;AAC7B,SAAK,UAAU,eAAe,SAAS,IAAI;AAAA,EAC7C;AAAA,EAEA,sBAAsB;AACpB,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AAEpB,WAAK,UAAU,IAAI,2BAAW,IAAI,WAAW,GAAG,MAAO,GAAG,CAAC,CAAC;AAAA,IAC9D;AAEA,SAAK,WAAY,OAAO;AACxB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,qBAAqB;AAtHvB;AAuHI,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,SAAS,EAAE;AAAA,IAClB;AAEA,SAAK,UAAW,OAAO;AACvB,eAAK,cAAL,mBAAgB,eAAe;AAC/B,eAAK,cAAL,mBAAgB,eAAe;AAC/B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,wBAAwB;AACtB,SAAK,gBAAgB;AACrB,SAAK;AAEL,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,QAAQ,KAAK,UAAU;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,yBAAyB;AACvB,SAAK,gBAAgB;AACrB,SAAK;AAAA,EACP;AAAA,EAEA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,MAAM,WAAoB;AAC9B,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AACA,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,UAAU,QAAQ;AAEvB,eAAW,YAAY,KAAK,QAAQ;AAClC,2CAAU,eAAe;AAAA,IAC3B;AAEA,SAAK,OAAO,KAAK,MAAS;AAC1B,SAAK,QAAQ,KAAK,MAAS;AAC3B,SAAK,cAAc,IAAI,CAAC;AACxB,SAAK,eAAe,IAAI,CAAC;AAEzB,UAAM,KAAK;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,WAAW;AACf,QAAI,QAAQ;AAEZ,WAAO,CAAC,OAAO;AACb,YAAM,KAAK,cAAc,KAAK;AAC9B,YAAM,KAAK,eAAe,KAAK;AAE/B,aAAO,KAAK,OAAO,UAAU,KAAK,QAAQ,QAAQ;AAChD,cAAM,WAAW,KAAK,OAAO,IAAI;AACjC,cAAM,YAAY,KAAK,QAAQ,IAAI;AAEnC,YAAI,EAAE,YAAY,YAAY;AAC5B,kBAAQ;AACR;AAAA,QACF;AAGA,eAAO,CAAC,KAAK,SAAS;AACpB,cAAI,KAAK,oBAAoB,SAAU;AACvC,gBAAM,KAAK,kBAAkB,GAAG;AAAA,QAClC;AAEA,cAAM,iBAAiB,SAAS;AAChC,cAAM,mBAAmB,KAAK,IAAI;AAElC,yBAAiB,MAAM,gBAAgB;AACrC,gBAAM,KAAK,cAAc,UAAU,kBAAkB,UAAU,WAAW,GAAG,KAAK;AAAA,QACpF;AAEA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,cACJ,UACA,cACA,UACA,WACA,UACA;AACA,QAAI;AACJ,QAAI,UAAU,iBAAiB,KAAK,UAAU,MAAM;AAClD,kBAAY,KAAK,aAAa,SAAS,UAAU,EAAE,SAAS,UAAU;AAAA,IACxE;AAEA,UAAM,YAAQ,wBAAU,KAAK;AAC7B,UAAM,QAAQ,KAAK,MAAM,WAAW,QAAQ;AAC5C,UAAM,iBAA2B,CAAC;AAElC,UAAM,SAAS,KAAK;AAEpB,eAAW,CAAC,MAAM,GAAG,GAAG,KAAK,OAAO;AAClC,UAAI,YAAY,KAAK,kBAAmB;AACxC,UAAI,KAAK,aAAc;AAEvB,YAAM,cAAc,KAAK,MAAM,cAAc,IAAI,EAAE;AACnD,qBAAe,KAAK,IAAI;AAExB,YAAM,UAAU,KAAK,IAAI,IAAI;AAC7B,YAAM,OAAO,SAAS,MAAM,GAAG,GAAG;AAElC,UAAI,QAAQ,KAAK;AACjB,UAAI;AACJ,UAAI,WAAW;AACb,gBAAQ;AACR,cAAM,oBAAoB,SAAS,qBAAqB,KAAK,MAAM;AACnE,cAAM,aAAa,oBAAoB;AACvC,cAAM,gBAAgB,KAAK,MAAM,QAAQ,OAAO;AAChD,cAAM,KAAK,gBAAgB,SAAS,mBAAmB;AACvD,cAAM,gBAAgB,KAAK,IAAI,GAAG,cAAc,EAAE;AAClD,gBAAQ,gBAAgB;AAAA,MAC1B,OAAO;AACL,gBAAQ,cAAc;AAAA,MACxB;AAEA,YAAM,aAAa,KAAK,IAAI,QAAQ,GAAG,IAAI,KAAK;AAChD,YAAM,KAAK,kBAAkB,aAAa,GAAO;AAEjD,WAAK;AAAA,QACH;AAAA,QACA,IAAI,qCAAqB;AAAA,UACvB,IAAI;AAAA,UACJ;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,OAAO;AAAA,UACP,UAAU,KAAK,MAAM;AAAA,QACvB,CAAC;AAAA,MACH;AAEA,WAAK,cAAc,GAAG,MAAM,IAAI,IAAI;AACpC,YAAM,KAAK,mBAAmB,QAAQ,cAAc,GAAO;AAC3D,eAAS,oBAAoB;AAAA,IAC/B;AAEA,SAAK;AAAA,MACH;AAAA,MACA,IAAI,qCAAqB;AAAA,QACvB,IAAI;AAAA,QACJ,MAAM;AAAA,QACN,WAAW,OAAO,CAAC;AAAA,QACnB,SAAS,OAAO,CAAC;AAAA,QACjB,OAAO;AAAA,QACP,UAAU,KAAK,MAAM;AAAA,MACvB,CAAC;AAAA,IACH;AAEA,SAAK,cAAc,GAAG,MAAM,IAAI,QAAQ;AAExC,UAAM,KAAK,kBAAkB,KAAK,MAAM,gBAAgB;AACxD,aAAS;AAAA,EACX;AAAA,EAEA,MAAM,kBAAkB,OAAe;AACrC,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,UAAU;AAAA,MACf,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,KAAK,CAAC;AAAA,IACrD,CAAC;AAAA,EACH;AAAA,EAEA,aAAa,MAAwB;AACnC,UAAM,UAAoB,CAAC;AAC3B,UAAM,QAAQ,KAAK,MAAM,WAAW,IAAI;AACxC,eAAW,QAAQ,OAAO;AACxB,YAAM,IAAI,KAAK,MAAM,cAAc,KAAK,CAAC,CAAC;AAC1C,cAAQ,KAAK,GAAG,CAAC;AAAA,IACnB;AACA,WAAO;AAAA,EACT;AAAA,EAEA,kBAAkB;AAChB,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAAA,EACF;AACF;","names":[]}
@@ -1 +1 @@
1
- {"version":3,"file":"transcription.d.ts","sourceRoot":"","sources":["../src/transcription.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,oBAAoB,EAAE,MAAM,mBAAmB,CAAC;AACzD,OAAO,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAC/C,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAIhF,OAAO,KAAK,EAAkB,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AAMjF,MAAM,WAAW,eAAe;IAC9B,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,MAAM,CAAC;IACd,gBAAgB,EAAE,MAAM,CAAC;IACzB,iBAAiB,EAAE,iBAAiB,CAAC;IACrC,aAAa,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,MAAM,EAAE,CAAC;IAC1C,UAAU,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;CAC3D;AAED,eAAO,MAAM,sBAAsB,EAAE,eAOpC,CAAC;AAeF,KAAK,aAAa,GAAG;IACnB,WAAW,EAAE,CAAC,IAAI,EAAE,oBAAoB,KAAK,IAAI,CAAC;CACnD,CAAC;oDAEoE,aAAa,aAAa,CAAC;AAAjG,qBAAa,qBAAsB,SAAQ,0BAAuD;;gBAsBpF,IAAI,EAAE,eAAe;IAOjC,SAAS,CAAC,KAAK,EAAE,UAAU;IAU3B,QAAQ,CAAC,IAAI,EAAE,MAAM;IAkBrB,mBAAmB;IAYnB,kBAAkB;IAalB,qBAAqB;IASrB,sBAAsB;IAKtB,IAAI,UAAU,IAAI,MAAM,CAEvB;IAEK,KAAK,CAAC,SAAS,EAAE,OAAO;CA6J/B"}
1
+ {"version":3,"file":"transcription.d.ts","sourceRoot":"","sources":["../src/transcription.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,oBAAoB,EAAE,MAAM,mBAAmB,CAAC;AACzD,OAAO,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAC/C,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAGhF,OAAO,KAAK,EAAkB,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AAMjF,MAAM,WAAW,eAAe;IAC9B,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,MAAM,CAAC;IACd,gBAAgB,EAAE,MAAM,CAAC;IACzB,iBAAiB,EAAE,iBAAiB,CAAC;IACrC,aAAa,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,MAAM,EAAE,CAAC;IAC1C,UAAU,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;CAC3D;AAED,eAAO,MAAM,sBAAsB,EAAE,eAOpC,CAAC;AAeF,KAAK,aAAa,GAAG;IACnB,WAAW,EAAE,CAAC,IAAI,EAAE,oBAAoB,KAAK,IAAI,CAAC;CACnD,CAAC;oDAEoE,aAAa,aAAa,CAAC;AAAjG,qBAAa,qBAAsB,SAAQ,0BAAuD;;gBAsBpF,IAAI,EAAE,eAAe;IAOjC,SAAS,CAAC,KAAK,EAAE,UAAU;IAU3B,QAAQ,CAAC,IAAI,EAAE,MAAM;IAkBrB,mBAAmB;IAYnB,kBAAkB;IAalB,qBAAqB;IASrB,sBAAsB;IAKtB,IAAI,UAAU,IAAI,MAAM,CAEvB;IAEK,KAAK,CAAC,SAAS,EAAE,OAAO;CA6J/B"}
@@ -1,9 +1,8 @@
1
1
  import { TranscriptionSegment } from "@livekit/protocol";
2
2
  import { AudioFrame } from "@livekit/rtc-node";
3
- import { randomUUID } from "node:crypto";
4
3
  import { EventEmitter } from "node:events";
5
4
