@livekit/agents 1.1.0 → 1.2.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (959) hide show
  1. package/dist/_exceptions.cjs.map +1 -1
  2. package/dist/_exceptions.d.ts.map +1 -1
  3. package/dist/_exceptions.js.map +1 -1
  4. package/dist/audio.cjs +89 -3
  5. package/dist/audio.cjs.map +1 -1
  6. package/dist/audio.d.cts +36 -1
  7. package/dist/audio.d.ts +36 -1
  8. package/dist/audio.d.ts.map +1 -1
  9. package/dist/audio.js +76 -2
  10. package/dist/audio.js.map +1 -1
  11. package/dist/beta/index.cjs +29 -0
  12. package/dist/beta/index.cjs.map +1 -0
  13. package/dist/beta/index.d.cts +2 -0
  14. package/dist/beta/index.d.ts +2 -0
  15. package/dist/beta/index.d.ts.map +1 -0
  16. package/dist/beta/index.js +7 -0
  17. package/dist/beta/index.js.map +1 -0
  18. package/dist/beta/workflows/index.cjs +29 -0
  19. package/dist/beta/workflows/index.cjs.map +1 -0
  20. package/dist/beta/workflows/index.d.cts +2 -0
  21. package/dist/beta/workflows/index.d.ts +2 -0
  22. package/dist/beta/workflows/index.d.ts.map +1 -0
  23. package/dist/beta/workflows/index.js +7 -0
  24. package/dist/beta/workflows/index.js.map +1 -0
  25. package/dist/beta/workflows/task_group.cjs +165 -0
  26. package/dist/beta/workflows/task_group.cjs.map +1 -0
  27. package/dist/beta/workflows/task_group.d.cts +32 -0
  28. package/dist/beta/workflows/task_group.d.ts +32 -0
  29. package/dist/beta/workflows/task_group.d.ts.map +1 -0
  30. package/dist/beta/workflows/task_group.js +141 -0
  31. package/dist/beta/workflows/task_group.js.map +1 -0
  32. package/dist/cli.cjs +44 -46
  33. package/dist/cli.cjs.map +1 -1
  34. package/dist/cli.d.cts +3 -3
  35. package/dist/cli.d.ts +3 -3
  36. package/dist/cli.d.ts.map +1 -1
  37. package/dist/cli.js +45 -47
  38. package/dist/cli.js.map +1 -1
  39. package/dist/connection_pool.cjs +242 -0
  40. package/dist/connection_pool.cjs.map +1 -0
  41. package/dist/connection_pool.d.cts +123 -0
  42. package/dist/connection_pool.d.ts +123 -0
  43. package/dist/connection_pool.d.ts.map +1 -0
  44. package/dist/connection_pool.js +218 -0
  45. package/dist/connection_pool.js.map +1 -0
  46. package/dist/connection_pool.test.cjs +256 -0
  47. package/dist/connection_pool.test.cjs.map +1 -0
  48. package/dist/connection_pool.test.js +255 -0
  49. package/dist/connection_pool.test.js.map +1 -0
  50. package/dist/constants.cjs +30 -0
  51. package/dist/constants.cjs.map +1 -1
  52. package/dist/constants.d.cts +10 -0
  53. package/dist/constants.d.ts +10 -0
  54. package/dist/constants.d.ts.map +1 -1
  55. package/dist/constants.js +20 -0
  56. package/dist/constants.js.map +1 -1
  57. package/dist/cpu.cjs +189 -0
  58. package/dist/cpu.cjs.map +1 -0
  59. package/dist/cpu.d.cts +24 -0
  60. package/dist/cpu.d.ts +24 -0
  61. package/dist/cpu.d.ts.map +1 -0
  62. package/dist/cpu.js +152 -0
  63. package/dist/cpu.js.map +1 -0
  64. package/dist/cpu.test.cjs +227 -0
  65. package/dist/cpu.test.cjs.map +1 -0
  66. package/dist/cpu.test.js +204 -0
  67. package/dist/cpu.test.js.map +1 -0
  68. package/dist/http_server.cjs +9 -6
  69. package/dist/http_server.cjs.map +1 -1
  70. package/dist/http_server.d.cts +5 -1
  71. package/dist/http_server.d.ts +5 -1
  72. package/dist/http_server.d.ts.map +1 -1
  73. package/dist/http_server.js +9 -6
  74. package/dist/http_server.js.map +1 -1
  75. package/dist/index.cjs +24 -9
  76. package/dist/index.cjs.map +1 -1
  77. package/dist/index.d.cts +15 -11
  78. package/dist/index.d.ts +15 -11
  79. package/dist/index.d.ts.map +1 -1
  80. package/dist/index.js +18 -9
  81. package/dist/index.js.map +1 -1
  82. package/dist/inference/api_protos.cjs +70 -2
  83. package/dist/inference/api_protos.cjs.map +1 -1
  84. package/dist/inference/api_protos.d.cts +373 -32
  85. package/dist/inference/api_protos.d.ts +373 -32
  86. package/dist/inference/api_protos.d.ts.map +1 -1
  87. package/dist/inference/api_protos.js +62 -2
  88. package/dist/inference/api_protos.js.map +1 -1
  89. package/dist/inference/index.cjs +8 -0
  90. package/dist/inference/index.cjs.map +1 -1
  91. package/dist/inference/index.d.cts +3 -4
  92. package/dist/inference/index.d.ts +3 -4
  93. package/dist/inference/index.d.ts.map +1 -1
  94. package/dist/inference/index.js +18 -3
  95. package/dist/inference/index.js.map +1 -1
  96. package/dist/inference/interruption/defaults.cjs +81 -0
  97. package/dist/inference/interruption/defaults.cjs.map +1 -0
  98. package/dist/inference/interruption/defaults.d.cts +19 -0
  99. package/dist/inference/interruption/defaults.d.ts +19 -0
  100. package/dist/inference/interruption/defaults.d.ts.map +1 -0
  101. package/dist/inference/interruption/defaults.js +46 -0
  102. package/dist/inference/interruption/defaults.js.map +1 -0
  103. package/dist/inference/interruption/errors.cjs +44 -0
  104. package/dist/inference/interruption/errors.cjs.map +1 -0
  105. package/dist/inference/interruption/errors.d.cts +12 -0
  106. package/dist/inference/interruption/errors.d.ts +12 -0
  107. package/dist/inference/interruption/errors.d.ts.map +1 -0
  108. package/dist/inference/interruption/errors.js +20 -0
  109. package/dist/inference/interruption/errors.js.map +1 -0
  110. package/dist/inference/interruption/http_transport.cjs +163 -0
  111. package/dist/inference/interruption/http_transport.cjs.map +1 -0
