@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
@@ -1,425 +0,0 @@
1
- import {
2
- AudioSource,
3
- AudioStream,
4
- LocalAudioTrack,
5
- RoomEvent,
6
- TrackPublishOptions,
7
- TrackSource
8
- } from "@livekit/rtc-node";
9
- import { EventEmitter } from "node:events";
10
- import { AudioByteStream } from "../audio.js";
11
- import {
12
- ATTRIBUTE_TRANSCRIPTION_FINAL,
13
- ATTRIBUTE_TRANSCRIPTION_TRACK_ID,
14
- TOPIC_TRANSCRIPTION
15
- } from "../constants.js";
16
- import * as llm from "../llm/index.js";
17
- import { log } from "../log.js";
18
- import { TextAudioSynchronizer, defaultTextSyncOptions } from "../transcription.js";
19
- import { findMicroTrackId } from "../utils.js";
20
- import { AgentPlayout } from "./agent_playout.js";
21
- class RealtimeSession extends EventEmitter {
22
- }
23
- class RealtimeModel {
24
- }
25
- const AGENT_STATE_ATTRIBUTE = "lk.agent.state";
26
- class MultimodalAgent extends EventEmitter {
27
- model;
28
- room = null;
29
- linkedParticipant = null;
30
- subscribedTrack = null;
31
- readMicroTask = null;
32
- #textResponseRetries = 0;
33
- #maxTextResponseRetries;
34
- constructor({
35
- model,
36
- chatCtx,
37
- fncCtx,
38
- maxTextResponseRetries = 5,
39
- noiseCancellation
40
- }) {
41
- super();
42
- this.model = model;
43
- this.#chatCtx = chatCtx;
44
- this.#fncCtx = fncCtx;
45
- this.#maxTextResponseRetries = maxTextResponseRetries;
46
- this.#noiseCancellation = noiseCancellation;
47
- }
48
- #participant = null;
49
- #agentPublication = null;
50
- #localTrackSid = null;
51
- #localSource = null;
52
- #agentPlayout = null;
53
- #playingHandle = void 0;
54
- #logger = log();
55
- #session = null;
56
- #fncCtx = void 0;
57
- #chatCtx = void 0;
58
- #noiseCancellation = void 0;
59
- #_started = false;
60
- #_pendingFunctionCalls = /* @__PURE__ */ new Set();
61
- #_speaking = false;
62
- get fncCtx() {
63
- return this.#fncCtx;
64
- }
65
- set fncCtx(ctx) {
66
- this.#fncCtx = ctx;
67
- if (this.#session) {
68
- this.#session.fncCtx = ctx;
69
- }
70
- }
71
- get #pendingFunctionCalls() {
72
- return this.#_pendingFunctionCalls;
73
- }
74
- set #pendingFunctionCalls(calls) {
75
- this.#_pendingFunctionCalls = calls;
76
- this.#updateState();
77
- }
78
- get #speaking() {
79
- return this.#_speaking;
80
- }
81
- set #speaking(isSpeaking) {
82
- this.#_speaking = isSpeaking;
83
- this.#updateState();
84
- }
85
- get #started() {
86
- return this.#_started;
87
- }
88
- set #started(started) {
89
- this.#_started = started;
90
- this.#updateState();
91
- }
92
- start(room, participant = null) {
93
- return new Promise(async (resolve, reject) => {
94
- var _a;
95
- if (this.#started) {
96
- reject(new Error("MultimodalAgent already started"));
97
- }
98
- this.#updateState();
99
- room.on(RoomEvent.ParticipantConnected, (participant2) => {
100
- if (this.linkedParticipant) {
101
- return;
102
- }
103
- this.#linkParticipant(participant2.identity);
104
- });
105
- room.on(
106
- RoomEvent.TrackPublished,
107
- (trackPublication, participant2) => {
108
- if (this.linkedParticipant && participant2.identity === this.linkedParticipant.identity && trackPublication.source === TrackSource.SOURCE_MICROPHONE && !trackPublication.subscribed) {
109
- trackPublication.setSubscribed(true);
110
- }
111
- }
112
- );
113
- room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));
114
- this.room = room;
115
- this.#participant = participant;
116
- this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);