  import { basic } from "./tokenize/index.js";
6
- import { AsyncIterableQueue, Future } from "./utils.js";
5
+ import { AsyncIterableQueue, Future, shortuuid } from "./utils.js";
7
6
  const STANDARD_SPEECH_RATE = 3830;
8
7
  const defaultTextSyncOptions = {
9
8
  language: "",
@@ -139,7 +138,7 @@ class TextAudioSynchronizer extends EventEmitter {
139
138
  if (audioData.pushedDuration > 0 && audioData.done) {
140
139
  realSpeed = this.#calcHyphens(textData.pushedText).length / audioData.pushedDuration;
141
140
  }
142
- const segId = "SG_" + randomUUID();
141
+ const segId = shortuuid("SG_");
143
142
  const words = this.#opts.splitWords(sentence);
144
143
  const processedWords = [];
145
144
  const ogText = this.#playedText;
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/transcription.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { TranscriptionSegment } from '@livekit/protocol';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { randomUUID } from 'node:crypto';\nimport { EventEmitter } from 'node:events';\nimport { basic } from './tokenize/index.js';\nimport type { SentenceStream, SentenceTokenizer } from './tokenize/tokenizer.js';\nimport { AsyncIterableQueue, Future } from './utils.js';\n\n// standard speech rate in hyphens/ms\nconst STANDARD_SPEECH_RATE = 3830;\n\nexport interface TextSyncOptions {\n language: string;\n speed: number;\n newSentenceDelay: number;\n sentenceTokenizer: SentenceTokenizer;\n hyphenateWord: (word: string) => string[];\n splitWords: (words: string) => [string, number, number][];\n}\n\nexport const defaultTextSyncOptions: TextSyncOptions = {\n language: '',\n speed: 1,\n newSentenceDelay: 400,\n sentenceTokenizer: new basic.SentenceTokenizer(),\n hyphenateWord: basic.hyphenateWord,\n splitWords: basic.splitWords,\n};\n\ninterface AudioData {\n pushedDuration: number;\n done: boolean;\n}\n\ninterface TextData {\n sentenceStream: SentenceStream;\n pushedText: string;\n done: boolean;\n forwardedHyphens: number;\n forwardedSentences: number;\n}\n\ntype SyncCallbacks = {\n textUpdated: (text: TranscriptionSegment) => void;\n};\n\nexport class TextAudioSynchronizer extends (EventEmitter as new () => TypedEmitter<SyncCallbacks>) {\n #opts: TextSyncOptions;\n #speed: number;\n\n #closed = false;\n #interrupted = false;\n #closeFut = new Future();\n\n #playingSegIndex = -1;\n #finishedSegIndex = -1;\n\n #textQChanged = new AsyncIterableQueue<number>();\n #textQ: (TextData | undefined)[] = [];\n #audioQChanged = new AsyncIterableQueue<number>();\n #audioQ: (AudioData | undefined)[] = [];\n\n #playedText = '';\n #task?: Promise<void>;\n\n #audioData?: AudioData;\n #textData?: TextData;\n\n constructor(opts: TextSyncOptions) {\n super();\n\n this.#opts = opts;\n this.#speed = opts.speed * STANDARD_SPEECH_RATE;\n }\n\n pushAudio(frame: AudioFrame) {\n this.#checkNotClosed();\n if (!this.#audioData) {\n this.#audioData = { pushedDuration: 0, done: false };\n this.#audioQ.push(this.#audioData);\n this.#audioQChanged.put(1);\n }\n this.#audioData.pushedDuration += frame.samplesPerChannel / frame.sampleRate;\n }\n\n pushText(text: string) {\n this.#checkNotClosed();\n if (!this.#textData) {\n this.#textData = {\n sentenceStream: this.#opts.sentenceTokenizer.stream(),\n pushedText: '',\n done: false,\n forwardedHyphens: 0,\n forwardedSentences: 0,\n };\n this.#textQ.push(this.#textData);\n this.#textQChanged.put(1);\n }\n\n this.#textData.pushedText += text;\n this.#textData.sentenceStream.pushText(text);\n }\n\n markAudioSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#audioData) {\n // create empty audio data if none exists\n this.pushAudio(new AudioFrame(new Int16Array(), 24000, 1, 0));\n }\n\n this.#audioData!.done = true;\n this.#audioData = undefined;\n }\n\n markTextSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#textData) {\n this.pushText('');\n }\n\n this.#textData!.done = true;\n this.#textData?.sentenceStream.flush();\n this.#textData?.sentenceStream.close();\n this.#textData = undefined;\n }\n\n segmentPlayoutStarted() {\n this.#checkNotClosed();\n this.#playingSegIndex++;\n\n if (!this.#task) {\n this.#task = this.#mainLoop();\n }\n }\n\n segmentPlayoutFinished() {\n this.#checkNotClosed();\n this.#finishedSegIndex++;\n }\n\n get playedText(): string {\n return this.#playedText;\n }\n\n async close(interrupt: boolean) {\n if (this.#closed) {\n return;\n }\n this.#closed = true;\n this.#interrupted = interrupt;\n this.#closeFut.resolve();\n\n for (const textData of this.#textQ) {\n textData?.sentenceStream.close();\n }\n\n this.#textQ.push(undefined);\n this.#audioQ.push(undefined);\n this.#textQChanged.put(1);\n this.#audioQChanged.put(1);\n\n await this.#task;\n }\n\n async #mainLoop() {\n let segIndex = 0;\n let qDone = false;\n\n while (!qDone) {\n await this.#textQChanged.next();\n await this.#audioQChanged.next();\n\n while (this.#textQ.length && this.#audioQ.length) {\n const textData = this.#textQ.pop();\n const audioData = this.#audioQ.pop();\n\n if (!(textData && audioData)) {\n qDone = true;\n break;\n }\n\n // wait for segment to start playing\n while (!this.#closed) {\n if (this.#playingSegIndex >= segIndex) break;\n await this.#sleepIfNotClosed(125);\n }\n\n const sentenceStream = textData.sentenceStream;\n const forwardStartTime = Date.now();\n\n for await (const ev of sentenceStream) {\n await this.#syncSentence(segIndex, forwardStartTime, textData, audioData, ev.token);\n }\n\n segIndex++;\n }\n }\n }\n\n async #syncSentence(\n segIndex: number,\n segStartTime: number,\n textData: TextData,\n audioData: AudioData,\n sentence: string,\n ) {\n let realSpeed: number | undefined;\n if (audioData.pushedDuration > 0 && audioData.done) {\n realSpeed = this.#calcHyphens(textData.pushedText).length / audioData.pushedDuration;\n }\n\n const segId = 'SG_' + randomUUID();\n const words = this.#opts.splitWords(sentence);\n const processedWords: string[] = [];\n\n const ogText = this.