  112. package/dist/inference/interruption/http_transport.d.cts +65 -0
  113. package/dist/inference/interruption/http_transport.d.ts +65 -0
  114. package/dist/inference/interruption/http_transport.d.ts.map +1 -0
  115. package/dist/inference/interruption/http_transport.js +137 -0
  116. package/dist/inference/interruption/http_transport.js.map +1 -0
  117. package/dist/inference/interruption/interruption_cache_entry.cjs +58 -0
  118. package/dist/inference/interruption/interruption_cache_entry.cjs.map +1 -0
  119. package/dist/inference/interruption/interruption_cache_entry.d.cts +30 -0
  120. package/dist/inference/interruption/interruption_cache_entry.d.ts +30 -0
  121. package/dist/inference/interruption/interruption_cache_entry.d.ts.map +1 -0
  122. package/dist/inference/interruption/interruption_cache_entry.js +34 -0
  123. package/dist/inference/interruption/interruption_cache_entry.js.map +1 -0
  124. package/dist/inference/interruption/interruption_detector.cjs +198 -0
  125. package/dist/inference/interruption/interruption_detector.cjs.map +1 -0
  126. package/dist/inference/interruption/interruption_detector.d.cts +59 -0
  127. package/dist/inference/interruption/interruption_detector.d.ts +59 -0
  128. package/dist/inference/interruption/interruption_detector.d.ts.map +1 -0
  129. package/dist/inference/interruption/interruption_detector.js +164 -0
  130. package/dist/inference/interruption/interruption_detector.js.map +1 -0
  131. package/dist/inference/interruption/interruption_stream.cjs +368 -0
  132. package/dist/inference/interruption/interruption_stream.cjs.map +1 -0
  133. package/dist/inference/interruption/interruption_stream.d.cts +46 -0
  134. package/dist/inference/interruption/interruption_stream.d.ts +46 -0
  135. package/dist/inference/interruption/interruption_stream.d.ts.map +1 -0
  136. package/dist/inference/interruption/interruption_stream.js +344 -0
  137. package/dist/inference/interruption/interruption_stream.js.map +1 -0
  138. package/dist/inference/interruption/types.cjs +17 -0
  139. package/dist/inference/interruption/types.cjs.map +1 -0
  140. package/dist/inference/interruption/types.d.cts +66 -0
  141. package/dist/inference/interruption/types.d.ts +66 -0
  142. package/dist/inference/interruption/types.d.ts.map +1 -0
  143. package/dist/inference/interruption/types.js +1 -0
  144. package/dist/inference/interruption/types.js.map +1 -0
  145. package/dist/inference/interruption/utils.cjs +130 -0
  146. package/dist/inference/interruption/utils.cjs.map +1 -0
  147. package/dist/inference/interruption/utils.d.cts +41 -0
  148. package/dist/inference/interruption/utils.d.ts +41 -0
  149. package/dist/inference/interruption/utils.d.ts.map +1 -0
  150. package/dist/inference/interruption/utils.js +105 -0
  151. package/dist/inference/interruption/utils.js.map +1 -0
  152. package/dist/inference/interruption/utils.test.cjs +105 -0
  153. package/dist/inference/interruption/utils.test.cjs.map +1 -0
  154. package/dist/inference/interruption/utils.test.js +104 -0
  155. package/dist/inference/interruption/utils.test.js.map +1 -0
  156. package/dist/inference/interruption/ws_transport.cjs +347 -0
  157. package/dist/inference/interruption/ws_transport.cjs.map +1 -0
  158. package/dist/inference/interruption/ws_transport.d.cts +33 -0
  159. package/dist/inference/interruption/ws_transport.d.ts +33 -0
  160. package/dist/inference/interruption/ws_transport.d.ts.map +1 -0
  161. package/dist/inference/interruption/ws_transport.js +313 -0
  162. package/dist/inference/interruption/ws_transport.js.map +1 -0
  163. package/dist/inference/llm.cjs +106 -66
  164. package/dist/inference/llm.cjs.map +1 -1
  165. package/dist/inference/llm.d.cts +65 -43
  166. package/dist/inference/llm.d.ts +65 -43
  167. package/dist/inference/llm.d.ts.map +1 -1
  168. package/dist/inference/llm.js +100 -66
  169. package/dist/inference/llm.js.map +1 -1
  170. package/dist/inference/stt.cjs +319 -170
  171. package/dist/inference/stt.cjs.map +1 -1
  172. package/dist/inference/stt.d.cts +64 -15
  173. package/dist/inference/stt.d.ts +64 -15
  174. package/dist/inference/stt.d.ts.map +1 -1
  175. package/dist/inference/stt.js +319 -170
  176. package/dist/inference/stt.js.map +1 -1
  177. package/dist/inference/stt.test.cjs +218 -0
  178. package/dist/inference/stt.test.cjs.map +1 -0
  179. package/dist/inference/stt.test.js +217 -0
  180. package/dist/inference/stt.test.js.map +1 -0
  181. package/dist/inference/tts.cjs +249 -71
  182. package/dist/inference/tts.cjs.map +1 -1
  183. package/dist/inference/tts.d.cts +94 -17
  184. package/dist/inference/tts.d.ts +94 -17
  185. package/dist/inference/tts.d.ts.map +1 -1
  186. package/dist/inference/tts.js +249 -77
  187. package/dist/inference/tts.js.map +1 -1
  188. package/dist/inference/tts.test.cjs +305 -0
  189. package/dist/inference/tts.test.cjs.map +1 -0
  190. package/dist/inference/tts.test.js +304 -0
  191. package/dist/inference/tts.test.js.map +1 -0
  192. package/dist/inference/utils.cjs +26 -7
  193. package/dist/inference/utils.cjs.map +1 -1
  194. package/dist/inference/utils.d.cts +14 -1
  195. package/dist/inference/utils.d.ts +14 -1
  196. package/dist/inference/utils.d.ts.map +1 -1
  197. package/dist/inference/utils.js +18 -2