117
- this.#agentPlayout = new AgentPlayout(
118
- this.#localSource,
119
- this.model.sampleRate,
120
- this.model.numChannels,
121
- this.model.inFrameSize,
122
- this.model.outFrameSize
123
- );
124
- const onPlayoutStarted = () => {
125
- this.emit("agent_started_speaking");
126
- this.#speaking = true;
127
- };
128
- const onPlayoutStopped = (interrupted) => {
129
- this.emit("agent_stopped_speaking");
130
- this.#speaking = false;
131
- if (this.#playingHandle) {
132
- let text = this.#playingHandle.synchronizer.playedText;
133
- if (interrupted) {
134
- text += "\u2026";
135
- }
136
- const msg = llm.ChatMessage.create({
137
- role: llm.ChatRole.ASSISTANT,
138
- text
139
- });
140
- if (interrupted) {
141
- this.emit("agent_speech_interrupted", msg);
142
- } else {
143
- this.emit("agent_speech_committed", msg);
144
- }
145
- this.#logger.child({ transcription: text, interrupted }).debug("committed agent speech");
146
- }
147
- };
148
- this.#agentPlayout.on("playout_started", onPlayoutStarted);
149
- this.#agentPlayout.on("playout_stopped", onPlayoutStopped);
150
- const track = LocalAudioTrack.createAudioTrack("assistant_voice", this.#localSource);
151
- const options = new TrackPublishOptions();
152
- options.source = TrackSource.SOURCE_MICROPHONE;
153
- this.#agentPublication = await ((_a = room.localParticipant) == null ? void 0 : _a.publishTrack(track, options)) || null;
154
- if (!this.#agentPublication) {
155
- this.#logger.error("Failed to publish track");
156
- reject(new Error("Failed to publish track"));
157
- return;
158
- }
159
- await this.#agentPublication.waitForSubscription();
160
- if (participant) {
161
- if (typeof participant === "string") {
162
- this.#linkParticipant(participant);
163
- } else {
164
- this.#linkParticipant(participant.identity);
165
- }
166
- } else {
167
- for (const participant2 of room.remoteParticipants.values()) {
168
- this.#linkParticipant(participant2.identity);
169
- break;
170
- }
171
- }
172
- this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });
173
- this.#started = true;
174
- this.#session.on("response_content_added", (message) => {
175
- var _a2;
176
- if (message.contentType === "text") return;
177
- const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);
178
- synchronizer.on("textUpdated", async (text) => {
179
- await this.#publishTranscription(
180
- this.room.localParticipant.identity,
181
- this.#getLocalTrackSid(),
182
- text.text,
183
- text.final,
184
- text.id
185
- );
186
- });
187
- const handle = (_a2 = this.#agentPlayout) == null ? void 0 : _a2.play(
188
- message.itemId,
189
- message.contentIndex,
190
- synchronizer,
191
- message.textStream,
192
- message.audioStream
193
- );
194
- this.#playingHandle = handle;
195
- });
196
- this.#session.on("response_content_done", (message) => {
197
- if (message.contentType === "text") {
198
- if (this.#textResponseRetries >= this.#maxTextResponseRetries) {
199
- throw new Error(
200
- `The OpenAI Realtime API returned a text response after ${this.#maxTextResponseRetries} retries. Please try to reduce the number of text system or assistant messages in the chat context.`
201
- );
202
- }
203
- this.#textResponseRetries++;
204
- this.#logger.child({
205
- itemId: message.itemId,
206
- text: message.text,
207
- retries: this.#textResponseRetries
208
- }).warn(
209
- "The OpenAI Realtime API returned a text response instead of audio. Attempting to recover to audio mode..."