#playedText;\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n for (const [word, _, end] of words) {\n if (segIndex <= this.#finishedSegIndex) break;\n if (this.#interrupted) return;\n\n const wordHyphens = this.#opts.hyphenateWord(word).length;\n processedWords.push(word);\n\n const elapsed = Date.now() - segStartTime;\n const text = sentence.slice(0, end); // TODO: rstrip punctuations\n\n let speed = this.#speed;\n let delay: number;\n if (realSpeed) {\n speed = realSpeed;\n const estimatedPausesMs = textData.forwardedSentences * this.#opts.newSentenceDelay;\n const hyphPauses = estimatedPausesMs * speed;\n const targetHyphens = Math.round(speed * elapsed);\n const dt = targetHyphens - textData.forwardedHyphens - hyphPauses;\n const toWaitHyphens = Math.max(0, wordHyphens - dt);\n delay = toWaitHyphens / speed;\n } else {\n delay = wordHyphens / speed;\n }\n\n const firstDelay = Math.min(delay / 2, 2 / speed);\n await this.#sleepIfNotClosed(firstDelay * 1000000);\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: text,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: false,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${text}`;\n await this.#sleepIfNotClosed((delay - firstDelay) * 1000000);\n textData.forwardedHyphens += wordHyphens;\n }\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: sentence,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: true,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${sentence}`;\n\n await this.#sleepIfNotClosed(this.#opts.newSentenceDelay);\n textData.forwardedSentences++;\n }\n\n async #sleepIfNotClosed(delay: number) {\n await Promise.race([\n this.#closeFut.await,\n new Promise((resolve) => setTimeout(resolve, delay)),\n ]);\n }\n\n #calcHyphens(text: string): string[] {\n const hyphens: string[] = [];\n const words = this.#opts.splitWords(text);\n for (const word of words) {\n const n = this.#opts.hyphenateWord(word[0]);\n hyphens.push(...n);\n }\n return hyphens;\n }\n\n #checkNotClosed() {\n if (this.#closed) {\n throw new Error('TextAudioSynchronizer is closed');\n }\n }\n}\n"],"mappings":"AAGA,SAAS,4BAA4B;AACrC,SAAS,kBAAkB;AAE3B,SAAS,kBAAkB;AAC3B,SAAS,oBAAoB;AAC7B,SAAS,aAAa;AAEtB,SAAS,oBAAoB,cAAc;AAG3C,MAAM,uBAAuB;AAWtB,MAAM,yBAA0C;AAAA,EACrD,UAAU;AAAA,EACV,OAAO;AAAA,EACP,kBAAkB;AAAA,EAClB,mBAAmB,IAAI,MAAM,kBAAkB;AAAA,EAC/C,eAAe,MAAM;AAAA,EACrB,YAAY,MAAM;AACpB;AAmBO,MAAM,8BAA+B,aAAuD;AAAA,EACjG;AAAA,EACA;AAAA,EAEA,UAAU;AAAA,EACV,eAAe;AAAA,EACf,YAAY,IAAI,OAAO;AAAA,EAEvB,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EAEpB,gBAAgB,IAAI,mBAA2B;AAAA,EAC/C,SAAmC,CAAC;AAAA,EACpC,iBAAiB,IAAI,mBAA2B;AAAA,EAChD,UAAqC,CAAC;AAAA,EAEtC,cAAc;AAAA,EACd;AAAA,EAEA;AAAA,EACA;AAAA,EAEA,YAAY,MAAuB;AACjC,UAAM;AAEN,SAAK,QAAQ;AACb,SAAK,SAAS,KAAK,QAAQ;AAAA,EAC7B;AAAA,EAEA,UAAU,OAAmB;AAC3B,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa,EAAE,gBAAgB,GAAG,MAAM,MAAM;AACnD,WAAK,QAAQ,KAAK,KAAK,UAAU;AACjC,WAAK,eAAe,IAAI,CAAC;AAAA,IAC3B;AACA,SAAK,WAAW,kBAAkB,MAAM,oBAAoB,MAAM;AAAA,EACpE;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,YAAY;AAAA,QACf,gBAAgB,KAAK,MAAM,kBAAkB,OAAO;AAAA,QACpD,YAAY;AAAA,QACZ,MAAM;AAAA,QACN,kBAAkB;AAAA,QAClB,oBAAoB;AAAA,MACtB;AACA,WAAK,OAAO,KAAK,KAAK,SAAS;AAC/B,WAAK,cAAc,IAAI,CAAC;AAAA,IAC1B;AAEA,SAAK,UAAU,cAAc;AAC7B,SAAK,UAAU,eAAe,SAAS,IAAI;AAAA,EAC7C;AAAA,EAEA,sBAAsB;AACpB,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AAEpB,WAAK,UAAU,IAAI,WAAW,IAAI,WAAW,GAAG,MAAO,GAAG,CAAC,CAAC;AAAA,IAC9D;AAEA,SAAK,WAAY,OAAO;AACxB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,qBAAqB;AAvHvB;AAwHI,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,SAAS,EAAE;AAAA,IAClB;AAEA,SAAK,UAAW,OAAO;AACvB,eAAK,cAAL,mBAAgB,eAAe;AAC/B,eAAK,cAAL,mBAAgB,eAAe;AAC/B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,wBAAwB;AACtB,SAAK,gBAAgB;AACrB,SAAK;AAEL,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,QAAQ,KAAK,UAAU;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,yBAAyB;AACvB,SAAK,gBAAgB;AACrB,SAAK;AAAA,EACP;AAAA,EAEA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,MAAM,WAAoB;AAC9B,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AACA,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,UAAU,QAAQ;AAEvB,eAAW,YAAY,KAAK,QAAQ;AAClC,2CAAU,eAAe;AAAA,IAC3B;AAEA,SAAK,OAAO,KAAK,MAAS;AAC1B,SAAK,QAAQ,KAAK,MAAS;AAC3B,SAAK,cAAc,IAAI,CAAC;AACxB,SAAK,eAAe,IAAI,CAAC;AAEzB,UAAM,KAAK;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,WAAW;AACf,QAAI,QAAQ;AAEZ,WAAO,CAAC,OAAO;AACb,YAAM,KAAK,cAAc,KAAK;AAC9B,YAAM,KAAK,eAAe,KAAK;AAE/B,aAAO,KAAK,OAAO,UAAU,KAAK,QAAQ,QAAQ;AAChD,cAAM,WAAW,KAAK,OAAO,IAAI;AACjC,cAAM,YAAY,KAAK,QAAQ,IAAI;AAEnC,YAAI,EAAE,YAAY,YAAY;AAC5B,kBAAQ;AACR;AAAA,QACF;AAGA,eAAO,CAAC,KAAK,SAAS;AACpB,cAAI,KAAK,oBAAoB,SAAU;AACvC,gBAAM,KAAK,kBAAkB,GAAG;AAAA,QAClC;AAEA,cAAM,iBAAiB,SAAS;AAChC,cAAM,mBAAmB,KAAK,IAAI;AAElC,yBAAiB,MAAM,gBAAgB;AACrC,gBAAM,KAAK,cAAc,UAAU,kBAAkB,UAAU,WAAW,GAAG,KAAK;AAAA,QACpF;AAEA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,cACJ,UACA,cACA,UACA,WACA,UACA;AACA,QAAI;AACJ,QAAI,UAAU,iBAAiB,KAAK,UAAU,MAAM;AAClD,kBAAY,KAAK,aAAa,SAAS,UAAU,EAAE,SAAS,UAAU;AAAA,IACxE;AAEA,UAAM,QAAQ,QAAQ,WAAW;AACjC,UAAM,QAAQ,KAAK,MAAM,WAAW,QAAQ;AAC5C,UAAM,iBAA2B,CAAC;AAElC,UAAM,SAAS,KAAK;AAEpB,eAAW,CAAC,MAAM,GAAG,GAAG,KAAK,OAAO;AAClC,UAAI,YAAY,KAAK,kBAAmB;AACxC,UAAI,KAAK,aAAc;AAEvB,YAAM,cAAc,KAAK,MAAM,cAAc,IAAI,EAAE;AACnD,qBAAe,KAAK,IAAI;AAExB,YAAM,UAAU,KAAK,IAAI,IAAI;AAC7B,YAAM,OAAO,SAAS,MAAM,GAAG,GAAG;AAElC,UAAI,QAAQ,KAAK;AACjB,UAAI;AACJ,UAAI,WAAW;AACb,gBAAQ;AACR,cAAM,oBAAoB,SAAS,qBAAqB,KAAK,MAAM;AACnE,cAAM,aAAa,oBAAoB;AACvC,cAAM,gBAAgB,KAAK,MAAM,QAAQ,OAAO;AAChD,cAAM,KAAK,gBAAgB,SAAS,mBAAmB;AACvD,cAAM,gBAAgB,KAAK,IAAI,GAAG,cAAc,EAAE;AAClD,gBAAQ,gBAAgB;AAAA,MAC1B,OAAO;AACL,gBAAQ,cAAc;AAAA,MACxB;AAEA,YAAM,aAAa,KAAK,IAAI,QAAQ,GAAG,IAAI,KAAK;AAChD,YAAM,KAAK,kBAAkB,aAAa,GAAO;AAEjD,WAAK;AAAA,QACH;AAAA,QACA,IAAI,qBAAqB;AAAA,UACvB,IAAI;AAAA,UACJ;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,OAAO;AAAA,UACP,UAAU,KAAK,MAAM;AAAA,QACvB,CAAC;AAAA,MACH;AAEA,WAAK,cAAc,GAAG,MAAM,IAAI,IAAI;AACpC,YAAM,KAAK,mBAAmB,QAAQ,cAAc,GAAO;AAC3D,eAAS,oBAAoB;AAAA,IAC/B;AAEA,SAAK;AAAA,MACH;AAAA,MACA,IAAI,qBAAqB;AAAA,QACvB,IAAI;AAAA,QACJ,MAAM;AAAA,QACN,WAAW,OAAO,CAAC;AAAA,QACnB,SAAS,OAAO,CAAC;AAAA,QACjB,OAAO;AAAA,QACP,UAAU,KAAK,MAAM;AAAA,MACvB,CAAC;AAAA,IACH;AAEA,SAAK,cAAc,GAAG,MAAM,IAAI,QAAQ;AAExC,UAAM,KAAK,kBAAkB,KAAK,MAAM,gBAAgB;AACxD,aAAS;AAAA,EACX;AAAA,EAEA,MAAM,kBAAkB,OAAe;AACrC,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,UAAU;AAAA,MACf,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,KAAK,CAAC;AAAA,IACrD,CAAC;AAAA,EACH;AAAA,EAEA,aAAa,MAAwB;AACnC,UAAM,UAAoB,CAAC;AAC3B,UAAM,QAAQ,KAAK,MAAM,WAAW,IAAI;AACxC,eAAW,QAAQ,OAAO;AACxB,YAAM,IAAI,KAAK,MAAM,cAAc,KAAK,CAAC,CAAC;AAC1C,cAAQ,KAAK,GAAG,CAAC;AAAA,IACnB;AACA,WAAO;AAAA,EACT;AAAA,EAEA,kBAAkB;AAChB,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAAA,EACF;AACF;","names":[]}