  198. package/dist/inference/utils.js.map +1 -1
  199. package/dist/ipc/inference_proc_executor.cjs +6 -3
  200. package/dist/ipc/inference_proc_executor.cjs.map +1 -1
  201. package/dist/ipc/inference_proc_executor.d.ts.map +1 -1
  202. package/dist/ipc/inference_proc_executor.js +6 -3
  203. package/dist/ipc/inference_proc_executor.js.map +1 -1
  204. package/dist/ipc/inference_proc_lazy_main.cjs +13 -1
  205. package/dist/ipc/inference_proc_lazy_main.cjs.map +1 -1
  206. package/dist/ipc/inference_proc_lazy_main.js +13 -1
  207. package/dist/ipc/inference_proc_lazy_main.js.map +1 -1
  208. package/dist/ipc/job_proc_executor.cjs +6 -1
  209. package/dist/ipc/job_proc_executor.cjs.map +1 -1
  210. package/dist/ipc/job_proc_executor.d.ts.map +1 -1
  211. package/dist/ipc/job_proc_executor.js +6 -1
  212. package/dist/ipc/job_proc_executor.js.map +1 -1
  213. package/dist/ipc/job_proc_lazy_main.cjs +89 -17
  214. package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
  215. package/dist/ipc/job_proc_lazy_main.js +68 -18
  216. package/dist/ipc/job_proc_lazy_main.js.map +1 -1
  217. package/dist/ipc/supervised_proc.cjs +34 -8
  218. package/dist/ipc/supervised_proc.cjs.map +1 -1
  219. package/dist/ipc/supervised_proc.d.cts +8 -0
  220. package/dist/ipc/supervised_proc.d.ts +8 -0
  221. package/dist/ipc/supervised_proc.d.ts.map +1 -1
  222. package/dist/ipc/supervised_proc.js +34 -8
  223. package/dist/ipc/supervised_proc.js.map +1 -1
  224. package/dist/ipc/supervised_proc.test.cjs +145 -0
  225. package/dist/ipc/supervised_proc.test.cjs.map +1 -0
  226. package/dist/ipc/supervised_proc.test.js +122 -0
  227. package/dist/ipc/supervised_proc.test.js.map +1 -0
  228. package/dist/job.cjs +109 -1
  229. package/dist/job.cjs.map +1 -1
  230. package/dist/job.d.cts +14 -0
  231. package/dist/job.d.ts +14 -0
  232. package/dist/job.d.ts.map +1 -1
  233. package/dist/job.js +99 -1
  234. package/dist/job.js.map +1 -1
  235. package/dist/language.cjs +394 -0
  236. package/dist/language.cjs.map +1 -0
  237. package/dist/language.d.cts +15 -0
  238. package/dist/language.d.ts +15 -0
  239. package/dist/language.d.ts.map +1 -0
  240. package/dist/language.js +363 -0
  241. package/dist/language.js.map +1 -0
  242. package/dist/language.test.cjs +43 -0
  243. package/dist/language.test.cjs.map +1 -0
  244. package/dist/language.test.js +49 -0
  245. package/dist/language.test.js.map +1 -0
  246. package/dist/llm/chat_context.cjs +345 -3
  247. package/dist/llm/chat_context.cjs.map +1 -1
  248. package/dist/llm/chat_context.d.cts +86 -2
  249. package/dist/llm/chat_context.d.ts +86 -2
  250. package/dist/llm/chat_context.d.ts.map +1 -1
  251. package/dist/llm/chat_context.js +344 -3
  252. package/dist/llm/chat_context.js.map +1 -1
  253. package/dist/llm/chat_context.test.cjs +692 -0
  254. package/dist/llm/chat_context.test.cjs.map +1 -1
  255. package/dist/llm/chat_context.test.js +692 -0
  256. package/dist/llm/chat_context.test.js.map +1 -1
  257. package/dist/llm/fallback_adapter.cjs +280 -0
  258. package/dist/llm/fallback_adapter.cjs.map +1 -0
  259. package/dist/llm/fallback_adapter.d.cts +73 -0
  260. package/dist/llm/fallback_adapter.d.ts +73 -0
  261. package/dist/llm/fallback_adapter.d.ts.map +1 -0
  262. package/dist/llm/fallback_adapter.js +256 -0
  263. package/dist/llm/fallback_adapter.js.map +1 -0
  264. package/dist/llm/fallback_adapter.test.cjs +176 -0
  265. package/dist/llm/fallback_adapter.test.cjs.map +1 -0
  266. package/dist/llm/fallback_adapter.test.js +175 -0
  267. package/dist/llm/fallback_adapter.test.js.map +1 -0
  268. package/dist/llm/index.cjs +11 -0
  269. package/dist/llm/index.cjs.map +1 -1
  270. package/dist/llm/index.d.cts +4 -3
  271. package/dist/llm/index.d.ts +4 -3
  272. package/dist/llm/index.d.ts.map +1 -1
  273. package/dist/llm/index.js +13 -1
  274. package/dist/llm/index.js.map +1 -1
  275. package/dist/llm/llm.cjs +65 -11
  276. package/dist/llm/llm.cjs.map +1 -1
  277. package/dist/llm/llm.d.cts +13 -2
  278. package/dist/llm/llm.d.ts +13 -2
  279. package/dist/llm/llm.d.ts.map +1 -1
  280. package/dist/llm/llm.js +65 -11
  281. package/dist/llm/llm.js.map +1 -1
  282. package/dist/llm/provider_format/google.cjs +6 -2
  283. package/dist/llm/provider_format/google.cjs.map +1 -1
  284. package/dist/llm/provider_format/google.d.cts +1 -1
  285. package/dist/llm/provider_format/google.d.ts +1 -1
  286. package/dist/llm/provider_format/google.d.ts.map +1 -1
  287. package/dist/llm/provider_format/google.js +6 -2
  288. package/dist/llm/provider_format/google.js.map +1 -1
  289. package/dist/llm/provider_format/google.test.cjs +48 -0
  290. package/dist/llm/provider_format/google.test.cjs.map +1 -1
  291. package/dist/llm/provider_format/google.test.js +54 -1
  292. package/dist/llm/provider_format/google.test.js.map +1 -1
  293. package/dist/llm/provider_format/index.cjs +2 -0
  294. package/dist/llm/provider_format/index.cjs.map +1 -1
  295. package/dist/llm/provider_format/index.d.cts +2 -2
  296. package/dist/llm/provider_format/index.d.ts +2 -2
  297. package/dist/llm/provider_format/index.d.ts.map +1 -1
  298. package/dist/llm/provider_format/index.js +6 -1
  299. package/dist/llm/provider_format/index.js.map +1 -1
  300. package/dist/llm/provider_format/openai.cjs +126 -24