210
- );
211
- this.#session.recoverFromTextResponse(message.itemId);
212
- } else {
213
- this.#textResponseRetries = 0;
214
- }
215
- });
216
- this.#session.on("input_speech_committed", async (ev) => {
217
- var _a2, _b;
218
- const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
219
- const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
220
- if (participantIdentity && trackSid) {
221
- await this.#publishTranscription(participantIdentity, trackSid, "\u2026", false, ev.itemId);
222
- } else {
223
- this.#logger.error("Participant or track not set");
224
- }
225
- });
226
- this.#session.on("input_speech_transcription_completed", async (ev) => {
227
- var _a2, _b;
228
- const transcription = ev.transcript;
229
- const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
230
- const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
231
- if (participantIdentity && trackSid) {
232
- await this.#publishTranscription(
233
- participantIdentity,
234
- trackSid,
235
- transcription,
236
- true,
237
- ev.itemId
238
- );
239
- } else {
240
- this.#logger.error("Participant or track not set");
241
- }
242
- const userMsg = llm.ChatMessage.create({
243
- role: llm.ChatRole.USER,
244
- text: transcription
245
- });
246
- this.emit("user_speech_committed", userMsg);
247
- this.#logger.child({ transcription }).debug("committed user speech");
248
- });
249
- this.#session.on("input_speech_started", async (ev) => {
250
- var _a2, _b;
251
- this.emit("user_started_speaking");
252
- if (this.#playingHandle && !this.#playingHandle.done) {
253
- this.#playingHandle.interrupt();
254
- this.#session.conversation.item.truncate(
255
- this.#playingHandle.itemId,
256
- this.#playingHandle.contentIndex,
257
- Math.floor(this.#playingHandle.audioSamples / 24e3 * 1e3)
258
- );
259
- this.#playingHandle = void 0;
260
- }
261
- const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
262
- const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
263
- if (participantIdentity && trackSid) {
264
- await this.#publishTranscription(participantIdentity, trackSid, "\u2026", false, ev.itemId);
265
- }
266
- });
267
- this.#session.on("input_speech_stopped", (ev) => {
268
- this.emit("user_stopped_speaking");
269
- });
270
- this.#session.on("function_call_started", (ev) => {
271
- this.#pendingFunctionCalls.add(ev.callId);
272
- this.#updateState();
273
- });
274
- this.#session.on("function_call_completed", (ev) => {
275
- this.#pendingFunctionCalls.delete(ev.callId);
276
- this.#updateState();
277
- });
278
- this.#session.on("function_call_failed", (ev) => {
279
- this.#pendingFunctionCalls.delete(ev.callId);
280
- this.#updateState();
281
- });
282
- this.#session.on("metrics_collected", (metrics) => {
283
- this.emit("metrics_collected", metrics);
284
- });
285
- resolve(this.#session);
286
- });
287
- }
288
- #linkParticipant(participantIdentity) {
289
- if (!this.room) {
290
- this.#logger.error("Room is not set");
291
- return;
292
- }
293
- this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;
294
- if (!this.linkedParticipant) {
295
- this.#logger.error(`Participant with identity ${participantIdentity} not found`);
296
- return;
297
- }
298
- if (this.linkedParticipant.trackPublications.size > 0) {
299
- this.#subscribeToMicrophone();
300
- }
301
- for (const publication of this.linkedParticipant.trackPublications.values()) {
302
- if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {
303
- this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);
304
- break;
305
- }
306
- }
307
- }
308
- #subscribeToMicrophone() {
309
- if (!this.linkedParticipant) {
310
- this.#logger.error("Participant is not set");
311
- return;
312
- }
313
- let microphonePublication = void 0;
314
- for (const publication of this.linkedParticipant.trackPublications.values()) {
315
- if (publication.source === TrackSource.SOURCE_MICROPHONE) {
316
- microphonePublication = publication;
317
- break;
318
- }
319
- }
320
- if (!microphonePublication) {
321
- return;
322
- }
323
- if (!microphonePublication.subscribed) {
324
- microphonePublication.setSubscribed(true);
325
- }
326
- }
327
- #handleTrackSubscription(track, publication, participant) {
328
- var _a;
329