1
+ {"version":3,"sources":["../src/transcription.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { TranscriptionSegment } from '@livekit/protocol';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport { basic } from './tokenize/index.js';\nimport type { SentenceStream, SentenceTokenizer } from './tokenize/tokenizer.js';\nimport { AsyncIterableQueue, Future, shortuuid } from './utils.js';\n\n// standard speech rate in hyphens/ms\nconst STANDARD_SPEECH_RATE = 3830;\n\nexport interface TextSyncOptions {\n language: string;\n speed: number;\n newSentenceDelay: number;\n sentenceTokenizer: SentenceTokenizer;\n hyphenateWord: (word: string) => string[];\n splitWords: (words: string) => [string, number, number][];\n}\n\nexport const defaultTextSyncOptions: TextSyncOptions = {\n language: '',\n speed: 1,\n newSentenceDelay: 400,\n sentenceTokenizer: new basic.SentenceTokenizer(),\n hyphenateWord: basic.hyphenateWord,\n splitWords: basic.splitWords,\n};\n\ninterface AudioData {\n pushedDuration: number;\n done: boolean;\n}\n\ninterface TextData {\n sentenceStream: SentenceStream;\n pushedText: string;\n done: boolean;\n forwardedHyphens: number;\n forwardedSentences: number;\n}\n\ntype SyncCallbacks = {\n textUpdated: (text: TranscriptionSegment) => void;\n};\n\nexport class TextAudioSynchronizer extends (EventEmitter as new () => TypedEmitter<SyncCallbacks>) {\n #opts: TextSyncOptions;\n #speed: number;\n\n #closed = false;\n #interrupted = false;\n #closeFut = new Future();\n\n #playingSegIndex = -1;\n #finishedSegIndex = -1;\n\n #textQChanged = new AsyncIterableQueue<number>();\n #textQ: (TextData | undefined)[] = [];\n #audioQChanged = new AsyncIterableQueue<number>();\n #audioQ: (AudioData | undefined)[] = [];\n\n #playedText = '';\n #task?: Promise<void>;\n\n #audioData?: AudioData;\n #textData?: TextData;\n\n constructor(opts: TextSyncOptions) {\n super();\n\n this.#opts = opts;\n this.#speed = opts.speed * STANDARD_SPEECH_RATE;\n }\n\n pushAudio(frame: AudioFrame) {\n this.#checkNotClosed();\n if (!this.#audioData) {\n this.#audioData = { pushedDuration: 0, done: false };\n this.#audioQ.push(this.#audioData);\n this.#audioQChanged.put(1);\n }\n this.#audioData.pushedDuration += frame.samplesPerChannel / frame.sampleRate;\n }\n\n pushText(text: string) {\n this.#checkNotClosed();\n if (!this.#textData) {\n this.#textData = {\n sentenceStream: this.#opts.sentenceTokenizer.stream(),\n pushedText: '',\n done: false,\n forwardedHyphens: 0,\n forwardedSentences: 0,\n };\n this.#textQ.push(this.#textData);\n this.#textQChanged.put(1);\n }\n\n this.#textData.pushedText += text;\n this.#textData.sentenceStream.pushText(text);\n }\n\n markAudioSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#audioData) {\n // create empty audio data if none exists\n this.pushAudio(new AudioFrame(new Int16Array(), 24000, 1, 0));\n }\n\n this.#audioData!.done = true;\n this.#audioData = undefined;\n }\n\n markTextSegmentEnd() {\n this.#checkNotClosed();\n\n if (!this.#textData) {\n this.pushText('');\n }\n\n this.#textData!.done = true;\n this.#textData?.sentenceStream.flush();\n this.#textData?.sentenceStream.close();\n this.#textData = undefined;\n }\n\n segmentPlayoutStarted() {\n this.#checkNotClosed();\n this.#playingSegIndex++;\n\n if (!this.#task) {\n this.#task = this.#mainLoop();\n }\n }\n\n segmentPlayoutFinished() {\n this.#checkNotClosed();\n this.#finishedSegIndex++;\n }\n\n get playedText(): string {\n return this.#playedText;\n }\n\n async close(interrupt: boolean) {\n if (this.#closed) {\n return;\n }\n this.#closed = true;\n this.#interrupted = interrupt;\n this.#closeFut.resolve();\n\n for (const textData of this.#textQ) {\n textData?.sentenceStream.close();\n }\n\n this.#textQ.push(undefined);\n this.#audioQ.push(undefined);\n this.#textQChanged.put(1);\n this.#audioQChanged.put(1);\n\n await this.#task;\n }\n\n async #mainLoop() {\n let segIndex = 0;\n let qDone = false;\n\n while (!qDone) {\n await this.#textQChanged.next();\n await this.#audioQChanged.next();\n\n while (this.#textQ.length && this.#audioQ.length) {\n const textData = this.#textQ.pop();\n const audioData = this.#audioQ.pop();\n\n if (!(textData && audioData)) {\n qDone = true;\n break;\n }\n\n // wait for segment to start playing\n while (!this.#closed) {\n if (this.#playingSegIndex >= segIndex) break;\n await this.#sleepIfNotClosed(125);\n }\n\n const sentenceStream = textData.sentenceStream;\n const forwardStartTime = Date.now();\n\n for await (const ev of sentenceStream) {\n await this.#syncSentence(segIndex, forwardStartTime, textData, audioData, ev.token);\n }\n\n segIndex++;\n }\n }\n }\n\n async #syncSentence(\n segIndex: number,\n segStartTime: number,\n textData: TextData,\n audioData: AudioData,\n sentence: string,\n ) {\n let realSpeed: number | undefined;\n if (audioData.pushedDuration > 0 && audioData.done) {\n realSpeed = this.#calcHyphens(textData.pushedText).length / audioData.pushedDuration;\n }\n\n const segId = shortuuid('SG_');\n const words = this.#opts.splitWords(sentence);\n const processedWords: string[] = [];\n\n const ogText = this.