  301. package/dist/llm/provider_format/openai.cjs.map +1 -1
  302. package/dist/llm/provider_format/openai.d.cts +1 -0
  303. package/dist/llm/provider_format/openai.d.ts +1 -0
  304. package/dist/llm/provider_format/openai.d.ts.map +1 -1
  305. package/dist/llm/provider_format/openai.js +124 -23
  306. package/dist/llm/provider_format/openai.js.map +1 -1
  307. package/dist/llm/provider_format/openai.test.cjs +393 -0
  308. package/dist/llm/provider_format/openai.test.cjs.map +1 -1
  309. package/dist/llm/provider_format/openai.test.js +400 -2
  310. package/dist/llm/provider_format/openai.test.js.map +1 -1
  311. package/dist/llm/provider_format/utils.cjs +5 -4
  312. package/dist/llm/provider_format/utils.cjs.map +1 -1
  313. package/dist/llm/provider_format/utils.d.ts.map +1 -1
  314. package/dist/llm/provider_format/utils.js +5 -4
  315. package/dist/llm/provider_format/utils.js.map +1 -1
  316. package/dist/llm/realtime.cjs +3 -0
  317. package/dist/llm/realtime.cjs.map +1 -1
  318. package/dist/llm/realtime.d.cts +15 -1
  319. package/dist/llm/realtime.d.ts +15 -1
  320. package/dist/llm/realtime.d.ts.map +1 -1
  321. package/dist/llm/realtime.js +3 -0
  322. package/dist/llm/realtime.js.map +1 -1
  323. package/dist/llm/remote_chat_context.cjs.map +1 -1
  324. package/dist/llm/remote_chat_context.d.cts +2 -0
  325. package/dist/llm/remote_chat_context.d.ts +2 -0
  326. package/dist/llm/remote_chat_context.d.ts.map +1 -1
  327. package/dist/llm/remote_chat_context.js.map +1 -1
  328. package/dist/llm/tool_context.cjs +50 -2
  329. package/dist/llm/tool_context.cjs.map +1 -1
  330. package/dist/llm/tool_context.d.cts +47 -11
  331. package/dist/llm/tool_context.d.ts +47 -11
  332. package/dist/llm/tool_context.d.ts.map +1 -1
  333. package/dist/llm/tool_context.js +48 -3
  334. package/dist/llm/tool_context.js.map +1 -1
  335. package/dist/llm/tool_context.test.cjs +197 -0
  336. package/dist/llm/tool_context.test.cjs.map +1 -1
  337. package/dist/llm/tool_context.test.js +175 -0
  338. package/dist/llm/tool_context.test.js.map +1 -1
  339. package/dist/llm/utils.cjs +107 -12
  340. package/dist/llm/utils.cjs.map +1 -1
  341. package/dist/llm/utils.d.cts +10 -3
  342. package/dist/llm/utils.d.ts +10 -3
  343. package/dist/llm/utils.d.ts.map +1 -1
  344. package/dist/llm/utils.js +106 -12
  345. package/dist/llm/utils.js.map +1 -1
  346. package/dist/llm/utils.test.cjs +90 -0
  347. package/dist/llm/utils.test.cjs.map +1 -1
  348. package/dist/llm/utils.test.js +98 -2
  349. package/dist/llm/utils.test.js.map +1 -1
  350. package/dist/llm/zod-utils.cjs +102 -0
  351. package/dist/llm/zod-utils.cjs.map +1 -0
  352. package/dist/llm/zod-utils.d.cts +65 -0
  353. package/dist/llm/zod-utils.d.ts +65 -0
  354. package/dist/llm/zod-utils.d.ts.map +1 -0
  355. package/dist/llm/zod-utils.js +64 -0
  356. package/dist/llm/zod-utils.js.map +1 -0
  357. package/dist/llm/zod-utils.test.cjs +472 -0
  358. package/dist/llm/zod-utils.test.cjs.map +1 -0
  359. package/dist/llm/zod-utils.test.js +455 -0
  360. package/dist/llm/zod-utils.test.js.map +1 -0
  361. package/dist/log.cjs +45 -14
  362. package/dist/log.cjs.map +1 -1
  363. package/dist/log.d.cts +8 -1
  364. package/dist/log.d.ts +8 -1
  365. package/dist/log.d.ts.map +1 -1
  366. package/dist/log.js +45 -15
  367. package/dist/log.js.map +1 -1
  368. package/dist/metrics/base.cjs.map +1 -1
  369. package/dist/metrics/base.d.cts +75 -19
  370. package/dist/metrics/base.d.ts +75 -19
  371. package/dist/metrics/base.d.ts.map +1 -1
  372. package/dist/metrics/index.cjs +5 -0
  373. package/dist/metrics/index.cjs.map +1 -1
  374. package/dist/metrics/index.d.cts +2 -1
  375. package/dist/metrics/index.d.ts +2 -1
  376. package/dist/metrics/index.d.ts.map +1 -1
  377. package/dist/metrics/index.js +6 -0
  378. package/dist/metrics/index.js.map +1 -1
  379. package/dist/metrics/model_usage.cjs +189 -0
  380. package/dist/metrics/model_usage.cjs.map +1 -0
  381. package/dist/metrics/model_usage.d.cts +92 -0
  382. package/dist/metrics/model_usage.d.ts +92 -0
  383. package/dist/metrics/model_usage.d.ts.map +1 -0
  384. package/dist/metrics/model_usage.js +164 -0
  385. package/dist/metrics/model_usage.js.map +1 -0
  386. package/dist/metrics/model_usage.test.cjs +474 -0
  387. package/dist/metrics/model_usage.test.cjs.map +1 -0
  388. package/dist/metrics/model_usage.test.js +476 -0
  389. package/dist/metrics/model_usage.test.js.map +1 -0
  390. package/dist/metrics/usage_collector.cjs +5 -2
  391. package/dist/metrics/usage_collector.cjs.map +1 -1
  392. package/dist/metrics/usage_collector.d.cts +10 -1
  393. package/dist/metrics/usage_collector.d.ts +10 -1
  394. package/dist/metrics/usage_collector.d.ts.map +1 -1
  395. package/dist/metrics/usage_collector.js +5 -2
  396. package/dist/metrics/usage_collector.js.map +1 -1
  397. package/dist/metrics/utils.cjs +23 -7
  398. package/dist/metrics/utils.cjs.map +1 -1
  399. package/dist/metrics/utils.d.ts.map +1 -1
  400. package/dist/metrics/utils.js +23 -7
  401. package/dist/metrics/utils.js.map +1 -1
  402. package/dist/stream/deferred_stream.cjs +31 -10
  403. package/dist/stream/deferred_stream.cjs.map +1 -1
  404. package/dist/stream/deferred_stream.d.cts +6 -1
  405. package/dist/stream/deferred_stream.d.ts +6 -1