- if (publication.source !== TrackSource.SOURCE_MICROPHONE || participant.identity !== ((_a = this.linkedParticipant) == null ? void 0 : _a.identity)) {
330
- return;
331
- }
332
- const readAudioStreamTask = async (audioStream) => {
333
- const bstream = new AudioByteStream(
334
- this.model.sampleRate,
335
- this.model.numChannels,
336
- this.model.inFrameSize
337
- );
338
- for await (const frame of audioStream) {
339
- const audioData = frame.data;
340
- for (const frame2 of bstream.write(audioData.buffer)) {
341
- this.#session.inputAudioBuffer.append(frame2);
342
- }
343
- }
344
- };
345
- this.subscribedTrack = track;
346
- this.readMicroTask = new Promise((resolve, reject) => {
347
- const audioStreamOptions = {
348
- sampleRate: this.model.sampleRate,
349
- numChannels: this.model.numChannels,
350
- ...this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}
351
- };
352
- readAudioStreamTask(new AudioStream(track, audioStreamOptions)).then(resolve).catch(reject);
353
- });
354
- }
355
- #getLocalTrackSid() {
356
- if (!this.#localTrackSid && this.room && this.room.localParticipant) {
357
- this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant.identity);
358
- }
359
- return this.#localTrackSid;
360
- }
361
- async #publishTranscription(participantIdentity, trackSid, text, isFinal, id) {
362
- var _a;
363
- this.#logger.debug(
364
- `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`
365
- );
366
- if (!((_a = this.room) == null ? void 0 : _a.localParticipant)) {
367
- this.#logger.error("Room or local participant not set");
368
- return;
369
- }
370
- this.room.localParticipant.publishTranscription({
371
- participantIdentity,
372
- trackSid,
373
- segments: [
374
- {
375
- text,
376
- final: isFinal,
377
- id,
378
- startTime: BigInt(0),
379
- endTime: BigInt(0),
380
- language: ""
381
- }
382
- ]
383
- });
384
- const stream = await this.room.localParticipant.streamText({
385
- topic: TOPIC_TRANSCRIPTION,
386
- senderIdentity: participantIdentity,
387
- attributes: {
388
- [ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,
389
- [ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString()
390
- }
391
- });
392
- await stream.write(text);
393
- await stream.close();
394
- }
395
- #updateState() {
396
- let newState = "initializing";
397
- if (this.#pendingFunctionCalls.size > 0) {
398
- newState = "thinking";
399
- } else if (this.#speaking) {
400
- newState = "speaking";
401
- } else if (this.#started) {
402
- newState = "listening";
403
- }
404
- this.#setState(newState);
405
- }
406
- #setState(state) {
407
- var _a;
408
- if (((_a = this.room) == null ? void 0 : _a.isConnected) && this.room.localParticipant) {
409
- const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];
410
- if (currentState !== state) {
411
- this.room.localParticipant.setAttributes({
412
- [AGENT_STATE_ATTRIBUTE]: state
413
- });
414
- this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);
415
- }
416
- }
417
- }
418
- }
419
- export {
420
- AGENT_STATE_ATTRIBUTE,
421
- MultimodalAgent,
422
- RealtimeModel,
423
- RealtimeSession
424
- };
425
- //# sourceMappingURL=multimodal_agent.js.map
@@ -1 +0,0 @@
1
- {"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n NoiseCancellationOptions,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport {\n ATTRIBUTE_TRANSCRIPTION_FINAL,\n ATTRIBUTE_TRANSCRIPTION_TRACK_ID,\n TOPIC_TRANSCRIPTION,\n} from '../constants.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { TextAudioSynchronizer, defaultTextSyncOptions } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n abstract recoverFromTextResponse(itemId: string): void;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n #textResponseRetries = 0;\n #maxTextResponseRetries: number;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n maxTextResponseRetries = 5,\n noiseCancellation,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n maxTextResponseRetries?: number;\n noiseCancellation?: NoiseCancellationOptions;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n this.#maxTextResponseRetries = maxTextResponseRetries;\n this.