#playedText;\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n for (const [word, _, end] of words) {\n if (segIndex <= this.#finishedSegIndex) break;\n if (this.#interrupted) return;\n\n const wordHyphens = this.#opts.hyphenateWord(word).length;\n processedWords.push(word);\n\n const elapsed = Date.now() - segStartTime;\n const text = sentence.slice(0, end); // TODO: rstrip punctuations\n\n let speed = this.#speed;\n let delay: number;\n if (realSpeed) {\n speed = realSpeed;\n const estimatedPausesMs = textData.forwardedSentences * this.#opts.newSentenceDelay;\n const hyphPauses = estimatedPausesMs * speed;\n const targetHyphens = Math.round(speed * elapsed);\n const dt = targetHyphens - textData.forwardedHyphens - hyphPauses;\n const toWaitHyphens = Math.max(0, wordHyphens - dt);\n delay = toWaitHyphens / speed;\n } else {\n delay = wordHyphens / speed;\n }\n\n const firstDelay = Math.min(delay / 2, 2 / speed);\n await this.#sleepIfNotClosed(firstDelay * 1000000);\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: text,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: false,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${text}`;\n await this.#sleepIfNotClosed((delay - firstDelay) * 1000000);\n textData.forwardedHyphens += wordHyphens;\n }\n\n this.emit(\n 'textUpdated',\n new TranscriptionSegment({\n id: segId,\n text: sentence,\n startTime: BigInt(0),\n endTime: BigInt(0),\n final: true,\n language: this.#opts.language,\n }),\n );\n\n this.#playedText = `${ogText} ${sentence}`;\n\n await this.#sleepIfNotClosed(this.#opts.newSentenceDelay);\n textData.forwardedSentences++;\n }\n\n async #sleepIfNotClosed(delay: number) {\n await Promise.race([\n this.#closeFut.await,\n new Promise((resolve) => setTimeout(resolve, delay)),\n ]);\n }\n\n #calcHyphens(text: string): string[] {\n const hyphens: string[] = [];\n const words = this.#opts.splitWords(text);\n for (const word of words) {\n const n = this.#opts.hyphenateWord(word[0]);\n hyphens.push(...n);\n }\n return hyphens;\n }\n\n #checkNotClosed() {\n if (this.#closed) {\n throw new Error('TextAudioSynchronizer is closed');\n }\n }\n}\n"],"mappings":"AAGA,SAAS,4BAA4B;AACrC,SAAS,kBAAkB;AAE3B,SAAS,oBAAoB;AAC7B,SAAS,aAAa;AAEtB,SAAS,oBAAoB,QAAQ,iBAAiB;AAGtD,MAAM,uBAAuB;AAWtB,MAAM,yBAA0C;AAAA,EACrD,UAAU;AAAA,EACV,OAAO;AAAA,EACP,kBAAkB;AAAA,EAClB,mBAAmB,IAAI,MAAM,kBAAkB;AAAA,EAC/C,eAAe,MAAM;AAAA,EACrB,YAAY,MAAM;AACpB;AAmBO,MAAM,8BAA+B,aAAuD;AAAA,EACjG;AAAA,EACA;AAAA,EAEA,UAAU;AAAA,EACV,eAAe;AAAA,EACf,YAAY,IAAI,OAAO;AAAA,EAEvB,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EAEpB,gBAAgB,IAAI,mBAA2B;AAAA,EAC/C,SAAmC,CAAC;AAAA,EACpC,iBAAiB,IAAI,mBAA2B;AAAA,EAChD,UAAqC,CAAC;AAAA,EAEtC,cAAc;AAAA,EACd;AAAA,EAEA;AAAA,EACA;AAAA,EAEA,YAAY,MAAuB;AACjC,UAAM;AAEN,SAAK,QAAQ;AACb,SAAK,SAAS,KAAK,QAAQ;AAAA,EAC7B;AAAA,EAEA,UAAU,OAAmB;AAC3B,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa,EAAE,gBAAgB,GAAG,MAAM,MAAM;AACnD,WAAK,QAAQ,KAAK,KAAK,UAAU;AACjC,WAAK,eAAe,IAAI,CAAC;AAAA,IAC3B;AACA,SAAK,WAAW,kBAAkB,MAAM,oBAAoB,MAAM;AAAA,EACpE;AAAA,EAEA,SAAS,MAAc;AACrB,SAAK,gBAAgB;AACrB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,YAAY;AAAA,QACf,gBAAgB,KAAK,MAAM,kBAAkB,OAAO;AAAA,QACpD,YAAY;AAAA,QACZ,MAAM;AAAA,QACN,kBAAkB;AAAA,QAClB,oBAAoB;AAAA,MACtB;AACA,WAAK,OAAO,KAAK,KAAK,SAAS;AAC/B,WAAK,cAAc,IAAI,CAAC;AAAA,IAC1B;AAEA,SAAK,UAAU,cAAc;AAC7B,SAAK,UAAU,eAAe,SAAS,IAAI;AAAA,EAC7C;AAAA,EAEA,sBAAsB;AACpB,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AAEpB,WAAK,UAAU,IAAI,WAAW,IAAI,WAAW,GAAG,MAAO,GAAG,CAAC,CAAC;AAAA,IAC9D;AAEA,SAAK,WAAY,OAAO;AACxB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,qBAAqB;AAtHvB;AAuHI,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,SAAS,EAAE;AAAA,IAClB;AAEA,SAAK,UAAW,OAAO;AACvB,eAAK,cAAL,mBAAgB,eAAe;AAC/B,eAAK,cAAL,mBAAgB,eAAe;AAC/B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,wBAAwB;AACtB,SAAK,gBAAgB;AACrB,SAAK;AAEL,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,QAAQ,KAAK,UAAU;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,yBAAyB;AACvB,SAAK,gBAAgB;AACrB,SAAK;AAAA,EACP;AAAA,EAEA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,MAAM,WAAoB;AAC9B,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AACA,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,UAAU,QAAQ;AAEvB,eAAW,YAAY,KAAK,QAAQ;AAClC,2CAAU,eAAe;AAAA,IAC3B;AAEA,SAAK,OAAO,KAAK,MAAS;AAC1B,SAAK,QAAQ,KAAK,MAAS;AAC3B,SAAK,cAAc,IAAI,CAAC;AACxB,SAAK,eAAe,IAAI,CAAC;AAEzB,UAAM,KAAK;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,WAAW;AACf,QAAI,QAAQ;AAEZ,WAAO,CAAC,OAAO;AACb,YAAM,KAAK,cAAc,KAAK;AAC9B,YAAM,KAAK,eAAe,KAAK;AAE/B,aAAO,KAAK,OAAO,UAAU,KAAK,QAAQ,QAAQ;AAChD,cAAM,WAAW,KAAK,OAAO,IAAI;AACjC,cAAM,YAAY,KAAK,QAAQ,IAAI;AAEnC,YAAI,EAAE,YAAY,YAAY;AAC5B,kBAAQ;AACR;AAAA,QACF;AAGA,eAAO,CAAC,KAAK,SAAS;AACpB,cAAI,KAAK,oBAAoB,SAAU;AACvC,gBAAM,KAAK,kBAAkB,GAAG;AAAA,QAClC;AAEA,cAAM,iBAAiB,SAAS;AAChC,cAAM,mBAAmB,KAAK,IAAI;AAElC,yBAAiB,MAAM,gBAAgB;AACrC,gBAAM,KAAK,cAAc,UAAU,kBAAkB,UAAU,WAAW,GAAG,KAAK;AAAA,QACpF;AAEA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,cACJ,UACA,cACA,UACA,WACA,UACA;AACA,QAAI;AACJ,QAAI,UAAU,iBAAiB,KAAK,UAAU,MAAM;AAClD,kBAAY,KAAK,aAAa,SAAS,UAAU,EAAE,SAAS,UAAU;AAAA,IACxE;AAEA,UAAM,QAAQ,UAAU,KAAK;AAC7B,UAAM,QAAQ,KAAK,MAAM,WAAW,QAAQ;AAC5C,UAAM,iBAA2B,CAAC;AAElC,UAAM,SAAS,KAAK;AAEpB,eAAW,CAAC,MAAM,GAAG,GAAG,KAAK,OAAO;AAClC,UAAI,YAAY,KAAK,kBAAmB;AACxC,UAAI,KAAK,aAAc;AAEvB,YAAM,cAAc,KAAK,MAAM,cAAc,IAAI,EAAE;AACnD,qBAAe,KAAK,IAAI;AAExB,YAAM,UAAU,KAAK,IAAI,IAAI;AAC7B,YAAM,OAAO,SAAS,MAAM,GAAG,GAAG;AAElC,UAAI,QAAQ,KAAK;AACjB,UAAI;AACJ,UAAI,WAAW;AACb,gBAAQ;AACR,cAAM,oBAAoB,SAAS,qBAAqB,KAAK,MAAM;AACnE,cAAM,aAAa,oBAAoB;AACvC,cAAM,gBAAgB,KAAK,MAAM,QAAQ,OAAO;AAChD,cAAM,KAAK,gBAAgB,SAAS,mBAAmB;AACvD,cAAM,gBAAgB,KAAK,IAAI,GAAG,cAAc,EAAE;AAClD,gBAAQ,gBAAgB;AAAA,MAC1B,OAAO;AACL,gBAAQ,cAAc;AAAA,MACxB;AAEA,YAAM,aAAa,KAAK,IAAI,QAAQ,GAAG,IAAI,KAAK;AAChD,YAAM,KAAK,kBAAkB,aAAa,GAAO;AAEjD,WAAK;AAAA,QACH;AAAA,QACA,IAAI,qBAAqB;AAAA,UACvB,IAAI;AAAA,UACJ;