  406. package/dist/stream/deferred_stream.d.ts.map +1 -1
  407. package/dist/stream/deferred_stream.js +31 -10
  408. package/dist/stream/deferred_stream.js.map +1 -1
  409. package/dist/stream/deferred_stream.test.cjs +2 -2
  410. package/dist/stream/deferred_stream.test.cjs.map +1 -1
  411. package/dist/stream/deferred_stream.test.js +2 -2
  412. package/dist/stream/deferred_stream.test.js.map +1 -1
  413. package/dist/stream/index.cjs +3 -0
  414. package/dist/stream/index.cjs.map +1 -1
  415. package/dist/stream/index.d.cts +1 -0
  416. package/dist/stream/index.d.ts +1 -0
  417. package/dist/stream/index.d.ts.map +1 -1
  418. package/dist/stream/index.js +2 -0
  419. package/dist/stream/index.js.map +1 -1
  420. package/dist/stream/multi_input_stream.cjs +139 -0
  421. package/dist/stream/multi_input_stream.cjs.map +1 -0
  422. package/dist/stream/multi_input_stream.d.cts +55 -0
  423. package/dist/stream/multi_input_stream.d.ts +55 -0
  424. package/dist/stream/multi_input_stream.d.ts.map +1 -0
  425. package/dist/stream/multi_input_stream.js +115 -0
  426. package/dist/stream/multi_input_stream.js.map +1 -0
  427. package/dist/stream/multi_input_stream.test.cjs +344 -0
  428. package/dist/stream/multi_input_stream.test.cjs.map +1 -0
  429. package/dist/stream/multi_input_stream.test.js +343 -0
  430. package/dist/stream/multi_input_stream.test.js.map +1 -0
  431. package/dist/stream/stream_channel.cjs +39 -1
  432. package/dist/stream/stream_channel.cjs.map +1 -1
  433. package/dist/stream/stream_channel.d.cts +5 -2
  434. package/dist/stream/stream_channel.d.ts +5 -2
  435. package/dist/stream/stream_channel.d.ts.map +1 -1
  436. package/dist/stream/stream_channel.js +39 -1
  437. package/dist/stream/stream_channel.js.map +1 -1
  438. package/dist/stream/stream_channel.test.cjs +27 -0
  439. package/dist/stream/stream_channel.test.cjs.map +1 -1
  440. package/dist/stream/stream_channel.test.js +27 -0
  441. package/dist/stream/stream_channel.test.js.map +1 -1
  442. package/dist/stt/stream_adapter.cjs +24 -9
  443. package/dist/stt/stream_adapter.cjs.map +1 -1
  444. package/dist/stt/stream_adapter.d.cts +7 -3
  445. package/dist/stt/stream_adapter.d.ts +7 -3
  446. package/dist/stt/stream_adapter.d.ts.map +1 -1
  447. package/dist/stt/stream_adapter.js +24 -9
  448. package/dist/stt/stream_adapter.js.map +1 -1
  449. package/dist/stt/stt.cjs +94 -19
  450. package/dist/stt/stt.cjs.map +1 -1
  451. package/dist/stt/stt.d.cts +68 -5
  452. package/dist/stt/stt.d.ts +68 -5
  453. package/dist/stt/stt.d.ts.map +1 -1
  454. package/dist/stt/stt.js +96 -21
  455. package/dist/stt/stt.js.map +1 -1
  456. package/dist/telemetry/index.cjs +72 -0
  457. package/dist/telemetry/index.cjs.map +1 -0
  458. package/dist/telemetry/index.d.cts +7 -0
  459. package/dist/telemetry/index.d.ts +7 -0
  460. package/dist/telemetry/index.d.ts.map +1 -0
  461. package/dist/telemetry/index.js +37 -0
  462. package/dist/telemetry/index.js.map +1 -0
  463. package/dist/telemetry/logging.cjs +65 -0
  464. package/dist/telemetry/logging.cjs.map +1 -0
  465. package/dist/telemetry/logging.d.cts +21 -0
  466. package/dist/telemetry/logging.d.ts +21 -0
  467. package/dist/telemetry/logging.d.ts.map +1 -0
  468. package/dist/telemetry/logging.js +40 -0
  469. package/dist/telemetry/logging.js.map +1 -0
  470. package/dist/telemetry/otel_http_exporter.cjs +166 -0
  471. package/dist/telemetry/otel_http_exporter.cjs.map +1 -0
  472. package/dist/telemetry/otel_http_exporter.d.cts +63 -0
  473. package/dist/telemetry/otel_http_exporter.d.ts +63 -0
  474. package/dist/telemetry/otel_http_exporter.d.ts.map +1 -0
  475. package/dist/telemetry/otel_http_exporter.js +142 -0
  476. package/dist/telemetry/otel_http_exporter.js.map +1 -0
  477. package/dist/telemetry/pino_otel_transport.cjs +217 -0
  478. package/dist/telemetry/pino_otel_transport.cjs.map +1 -0
  479. package/dist/telemetry/pino_otel_transport.d.cts +58 -0
  480. package/dist/telemetry/pino_otel_transport.d.ts +58 -0
  481. package/dist/telemetry/pino_otel_transport.d.ts.map +1 -0
  482. package/dist/telemetry/pino_otel_transport.js +189 -0
  483. package/dist/telemetry/pino_otel_transport.js.map +1 -0
  484. package/dist/telemetry/trace_types.cjs +233 -0
  485. package/dist/telemetry/trace_types.cjs.map +1 -0
  486. package/dist/telemetry/trace_types.d.cts +74 -0
  487. package/dist/telemetry/trace_types.d.ts +74 -0
  488. package/dist/telemetry/trace_types.d.ts.map +1 -0
  489. package/dist/telemetry/trace_types.js +141 -0
  490. package/dist/telemetry/trace_types.js.map +1 -0
  491. package/dist/telemetry/traces.cjs +484 -0
  492. package/dist/telemetry/traces.cjs.map +1 -0
  493. package/dist/telemetry/traces.d.cts +116 -0
  494. package/dist/telemetry/traces.d.ts +116 -0
  495. package/dist/telemetry/traces.d.ts.map +1 -0
  496. package/dist/telemetry/traces.js +449 -0
  497. package/dist/telemetry/traces.js.map +1 -0
  498. package/dist/telemetry/utils.cjs +86 -0
  499. package/dist/telemetry/utils.cjs.map +1 -0
  500. package/dist/telemetry/utils.d.cts +5 -0
  501. package/dist/telemetry/utils.d.ts +5 -0
  502. package/dist/telemetry/utils.d.ts.map +1 -0
  503. package/dist/telemetry/utils.js +51 -0
  504. package/dist/telemetry/utils.js.map +1 -0