#noiseCancellation = noiseCancellation;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n #noiseCancellation: NoiseCancellationOptions | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity!);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.synchronizer.playedText;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity!);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') return;\n\n const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);\n synchronizer.on('textUpdated', async (text) => {\n await this.#publishTranscription(\n this.room!.localParticipant!.identity!,\n this.#getLocalTrackSid()!,\n text.text,\n text.final,\n text.id,\n );\n });\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n synchronizer,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_done', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') {\n if (this.#textResponseRetries >= this.#maxTextResponseRetries) {\n throw new Error(\n 'The OpenAI Realtime API returned a text response ' +\n `after ${this.#maxTextResponseRetries} retries. ` +\n 'Please try to reduce the number of text system or ' +\n 'assistant messages in the chat context.',\n );\n }\n\n this.#textResponseRetries++;\n this.#logger\n .child({\n itemId: message.itemId,\n text: message.text,\n retries: this.#textResponseRetries,\n })\n .warn(\n 'The OpenAI Realtime API returned a text response instead of audio. ' +\n 'Attempting to recover to audio mode...',\n );\n this.#session!.recoverFromTextResponse(message.itemId);\n } else {\n this.#textResponseRetries = 0;\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', async (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n await this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', async (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n await this.#publishTranscription(\n participantIdentity,\n trackSid,\n transcription,\n true,\n ev.itemId,\n );\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', async (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n await this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n const audioStreamOptions = {\n sampleRate: this.model.sampleRate,\n numChannels: this.model.numChannels,\n ...(this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}),\n };\n readAudioStreamTask(new AudioStream(track, audioStreamOptions)).then(resolve).catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant!.identity!);\n }\n return this.#localTrackSid;\n }\n\n async #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): Promise<void> {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n\n const stream = await this.room.localParticipant.streamText({\n topic: TOPIC_TRANSCRIPTION,\n senderIdentity: participantIdentity,\n attributes: {\n [ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,\n [ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString(),\n },\n });\n await stream.write(text);\n await stream.close();\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes![AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":"AAYA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAChC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,YAAY,SAAS;AACrB,SAAS,WAAW;AAEpB,SAAS,uBAAuB,8BAA8B;AAC9D,SAAS,wBAAwB;AACjC,SAAS,oBAAwC;AAM1C,MAAe,wBAAwB,aAAa;AAO3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,aAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,uBAAuB;AAAA,EACvB;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA,yBAAyB;AAAA,IACzB;AAAA,EACF,GAMG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,0BAA0B;AAC/B,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,UAAU,IAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EACxC,qBAA2D;AAAA,EAE3D,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AA1JlD;AA2JM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,UAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAS;AAAA,MAC7C,CAAC;AACD,WAAK;AAAA,QACH,UAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,YAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,UAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,YAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,aAAa;AAC5C,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oBAAoB;AACxC,cAAQ,SAAS,YAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAS;AAAA,QAC7C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAS;AAC3C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AA7PnE,YAAAC;AA+PQ,YAAI,QAAQ,gBAAgB,OAAQ;AAEpC,cAAM,eAAe,IAAI,sBAAsB,sBAAsB;AACrE,qBAAa,GAAG,eAAe,OAAO,SAAS;AAC7C,gBAAM,KAAK;AAAA,YACT,KAAK,KAAM,iBAAkB;AAAA,YAC7B,KAAK,kBAAkB;AAAA,YACvB,KAAK;AAAA,YACL,KAAK;AAAA,YACL,KAAK;AAAA,UACP;AAAA,QACF,CAAC;AAED,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,YAAiB;AAE1D,YAAI,QAAQ,gBAAgB,QAAQ;AAClC,cAAI,KAAK,wBAAwB,KAAK,yBAAyB;AAC7D,kBAAM,IAAI;AAAA,cACR,0DACW,KAAK,uBAAuB;AAAA,YAGzC;AAAA,UACF;AAEA,eAAK;AACL,eAAK,QACF,MAAM;AAAA,YACL,QAAQ,QAAQ;AAAA,YAChB,MAAM,QAAQ;AAAA,YACd,SAAS,KAAK;AAAA,UAChB,CAAC,EACA;AAAA,YACC;AAAA,UAEF;AACF,eAAK,SAAU,wBAAwB,QAAQ,MAAM;AAAA,QACvD,OAAO;AACL,eAAK,uBAAuB;AAAA,QAC9B;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,OAAO,OAAY;AArTpE,YAAAA,KAAA;AAuTQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,gBAAM,KAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACvF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,OAAO,OAAY;AAjUlF,YAAAA,KAAA;AAmUQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,gBAAM,KAAK;AAAA,YACT;AAAA,YACA;AAAA,YACA;AAAA,YACA;AAAA,YACA,GAAG;AAAA,UACL;AAAA,QACF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,OAAO,OAAY;AAzVlE,YAAAA,KAAA;AA0VQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,gBAAM,KAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACvF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AAhcJ;AAicI,QACE,YAAY,WAAW,YAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,YAAM,qBAAqB;AAAA,QACzB,YAAY,KAAK,MAAM;AAAA,QACvB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,qBAAqB,EAAE,mBAAmB,KAAK,mBAAmB,IAAI,CAAC;AAAA,MAClF;AACA,0BAAoB,IAAI,YAAY,OAAO,kBAAkB,CAAC,EAAE,KAAK,OAAO,EAAE,MAAM,MAAM;AAAA,IAC5F,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AACjC,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,iBAAiB,iBAAiB,KAAK,MAAM,KAAK,KAAK,iBAAkB,QAAS;AAAA,IACzF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,sBACJ,qBACA,UACA,MACA,SACA,IACe;AA9enB;AA+eI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAED,UAAM,SAAS,MAAM,KAAK,KAAK,iBAAiB,WAAW;AAAA,MACzD,OAAO;AAAA,MACP,gBAAgB;AAAA,MAChB,YAAY;AAAA,QACV,CAAC,gCAAgC,GAAG;AAAA,QACpC,CAAC,6BAA6B,GAAG,QAAQ,SAAS;AAAA,MACpD;AAAA,IACF,CAAC;AACD,UAAM,OAAO,MAAM,IAAI;AACvB,UAAM,OAAO,MAAM;AAAA,EACrB;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AA/hB/B;AAgiBI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAY,qBAAqB;AACjF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
@@ -1,197 +0,0 @@
1
- "use strict";
2
- var __defProp = Object.defineProperty;
3
- var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
4
- var __getOwnPropNames = Object.getOwnPropertyNames;
5
- var __hasOwnProp = Object.prototype.hasOwnProperty;
6
- var __export = (target, all) => {
7
- for (var name in all)
8
- __defProp(target, name, { get: all[name], enumerable: true });
9
- };
10
- var __copyProps = (to, from, except, desc) => {
11
- if (from && typeof from === "object" || typeof from === "function") {
12
- for (let key of __getOwnPropNames(from))
13
- if (!__hasOwnProp.call(to, key) && key !== except)
14
- __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
15
- }
16
- return to;
17
- };
18
- var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
19
- var agent_output_exports = {};
20
- __export(agent_output_exports, {
21
- AgentOutput: () => AgentOutput,
22
- SynthesisHandle: () => SynthesisHandle
23
- });
24
- module.exports = __toCommonJS(agent_output_exports);
25
- var import_log = require("../log.cjs");
26
- var import_tts = require("../tts/index.cjs");
27
- var import_utils = require("../utils.cjs");
28
- class SynthesisHandle {
29
- static FLUSH_SENTINEL = Symbol("FLUSH_SENTINEL");
30
- #speechId;
31
- text;
32
- ttsSource;
33
- #agentPlayout;
34
- tts;
35
- queue = new import_utils.AsyncIterableQueue();
36
- #playHandle;
37
- intFut = new import_utils.Future();
38
- #logger = (0, import_log.log)();
39
- synchronizer;
40
- constructor(speechId, ttsSource, agentPlayout, tts, synchronizer) {
41
- this.#speechId = speechId;
42
- this.ttsSource = ttsSource;
43
- this.#agentPlayout = agentPlayout;
44
- this.tts = tts;
45
- this.synchronizer = synchronizer;
46
- }
47
- get speechId() {
48
- return this.#speechId;
49
- }
50
- get validated() {
51
- return !!this.#playHandle;
52
- }
53
- get interrupted() {