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,OAAO;AAAA,UACP,UAAU,KAAK,MAAM;AAAA,QACvB,CAAC;AAAA,MACH;AAEA,WAAK,cAAc,GAAG,MAAM,IAAI,IAAI;AACpC,YAAM,KAAK,mBAAmB,QAAQ,cAAc,GAAO;AAC3D,eAAS,oBAAoB;AAAA,IAC/B;AAEA,SAAK;AAAA,MACH;AAAA,MACA,IAAI,qBAAqB;AAAA,QACvB,IAAI;AAAA,QACJ,MAAM;AAAA,QACN,WAAW,OAAO,CAAC;AAAA,QACnB,SAAS,OAAO,CAAC;AAAA,QACjB,OAAO;AAAA,QACP,UAAU,KAAK,MAAM;AAAA,MACvB,CAAC;AAAA,IACH;AAEA,SAAK,cAAc,GAAG,MAAM,IAAI,QAAQ;AAExC,UAAM,KAAK,kBAAkB,KAAK,MAAM,gBAAgB;AACxD,aAAS;AAAA,EACX;AAAA,EAEA,MAAM,kBAAkB,OAAe;AACrC,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,UAAU;AAAA,MACf,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,KAAK,CAAC;AAAA,IACrD,CAAC;AAAA,EACH;AAAA,EAEA,aAAa,MAAwB;AACnC,UAAM,UAAoB,CAAC;AAC3B,UAAM,QAAQ,KAAK,MAAM,WAAW,IAAI;AACxC,eAAW,QAAQ,OAAO;AACxB,YAAM,IAAI,KAAK,MAAM,cAAc,KAAK,CAAC,CAAC;AAC1C,cAAQ,KAAK,GAAG,CAAC;AAAA,IACnB;AACA,WAAO;AAAA,EACT;AAAA,EAEA,kBAAkB;AAChB,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAAA,EACF;AACF;","names":[]}
@@ -22,8 +22,7 @@ __export(tts_exports, {
22
22
  StreamAdapter: () => import_stream_adapter.StreamAdapter,
23
23
  StreamAdapterWrapper: () => import_stream_adapter.StreamAdapterWrapper,
24
24
  SynthesizeStream: () => import_tts.SynthesizeStream,
25
- TTS: () => import_tts.TTS,
26
- TTSEvent: () => import_tts.TTSEvent
25
+ TTS: () => import_tts.TTS
27
26
  });
28
27
  module.exports = __toCommonJS(tts_exports);
29
28
  var import_tts = require("./tts.cjs");
@@ -34,7 +33,6 @@ var import_stream_adapter = require("./stream_adapter.cjs");
34
33
  StreamAdapter,
35
34
  StreamAdapterWrapper,
36
35
  SynthesizeStream,
37
- TTS,
38
- TTSEvent
36
+ TTS
39
37
  });
40
38
  //# sourceMappingURL=index.cjs.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/tts/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport {\n type SynthesizedAudio,\n type TTSCapabilities,\n type TTSCallbacks,\n TTS,\n TTSEvent,\n SynthesizeStream,\n ChunkedStream,\n} from './tts.js';\nexport { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,iBAQO;AACP,4BAAoD;","names":[]}
1
+ {"version":3,"sources":["../../src/tts/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport {\n type SynthesizedAudio,\n type TTSCapabilities,\n type TTSCallbacks,\n TTS,\n SynthesizeStream,\n ChunkedStream,\n} from './tts.js';\nexport { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,iBAOO;AACP,4BAAoD;","names":[]}
@@ -1,3 +1,3 @@
1
- export { type SynthesizedAudio, type TTSCapabilities, type TTSCallbacks, TTS, TTSEvent, SynthesizeStream, ChunkedStream, } from './tts.js';
1
+ export { type SynthesizedAudio, type TTSCapabilities, type TTSCallbacks, TTS, SynthesizeStream, ChunkedStream, } from './tts.js';
2
2
  export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';
3
3
  //# sourceMappingURL=index.d.ts.map
@@ -1,3 +1,3 @@
1
- export { type SynthesizedAudio, type TTSCapabilities, type TTSCallbacks, TTS, TTSEvent, SynthesizeStream, ChunkedStream, } from './tts.js';
1
+ export { type SynthesizedAudio, type TTSCapabilities, type TTSCallbacks, TTS, SynthesizeStream, ChunkedStream, } from './tts.js';
2
2
  export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';
3
3
  //# sourceMappingURL=index.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/tts/index.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,gBAAgB,EACrB,KAAK,eAAe,EACpB,KAAK,YAAY,EACjB,GAAG,EACH,QAAQ,EACR,gBAAgB,EAChB,aAAa,GACd,MAAM,UAAU,CAAC;AAClB,OAAO,EAAE,aAAa,EAAE,oBAAoB,EAAE,MAAM,qBAAqB,CAAC"}
1
+ {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/tts/index.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,gBAAgB,EACrB,KAAK,eAAe,EACpB,KAAK,YAAY,EACjB,GAAG,EACH,gBAAgB,EAChB,aAAa,GACd,MAAM,UAAU,CAAC;AAClB,OAAO,EAAE,aAAa,EAAE,oBAAoB,EAAE,MAAM,qBAAqB,CAAC"}
package/dist/tts/index.js CHANGED
@@ -1,6 +1,5 @@
1
1
  import {
2
2
  TTS,
3
- TTSEvent,
4
3
  SynthesizeStream,
5
4
  ChunkedStream
6
5
  } from "./tts.js";
@@ -10,7 +9,6 @@ export {
10
9
  StreamAdapter,
11
10
  StreamAdapterWrapper,
12
11
  SynthesizeStream,
13
- TTS,
14
- TTSEvent
12
+ TTS
15
13
  };
16
14
  //# sourceMappingURL=index.js.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/tts/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport {\n type SynthesizedAudio,\n type TTSCapabilities,\n type TTSCallbacks,\n TTS,\n TTSEvent,\n SynthesizeStream,\n ChunkedStream,\n} from './tts.js';\nexport { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';\n"],"mappings":"AAGA;AAAA,EAIE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,eAAe,4BAA4B;","names":[]}
1
+ {"version":3,"sources":["../../src/tts/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport {\n type SynthesizedAudio,\n type TTSCapabilities,\n type TTSCallbacks,\n TTS,\n SynthesizeStream,\n ChunkedStream,\n} from './tts.js';\nexport { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';\n"],"mappings":"AAGA;AAAA,EAIE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,eAAe,4BAA4B;","names":[]}
@@ -22,6 +22,7 @@ __export(stream_adapter_exports, {
22
22
  StreamAdapterWrapper: () => StreamAdapterWrapper
23
23
  });
24
24
  module.exports = __toCommonJS(stream_adapter_exports);
25
+ var import_utils = require("../utils.cjs");
25
26
  var import_tts = require("./tts.cjs");
26
27
  class StreamAdapter extends import_tts.TTS {
27
28
  #tts;
@@ -33,8 +34,8 @@ class StreamAdapter extends import_tts.TTS {
33
34
  this.#sentenceTokenizer = sentenceTokenizer;
34
35
  this.label = this.#tts.label;
35
36
  this.label = `tts.StreamAdapter<${this.#tts.label}>`;
36
- this.#tts.on(import_tts.TTSEvent.METRICS_COLLECTED, (metrics) => {
37
- this.emit(import_tts.TTSEvent.METRICS_COLLECTED, metrics);
37
+ this.#tts.on("metrics_collected", (metrics) => {
38
+ this.emit("metrics_collected", metrics);
38
39
  });
39
40
  }
40
41
  synthesize(text) {
@@ -53,14 +54,11 @@ class StreamAdapterWrapper extends import_tts.SynthesizeStream {
53
54
  this.#tts = tts;
54
55
  this.#sentenceStream = sentenceTokenizer.stream();