  505. package/dist/tokenize/basic/sentence.cjs +3 -3
  506. package/dist/tokenize/basic/sentence.cjs.map +1 -1
  507. package/dist/tokenize/basic/sentence.js +3 -3
  508. package/dist/tokenize/basic/sentence.js.map +1 -1
  509. package/dist/tokenize/tokenizer.test.cjs +3 -1
  510. package/dist/tokenize/tokenizer.test.cjs.map +1 -1
  511. package/dist/tokenize/tokenizer.test.js +3 -1
  512. package/dist/tokenize/tokenizer.test.js.map +1 -1
  513. package/dist/transcription.cjs.map +1 -1
  514. package/dist/transcription.d.cts +6 -0
  515. package/dist/transcription.d.ts +6 -0
  516. package/dist/transcription.d.ts.map +1 -1
  517. package/dist/transcription.js.map +1 -1
  518. package/dist/tts/fallback_adapter.cjs +472 -0
  519. package/dist/tts/fallback_adapter.cjs.map +1 -0
  520. package/dist/tts/fallback_adapter.d.cts +110 -0
  521. package/dist/tts/fallback_adapter.d.ts +110 -0
  522. package/dist/tts/fallback_adapter.d.ts.map +1 -0
  523. package/dist/tts/fallback_adapter.js +448 -0
  524. package/dist/tts/fallback_adapter.js.map +1 -0
  525. package/dist/tts/index.cjs +3 -0
  526. package/dist/tts/index.cjs.map +1 -1
  527. package/dist/tts/index.d.cts +1 -0
  528. package/dist/tts/index.d.ts +1 -0
  529. package/dist/tts/index.d.ts.map +1 -1
  530. package/dist/tts/index.js +2 -0
  531. package/dist/tts/index.js.map +1 -1
  532. package/dist/tts/stream_adapter.cjs +25 -8
  533. package/dist/tts/stream_adapter.cjs.map +1 -1
  534. package/dist/tts/stream_adapter.d.cts +6 -3
  535. package/dist/tts/stream_adapter.d.ts +6 -3
  536. package/dist/tts/stream_adapter.d.ts.map +1 -1
  537. package/dist/tts/stream_adapter.js +25 -8
  538. package/dist/tts/stream_adapter.js.map +1 -1
  539. package/dist/tts/tts.cjs +189 -57
  540. package/dist/tts/tts.cjs.map +1 -1
  541. package/dist/tts/tts.d.cts +58 -6
  542. package/dist/tts/tts.d.ts +58 -6
  543. package/dist/tts/tts.d.ts.map +1 -1
  544. package/dist/tts/tts.js +191 -59
  545. package/dist/tts/tts.js.map +1 -1
  546. package/dist/typed_promise.cjs +48 -0
  547. package/dist/typed_promise.cjs.map +1 -0
  548. package/dist/typed_promise.d.cts +24 -0
  549. package/dist/typed_promise.d.ts +24 -0
  550. package/dist/typed_promise.d.ts.map +1 -0
  551. package/dist/typed_promise.js +28 -0
  552. package/dist/typed_promise.js.map +1 -0
  553. package/dist/types.cjs +24 -32
  554. package/dist/types.cjs.map +1 -1
  555. package/dist/types.d.cts +45 -10
  556. package/dist/types.d.ts +45 -10
  557. package/dist/types.d.ts.map +1 -1
  558. package/dist/types.js +20 -30
  559. package/dist/types.js.map +1 -1
  560. package/dist/utils.cjs +124 -28
  561. package/dist/utils.cjs.map +1 -1
  562. package/dist/utils.d.cts +41 -1
  563. package/dist/utils.d.ts +41 -1
  564. package/dist/utils.d.ts.map +1 -1
  565. package/dist/utils.js +119 -27
  566. package/dist/utils.js.map +1 -1
  567. package/dist/utils.test.cjs +73 -1
  568. package/dist/utils.test.cjs.map +1 -1
  569. package/dist/utils.test.js +74 -10
  570. package/dist/utils.test.js.map +1 -1
  571. package/dist/vad.cjs +35 -15
  572. package/dist/vad.cjs.map +1 -1
  573. package/dist/vad.d.cts +15 -5
  574. package/dist/vad.d.ts +15 -5
  575. package/dist/vad.d.ts.map +1 -1
  576. package/dist/vad.js +35 -15
  577. package/dist/vad.js.map +1 -1
  578. package/dist/version.cjs +1 -1
  579. package/dist/version.cjs.map +1 -1
  580. package/dist/version.d.cts +1 -1
  581. package/dist/version.d.ts +1 -1
  582. package/dist/version.d.ts.map +1 -1
  583. package/dist/version.js +1 -1
  584. package/dist/version.js.map +1 -1
  585. package/dist/voice/agent.cjs +258 -35
  586. package/dist/voice/agent.cjs.map +1 -1
  587. package/dist/voice/agent.d.cts +54 -13
  588. package/dist/voice/agent.d.ts +54 -13
  589. package/dist/voice/agent.d.ts.map +1 -1
  590. package/dist/voice/agent.js +254 -34
  591. package/dist/voice/agent.js.map +1 -1
  592. package/dist/voice/agent.test.cjs +314 -0
  593. package/dist/voice/agent.test.cjs.map +1 -1
  594. package/dist/voice/agent.test.js +316 -2
  595. package/dist/voice/agent.test.js.map +1 -1
  596. package/dist/voice/agent_activity.cjs +1116 -385
  597. package/dist/voice/agent_activity.cjs.map +1 -1
  598. package/dist/voice/agent_activity.d.cts +72 -11
  599. package/dist/voice/agent_activity.d.ts +72 -11
  600. package/dist/voice/agent_activity.d.ts.map +1 -1
  601. package/dist/voice/agent_activity.js +1119 -383
  602. package/dist/voice/agent_activity.js.map +1 -1
  603. package/dist/voice/agent_activity.test.cjs +135 -0
  604. package/dist/voice/agent_activity.test.cjs.map +1 -0
  605. package/dist/voice/agent_activity.test.js +134 -0
  606. package/dist/voice/agent_activity.test.js.map +1 -0
  607. package/dist/voice/agent_session.cjs +550 -90
  608. package/dist/voice/agent_session.cjs.map +1 -1
  609. package/dist/voice/agent_session.d.cts +185 -25
  610. package/dist/voice/agent_session.d.ts +185 -25
  611. package/dist/voice/agent_session.d.ts.map +1 -1
  612. package/dist/voice/agent_session.js +556 -91
  613. package/dist/voice/agent_session.js.map +1 -1
  614. package/dist/voice/audio_recognition.cjs +605 -46
  615. package/dist/voice/audio_recognition.cjs.map +1 -1
  616. package/dist/voice/audio_recognition.d.cts +96 -4
  617. package/dist/voice/audio_recognition.d.ts +96 -4