54
- return this.intFut.done;
55
- }
56
- get playHandle() {
57
- return this.#playHandle;
58
- }
59
- /** Validate the speech for playout. */
60
- play() {
61
- if (this.interrupted) {
62
- throw new Error("synthesis was interrupted");
63
- }
64
- this.#playHandle = this.#agentPlayout.play(this.#speechId, this.queue, this.synchronizer);
65
- return this.#playHandle;
66
- }
67
- /** Interrupt the speech. */
68
- interrupt() {
69
- var _a;
70
- if (this.interrupted) {
71
- return;
72
- }
73
- this.#logger.child({ speechId: this.#speechId }).debug("interrupting synthesis/playout");
74
- (_a = this.#playHandle) == null ? void 0 : _a.interrupt();
75
- this.intFut.resolve();
76
- }
77
- }
78
- class AgentOutput {
79
- #agentPlayout;
80
- #tts;
81
- #tasks = [];
82
- constructor(agentPlayout, tts) {
83
- this.#agentPlayout = agentPlayout;
84
- this.#tts = tts;
85
- }
86
- get playout() {
87
- return this.#agentPlayout;
88
- }
89
- async close() {
90
- this.#tasks.forEach((task) => task.cancel());
91
- await Promise.all(this.#tasks);
92
- }
93
- synthesize(speechId, ttsSource, synchronizer) {
94
- const handle = new SynthesisHandle(
95
- speechId,
96
- ttsSource,
97
- this.#agentPlayout,
98
- this.#tts,
99
- synchronizer
100
- );
101
- const task = this.#synthesize(handle);
102
- this.#tasks.push(task);
103
- task.finally(() => this.#tasks.splice(this.#tasks.indexOf(task)));
104
- return handle;
105
- }
106
- #synthesize(handle) {
107
- return new import_utils.CancellablePromise(async (resolve, _, onCancel) => {
108
- const ttsSource = await handle.ttsSource;
109
- let task;
110
- if (typeof ttsSource === "string") {
111
- task = stringSynthesisTask(ttsSource, handle);
112
- } else {
113
- task = streamSynthesisTask(ttsSource, handle);
114
- }
115
- onCancel(() => {
116
- (0, import_utils.gracefullyCancel)(task);
117
- });
118
- try {
119
- await Promise.any([task, handle.intFut.await]);
120
- } finally {
121
- if (handle.intFut.done) {
122
- (0, import_utils.gracefullyCancel)(task);
123
- } else {
124
- task.then((text) => {
125
- handle.text = text;
126
- });
127
- }
128
- }
129
- resolve();
130
- });
131
- }
132
- }
133
- const stringSynthesisTask = (text, handle) => {
134
- return new import_utils.CancellablePromise(async (resolve, _, onCancel) => {
135
- let cancelled = false;
136
- onCancel(() => {
137
- cancelled = true;
138
- });
139
- const ttsStream = handle.tts.stream();
140
- ttsStream.pushText(text);
141
- handle.synchronizer.pushText(text);
142
- handle.synchronizer.markTextSegmentEnd();
143
- ttsStream.flush();
144
- ttsStream.endInput();
145
- for await (const audio of ttsStream) {
146
- if (cancelled || audio === import_tts.SynthesizeStream.END_OF_STREAM) {
147
- break;
148
- }
149
- handle.queue.put(audio.frame);
150
- }
151
- handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);
152
- resolve(text);
153
- });
154
- };
155
- const streamSynthesisTask = (stream, handle) => {
156
- return new import_utils.CancellablePromise(async (resolve, _, onCancel) => {
157
- let fullText = "";
158
- let cancelled = false;
159
- onCancel(() => {
160
- cancelled = true;
161
- });
162
- const ttsStream = handle.tts.stream();
163
- const readGeneratedAudio = async () => {
164
- for await (const audio of ttsStream) {
165
- if (cancelled) break;
166
- if (audio === import_tts.SynthesizeStream.END_OF_STREAM) {
167
- break;
168
- }
169
- handle.queue.put(audio.frame);
170
- }
171
- handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);
172
- };
173
- readGeneratedAudio();
174
- for await (const text of stream) {
175
- fullText += text;
176
- if (cancelled) break;
177
- handle.synchronizer.pushText(text);
178
- ttsStream.pushText(text);
179
- }
180
- if (!cancelled) {
181
- handle.synchronizer.markTextSegmentEnd();
182
- }
183
- if (!fullText || fullText.trim().length === 0) {
184
- cancelled = true;
185
- handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);
186
- }
187
- ttsStream.flush();
188
- ttsStream.endInput();
189
- resolve(fullText);
190
- });
191
- };
192
- // Annotate the CommonJS export names for ESM import in node:
193
- 0 && (module.exports = {
194
- AgentOutput,
195
- SynthesisHandle
196
- });
197
- //# sourceMappingURL=agent_output.cjs.map