55
56
  this.label = `tts.StreamAdapterWrapper<${this.#tts.label}>`;
56
- this.#run();
57
57
  }
58
- async monitorMetrics() {
59
- return;
60
- }
61
- async #run() {
58
+ async run() {
62
59
  const forwardInput = async () => {
63
60
  for await (const input of this.input) {
61
+ if (this.abortController.signal.aborted) break;
64
62
  if (input === import_tts.SynthesizeStream.FLUSH_SENTINEL) {
65
63
  this.#sentenceStream.flush();
66
64
  } else {
@@ -70,15 +68,30 @@ class StreamAdapterWrapper extends import_tts.SynthesizeStream {
70
68
  this.#sentenceStream.endInput();
71
69
  this.#sentenceStream.close();
72
70
  };
73
- const synthesize = async () => {
71
+ const synthesizeSentenceStream = async () => {
72
+ let task;
73
+ const tokenCompletionTasks = [];
74
74
  for await (const ev of this.#sentenceStream) {
75
- for await (const audio of this.#tts.synthesize(ev.token)) {
76
- this.output.put(audio);
77
- }
75
+ if (this.abortController.signal.aborted) break;
76
+ task = import_utils.Task.from(
77
+ (controller) => synthesize(ev.token, task, controller),
78
+ this.abortController
79
+ );
80
+ tokenCompletionTasks.push(task);
81
+ }
82
+ await Promise.all(tokenCompletionTasks.map((t) => t.result));
83
+ this.queue.put(import_tts.SynthesizeStream.END_OF_STREAM);
84
+ };
85
+ const synthesize = async (token, prevTask, controller) => {
86
+ const audioStream = this.#tts.synthesize(token);
87
+ await (prevTask == null ? void 0 : prevTask.result);
88
+ if (controller.signal.aborted) return;
89
+ for await (const audio of audioStream) {
90
+ if (controller.signal.aborted) break;
91
+ this.queue.put(audio);
78
92
  }
79
- this.output.put(import_tts.SynthesizeStream.END_OF_STREAM);
80
93
  };
81
- Promise.all([forwardInput(), synthesize()]);
94
+ await Promise.all([forwardInput(), synthesizeSentenceStream()]);
82
95
  }
83
96
  }
84
97
  // Annotate the CommonJS export names for ESM import in node:
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/tts/stream_adapter.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { SentenceStream, SentenceTokenizer } from '../tokenize/index.js';\nimport type { ChunkedStream } from './tts.js';\nimport { SynthesizeStream, TTS, TTSEvent } from './tts.js';\n\nexport class StreamAdapter extends TTS {\n #tts: TTS;\n #sentenceTokenizer: SentenceTokenizer;\n label: string;\n\n constructor(tts: TTS, sentenceTokenizer: SentenceTokenizer) {\n super(tts.sampleRate, tts.numChannels, { streaming: true });\n this.#tts = tts;\n this.#sentenceTokenizer = sentenceTokenizer;\n this.label = this.#tts.label;\n this.label = `tts.StreamAdapter<${this.#tts.label}>`;\n\n this.#tts.on(TTSEvent.METRICS_COLLECTED, (metrics) => {\n this.emit(TTSEvent.METRICS_COLLECTED, metrics);\n });\n }\n\n synthesize(text: string): ChunkedStream {\n return this.#tts.synthesize(text);\n }\n\n stream(): StreamAdapterWrapper {\n return new StreamAdapterWrapper(this.#tts, this.#sentenceTokenizer);\n }\n}\n\nexport class StreamAdapterWrapper extends SynthesizeStream {\n #tts: TTS;\n #sentenceStream: SentenceStream;\n label: string;\n\n constructor(tts: TTS, sentenceTokenizer: SentenceTokenizer) {\n super(tts);\n this.#tts = tts;\n this.#sentenceStream = sentenceTokenizer.stream();\n this.label = `tts.StreamAdapterWrapper<${this.#tts.label}>`;\n\n this.#run();\n }\n\n async monitorMetrics() {\n return; // do nothing\n }\n\n async #run() {\n const forwardInput = async () => {\n for await (const input of this.input) {\n if (input === SynthesizeStream.FLUSH_SENTINEL) {\n this.#sentenceStream.flush();\n } else {\n this.#sentenceStream.pushText(input);\n }\n }\n this.#sentenceStream.endInput();\n this.#sentenceStream.close();\n };\n\n const synthesize = async () => {\n for await (const ev of this.#sentenceStream) {\n for await (const audio of this.#tts.synthesize(ev.token)) {\n this.output.put(audio);\n }\n }\n this.output.put(SynthesizeStream.END_OF_STREAM);\n };\n\n Promise.all([forwardInput(), synthesize()]);\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,iBAAgD;AAEzC,MAAM,sBAAsB,eAAI;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,mBAAsC;AAC1D,UAAM,IAAI,YAAY,IAAI,aAAa,EAAE,WAAW,KAAK,CAAC;AAC1D,SAAK,OAAO;AACZ,SAAK,qBAAqB;AAC1B,SAAK,QAAQ,KAAK,KAAK;AACvB,SAAK,QAAQ,qBAAqB,KAAK,KAAK,KAAK;AAEjD,SAAK,KAAK,GAAG,oBAAS,mBAAmB,CAAC,YAAY;AACpD,WAAK,KAAK,oBAAS,mBAAmB,OAAO;AAAA,IAC/C,CAAC;AAAA,EACH;AAAA,EAEA,WAAW,MAA6B;AACtC,WAAO,KAAK,KAAK,WAAW,IAAI;AAAA,EAClC;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,qBAAqB,KAAK,MAAM,KAAK,kBAAkB;AAAA,EACpE;AACF;AAEO,MAAM,6BAA6B,4BAAiB;AAAA,EACzD;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,mBAAsC;AAC1D,UAAM,GAAG;AACT,SAAK,OAAO;AACZ,SAAK,kBAAkB,kBAAkB,OAAO;AAChD,SAAK,QAAQ,4BAA4B,KAAK,KAAK,KAAK;AAExD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,iBAAiB;AACrB;AAAA,EACF;AAAA,EAEA,MAAM,OAAO;AACX,UAAM,eAAe,YAAY;AAC/B,uBAAiB,SAAS,KAAK,OAAO;AACpC,YAAI,UAAU,4BAAiB,gBAAgB;AAC7C,eAAK,gBAAgB,MAAM;AAAA,QAC7B,OAAO;AACL,eAAK,gBAAgB,SAAS,KAAK;AAAA,QACrC;AAAA,MACF;AACA,WAAK,gBAAgB,SAAS;AAC9B,WAAK,gBAAgB,MAAM;AAAA,IAC7B;AAEA,UAAM,aAAa,YAAY;AAC7B,uBAAiB,MAAM,KAAK,iBAAiB;AAC3C,yBAAiB,SAAS,KAAK,KAAK,WAAW,GAAG,KAAK,GAAG;AACxD,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB;AAAA,MACF;AACA,WAAK,OAAO,IAAI,4BAAiB,aAAa;AAAA,IAChD;AAEA,YAAQ,IAAI,CAAC,aAAa,GAAG,WAAW,CAAC,CAAC;AAAA,EAC5C;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/tts/stream_adapter.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { SentenceStream, SentenceTokenizer } from '../tokenize/index.js';\nimport { Task } from '../utils.js';\nimport type { ChunkedStream } from './tts.js';\nimport { SynthesizeStream, TTS } from './tts.js';\n\nexport class StreamAdapter extends TTS {\n #tts: TTS;\n #sentenceTokenizer: SentenceTokenizer;\n label: string;\n\n constructor(tts: TTS, sentenceTokenizer: SentenceTokenizer) {\n super(tts.sampleRate, tts.numChannels, { streaming: true });\n this.#tts = tts;\n this.#sentenceTokenizer = sentenceTokenizer;\n this.label = this.