  618. package/dist/voice/audio_recognition.d.ts.map +1 -1
  619. package/dist/voice/audio_recognition.js +611 -47
  620. package/dist/voice/audio_recognition.js.map +1 -1
  621. package/dist/voice/audio_recognition_span.test.cjs +295 -0
  622. package/dist/voice/audio_recognition_span.test.cjs.map +1 -0
  623. package/dist/voice/audio_recognition_span.test.js +299 -0
  624. package/dist/voice/audio_recognition_span.test.js.map +1 -0
  625. package/dist/voice/avatar/datastream_io.cjs +7 -1
  626. package/dist/voice/avatar/datastream_io.cjs.map +1 -1
  627. package/dist/voice/avatar/datastream_io.d.cts +1 -0
  628. package/dist/voice/avatar/datastream_io.d.ts +1 -0
  629. package/dist/voice/avatar/datastream_io.d.ts.map +1 -1
  630. package/dist/voice/avatar/datastream_io.js +7 -1
  631. package/dist/voice/avatar/datastream_io.js.map +1 -1
  632. package/dist/voice/background_audio.cjs +367 -0
  633. package/dist/voice/background_audio.cjs.map +1 -0
  634. package/dist/voice/background_audio.d.cts +123 -0
  635. package/dist/voice/background_audio.d.ts +123 -0
  636. package/dist/voice/background_audio.d.ts.map +1 -0
  637. package/dist/voice/background_audio.js +343 -0
  638. package/dist/voice/background_audio.js.map +1 -0
  639. package/dist/voice/events.cjs +3 -0
  640. package/dist/voice/events.cjs.map +1 -1
  641. package/dist/voice/events.d.cts +16 -9
  642. package/dist/voice/events.d.ts +16 -9
  643. package/dist/voice/events.d.ts.map +1 -1
  644. package/dist/voice/events.js +3 -0
  645. package/dist/voice/events.js.map +1 -1
  646. package/dist/voice/generation.cjs +205 -41
  647. package/dist/voice/generation.cjs.map +1 -1
  648. package/dist/voice/generation.d.cts +21 -5
  649. package/dist/voice/generation.d.ts +21 -5
  650. package/dist/voice/generation.d.ts.map +1 -1
  651. package/dist/voice/generation.js +215 -43
  652. package/dist/voice/generation.js.map +1 -1
  653. package/dist/voice/generation_tools.test.cjs +236 -0
  654. package/dist/voice/generation_tools.test.cjs.map +1 -0
  655. package/dist/voice/generation_tools.test.js +235 -0
  656. package/dist/voice/generation_tools.test.js.map +1 -0
  657. package/dist/voice/index.cjs +33 -2
  658. package/dist/voice/index.cjs.map +1 -1
  659. package/dist/voice/index.d.cts +8 -2
  660. package/dist/voice/index.d.ts +8 -2
  661. package/dist/voice/index.d.ts.map +1 -1
  662. package/dist/voice/index.js +19 -2
  663. package/dist/voice/index.js.map +1 -1
  664. package/dist/voice/interruption_detection.test.cjs +114 -0
  665. package/dist/voice/interruption_detection.test.cjs.map +1 -0
  666. package/dist/voice/interruption_detection.test.js +113 -0
  667. package/dist/voice/interruption_detection.test.js.map +1 -0
  668. package/dist/voice/io.cjs +66 -6
  669. package/dist/voice/io.cjs.map +1 -1
  670. package/dist/voice/io.d.cts +67 -7
  671. package/dist/voice/io.d.ts +67 -7
  672. package/dist/voice/io.d.ts.map +1 -1
  673. package/dist/voice/io.js +62 -5
  674. package/dist/voice/io.js.map +1 -1
  675. package/dist/voice/recorder_io/index.cjs +23 -0
  676. package/dist/voice/recorder_io/index.cjs.map +1 -0
  677. package/dist/voice/recorder_io/index.d.cts +2 -0
  678. package/dist/voice/recorder_io/index.d.ts +2 -0
  679. package/dist/voice/recorder_io/index.d.ts.map +1 -0
  680. package/dist/voice/recorder_io/index.js +2 -0
  681. package/dist/voice/recorder_io/index.js.map +1 -0
  682. package/dist/voice/recorder_io/recorder_io.cjs +607 -0
  683. package/dist/voice/recorder_io/recorder_io.cjs.map +1 -0
  684. package/dist/voice/recorder_io/recorder_io.d.cts +106 -0
  685. package/dist/voice/recorder_io/recorder_io.d.ts +106 -0
  686. package/dist/voice/recorder_io/recorder_io.d.ts.map +1 -0
  687. package/dist/voice/recorder_io/recorder_io.js +573 -0
  688. package/dist/voice/recorder_io/recorder_io.js.map +1 -0
  689. package/dist/voice/remote_session.cjs +922 -0
  690. package/dist/voice/remote_session.cjs.map +1 -0
  691. package/dist/voice/remote_session.d.cts +108 -0
  692. package/dist/voice/remote_session.d.ts +108 -0
  693. package/dist/voice/remote_session.d.ts.map +1 -0
  694. package/dist/voice/remote_session.js +887 -0
  695. package/dist/voice/remote_session.js.map +1 -0
  696. package/dist/voice/report.cjs +88 -0
  697. package/dist/voice/report.cjs.map +1 -0
  698. package/dist/voice/report.d.cts +49 -0
  699. package/dist/voice/report.d.ts +49 -0
  700. package/dist/voice/report.d.ts.map +1 -0
  701. package/dist/voice/report.js +63 -0
  702. package/dist/voice/report.js.map +1 -0
  703. package/dist/voice/report.test.cjs +121 -0
  704. package/dist/voice/report.test.cjs.map +1 -0
  705. package/dist/voice/report.test.js +120 -0
  706. package/dist/voice/report.test.js.map +1 -0
  707. package/dist/voice/room_io/_input.cjs +40 -7
  708. package/dist/voice/room_io/_input.cjs.map +1 -1
  709. package/dist/voice/room_io/_input.d.cts +5 -2
  710. package/dist/voice/room_io/_input.d.ts +5 -2
  711. package/dist/voice/room_io/_input.d.ts.map +1 -1
  712. package/dist/voice/room_io/_input.js +41 -8
  713. package/dist/voice/room_io/_input.js.map +1 -1
  714. package/dist/voice/room_io/_output.cjs +19 -11
  715. package/dist/voice/room_io/_output.cjs.map +1 -1
  716. package/dist/voice/room_io/_output.d.cts +7 -4
  717. package/dist/voice/room_io/_output.d.ts +7 -4