#tts.label;\n this.label = `tts.StreamAdapter<${this.#tts.label}>`;\n\n this.#tts.on('metrics_collected', (metrics) => {\n this.emit('metrics_collected', metrics);\n });\n }\n\n synthesize(text: string): ChunkedStream {\n return this.#tts.synthesize(text);\n }\n\n stream(): StreamAdapterWrapper {\n return new StreamAdapterWrapper(this.#tts, this.#sentenceTokenizer);\n }\n}\n\nexport class StreamAdapterWrapper extends SynthesizeStream {\n #tts: TTS;\n #sentenceStream: SentenceStream;\n label: string;\n\n constructor(tts: TTS, sentenceTokenizer: SentenceTokenizer) {\n super(tts);\n this.#tts = tts;\n this.#sentenceStream = sentenceTokenizer.stream();\n this.label = `tts.StreamAdapterWrapper<${this.#tts.label}>`;\n }\n\n protected async run() {\n const forwardInput = async () => {\n for await (const input of this.input) {\n if (this.abortController.signal.aborted) break;\n\n if (input === SynthesizeStream.FLUSH_SENTINEL) {\n this.#sentenceStream.flush();\n } else {\n this.#sentenceStream.pushText(input);\n }\n }\n this.#sentenceStream.endInput();\n this.#sentenceStream.close();\n };\n\n const synthesizeSentenceStream = async () => {\n let task: Task<void> | undefined;\n const tokenCompletionTasks: Task<void>[] = [];\n\n for await (const ev of this.#sentenceStream) {\n if (this.abortController.signal.aborted) break;\n\n // this will enable non-blocking synthesis of the stream of tokens\n task = Task.from(\n (controller) => synthesize(ev.token, task, controller),\n this.abortController,\n );\n\n tokenCompletionTasks.push(task);\n }\n\n await Promise.all(tokenCompletionTasks.map((t) => t.result));\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n };\n\n const synthesize = async (\n token: string,\n prevTask: Task<void> | undefined,\n controller: AbortController,\n ) => {\n const audioStream = this.#tts.synthesize(token);\n\n // wait for previous audio transcription to complete before starting\n // to queuing audio frames of the current token\n await prevTask?.result;\n if (controller.signal.aborted) return;\n\n for await (const audio of audioStream) {\n if (controller.signal.aborted) break;\n this.queue.put(audio);\n }\n };\n\n await Promise.all([forwardInput(), synthesizeSentenceStream()]);\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,mBAAqB;AAErB,iBAAsC;AAE/B,MAAM,sBAAsB,eAAI;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,mBAAsC;AAC1D,UAAM,IAAI,YAAY,IAAI,aAAa,EAAE,WAAW,KAAK,CAAC;AAC1D,SAAK,OAAO;AACZ,SAAK,qBAAqB;AAC1B,SAAK,QAAQ,KAAK,KAAK;AACvB,SAAK,QAAQ,qBAAqB,KAAK,KAAK,KAAK;AAEjD,SAAK,KAAK,GAAG,qBAAqB,CAAC,YAAY;AAC7C,WAAK,KAAK,qBAAqB,OAAO;AAAA,IACxC,CAAC;AAAA,EACH;AAAA,EAEA,WAAW,MAA6B;AACtC,WAAO,KAAK,KAAK,WAAW,IAAI;AAAA,EAClC;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,qBAAqB,KAAK,MAAM,KAAK,kBAAkB;AAAA,EACpE;AACF;AAEO,MAAM,6BAA6B,4BAAiB;AAAA,EACzD;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,mBAAsC;AAC1D,UAAM,GAAG;AACT,SAAK,OAAO;AACZ,SAAK,kBAAkB,kBAAkB,OAAO;AAChD,SAAK,QAAQ,4BAA4B,KAAK,KAAK,KAAK;AAAA,EAC1D;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,eAAe,YAAY;AAC/B,uBAAiB,SAAS,KAAK,OAAO;AACpC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,YAAI,UAAU,4BAAiB,gBAAgB;AAC7C,eAAK,gBAAgB,MAAM;AAAA,QAC7B,OAAO;AACL,eAAK,gBAAgB,SAAS,KAAK;AAAA,QACrC;AAAA,MACF;AACA,WAAK,gBAAgB,SAAS;AAC9B,WAAK,gBAAgB,MAAM;AAAA,IAC7B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,UAAI;AACJ,YAAM,uBAAqC,CAAC;AAE5C,uBAAiB,MAAM,KAAK,iBAAiB;AAC3C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAGzC,eAAO,kBAAK;AAAA,UACV,CAAC,eAAe,WAAW,GAAG,OAAO,MAAM,UAAU;AAAA,UACrD,KAAK;AAAA,QACP;AAEA,6BAAqB,KAAK,IAAI;AAAA,MAChC;AAEA,YAAM,QAAQ,IAAI,qBAAqB,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAC3D,WAAK,MAAM,IAAI,4BAAiB,aAAa;AAAA,IAC/C;AAEA,UAAM,aAAa,OACjB,OACA,UACA,eACG;AACH,YAAM,cAAc,KAAK,KAAK,WAAW,KAAK;AAI9C,aAAM,qCAAU;AAChB,UAAI,WAAW,OAAO,QAAS;AAE/B,uBAAiB,SAAS,aAAa;AACrC,YAAI,WAAW,OAAO,QAAS;AAC/B,aAAK,MAAM,IAAI,KAAK;AAAA,MACtB;AAAA,IACF;AAEA,UAAM,QAAQ,IAAI,CAAC,aAAa,GAAG,yBAAyB,CAAC,CAAC;AAAA,EAChE;AACF;","names":[]}
@@ -12,6 +12,6 @@ export declare class StreamAdapterWrapper extends SynthesizeStream {
12
12
  #private;
13
13
  label: string;
14
14
  constructor(tts: TTS, sentenceTokenizer: SentenceTokenizer);
15
- monitorMetrics(): Promise<void>;
15
+ protected run(): Promise<void>;
16
16
  }
17
17
  //# sourceMappingURL=stream_adapter.d.ts.map
@@ -12,6 +12,6 @@ export declare class StreamAdapterWrapper extends SynthesizeStream {
12
12
  #private;
13
13
  label: string;
14
14
  constructor(tts: TTS, sentenceTokenizer: SentenceTokenizer);
15
- monitorMetrics(): Promise<void>;
15
+ protected run(): Promise<void>;
16
16
  }
17
17
  //# sourceMappingURL=stream_adapter.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"stream_adapter.d.ts","sourceRoot":"","sources":["../../src/tts/stream_adapter.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAkB,iBAAiB,EAAE,MAAM,sBAAsB,CAAC;AAC9E,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,UAAU,CAAC;AAC9C,OAAO,EAAE,gBAAgB,EAAE,GAAG,EAAY,MAAM,UAAU,CAAC;AAE3D,qBAAa,aAAc,SAAQ,GAAG;;IAGpC,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,iBAAiB,EAAE,iBAAiB;IAY1D,UAAU,CAAC,IAAI,EAAE,MAAM,GAAG,aAAa;IAIvC,MAAM,IAAI,oBAAoB;CAG/B;AAED,qBAAa,oBAAqB,SAAQ,gBAAgB;;IAGxD,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,iBAAiB,EAAE,iBAAiB;IASpD,cAAc;CA4BrB"}
1
+ {"version":3,"file":"stream_adapter.d.ts","sourceRoot":"","sources":["../../src/tts/stream_adapter.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAkB,iBAAiB,EAAE,MAAM,sBAAsB,CAAC;AAE9E,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,UAAU,CAAC;AAC9C,OAAO,EAAE,gBAAgB,EAAE,GAAG,EAAE,MAAM,UAAU,CAAC;AAEjD,qBAAa,aAAc,SAAQ,GAAG;;IAGpC,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,iBAAiB,EAAE,iBAAiB;IAY1D,UAAU,CAAC,IAAI,EAAE,MAAM,GAAG,aAAa;IAIvC,MAAM,IAAI,oBAAoB;CAG/B;AAED,qBAAa,oBAAqB,SAAQ,gBAAgB;;IAGxD,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,iBAAiB,EAAE,iBAAiB;cAO1C,GAAG;CAuDpB"}