  718. package/dist/voice/room_io/_output.d.ts.map +1 -1
  719. package/dist/voice/room_io/_output.js +20 -12
  720. package/dist/voice/room_io/_output.js.map +1 -1
  721. package/dist/voice/room_io/room_io.cjs +33 -6
  722. package/dist/voice/room_io/room_io.cjs.map +1 -1
  723. package/dist/voice/room_io/room_io.d.cts +29 -9
  724. package/dist/voice/room_io/room_io.d.ts +29 -9
  725. package/dist/voice/room_io/room_io.d.ts.map +1 -1
  726. package/dist/voice/room_io/room_io.js +33 -7
  727. package/dist/voice/room_io/room_io.js.map +1 -1
  728. package/dist/voice/speech_handle.cjs +22 -4
  729. package/dist/voice/speech_handle.cjs.map +1 -1
  730. package/dist/voice/speech_handle.d.cts +17 -2
  731. package/dist/voice/speech_handle.d.ts +17 -2
  732. package/dist/voice/speech_handle.d.ts.map +1 -1
  733. package/dist/voice/speech_handle.js +21 -4
  734. package/dist/voice/speech_handle.js.map +1 -1
  735. package/dist/voice/testing/fake_llm.cjs +127 -0
  736. package/dist/voice/testing/fake_llm.cjs.map +1 -0
  737. package/dist/voice/testing/fake_llm.d.cts +30 -0
  738. package/dist/voice/testing/fake_llm.d.ts +30 -0
  739. package/dist/voice/testing/fake_llm.d.ts.map +1 -0
  740. package/dist/voice/testing/fake_llm.js +103 -0
  741. package/dist/voice/testing/fake_llm.js.map +1 -0
  742. package/dist/voice/testing/index.cjs +57 -0
  743. package/dist/voice/testing/index.cjs.map +1 -0
  744. package/dist/voice/testing/index.d.cts +21 -0
  745. package/dist/voice/testing/index.d.ts +21 -0
  746. package/dist/voice/testing/index.d.ts.map +1 -0
  747. package/dist/voice/testing/index.js +35 -0
  748. package/dist/voice/testing/index.js.map +1 -0
  749. package/dist/voice/testing/run_result.cjs +817 -0
  750. package/dist/voice/testing/run_result.cjs.map +1 -0
  751. package/dist/voice/testing/run_result.d.cts +385 -0
  752. package/dist/voice/testing/run_result.d.ts +385 -0
  753. package/dist/voice/testing/run_result.d.ts.map +1 -0
  754. package/dist/voice/testing/run_result.js +790 -0
  755. package/dist/voice/testing/run_result.js.map +1 -0
  756. package/dist/voice/testing/types.cjs +46 -0
  757. package/dist/voice/testing/types.cjs.map +1 -0
  758. package/dist/voice/testing/types.d.cts +83 -0
  759. package/dist/voice/testing/types.d.ts +83 -0
  760. package/dist/voice/testing/types.d.ts.map +1 -0
  761. package/dist/voice/testing/types.js +19 -0
  762. package/dist/voice/testing/types.js.map +1 -0
  763. package/dist/voice/transcription/synchronizer.cjs +139 -15
  764. package/dist/voice/transcription/synchronizer.cjs.map +1 -1
  765. package/dist/voice/transcription/synchronizer.d.cts +35 -4
  766. package/dist/voice/transcription/synchronizer.d.ts +35 -4
  767. package/dist/voice/transcription/synchronizer.d.ts.map +1 -1
  768. package/dist/voice/transcription/synchronizer.js +143 -16
  769. package/dist/voice/transcription/synchronizer.js.map +1 -1
  770. package/dist/voice/transcription/synchronizer.test.cjs +151 -0
  771. package/dist/voice/transcription/synchronizer.test.cjs.map +1 -0
  772. package/dist/voice/transcription/synchronizer.test.js +150 -0
  773. package/dist/voice/transcription/synchronizer.test.js.map +1 -0
  774. package/dist/voice/turn_config/endpointing.cjs +33 -0
  775. package/dist/voice/turn_config/endpointing.cjs.map +1 -0
  776. package/dist/voice/turn_config/endpointing.d.cts +30 -0
  777. package/dist/voice/turn_config/endpointing.d.ts +30 -0
  778. package/dist/voice/turn_config/endpointing.d.ts.map +1 -0
  779. package/dist/voice/turn_config/endpointing.js +9 -0
  780. package/dist/voice/turn_config/endpointing.js.map +1 -0
  781. package/dist/voice/turn_config/interruption.cjs +37 -0
  782. package/dist/voice/turn_config/interruption.cjs.map +1 -0
  783. package/dist/voice/turn_config/interruption.d.cts +53 -0
  784. package/dist/voice/turn_config/interruption.d.ts +53 -0
  785. package/dist/voice/turn_config/interruption.d.ts.map +1 -0
  786. package/dist/voice/turn_config/interruption.js +13 -0
  787. package/dist/voice/turn_config/interruption.js.map +1 -0
  788. package/dist/voice/turn_config/turn_handling.cjs +35 -0
  789. package/dist/voice/turn_config/turn_handling.cjs.map +1 -0
  790. package/dist/voice/turn_config/turn_handling.d.cts +36 -0
  791. package/dist/voice/turn_config/turn_handling.d.ts +36 -0
  792. package/dist/voice/turn_config/turn_handling.d.ts.map +1 -0
  793. package/dist/voice/turn_config/turn_handling.js +11 -0
  794. package/dist/voice/turn_config/turn_handling.js.map +1 -0
  795. package/dist/voice/turn_config/utils.cjs +157 -0
  796. package/dist/voice/turn_config/utils.cjs.map +1 -0
  797. package/dist/voice/turn_config/utils.d.cts +37 -0
  798. package/dist/voice/turn_config/utils.d.ts +37 -0
  799. package/dist/voice/turn_config/utils.d.ts.map +1 -0
  800. package/dist/voice/turn_config/utils.js +131 -0
  801. package/dist/voice/turn_config/utils.js.map +1 -0
  802. package/dist/voice/turn_config/utils.test.cjs +128 -0
  803. package/dist/voice/turn_config/utils.test.cjs.map +1 -0
  804. package/dist/voice/turn_config/utils.test.js +127 -0
  805. package/dist/voice/turn_config/utils.test.js.map +1 -0
  806. package/dist/voice/utils.cjs +47 -0
  807. package/dist/voice/utils.cjs.map +1 -0
  808. package/dist/voice/utils.d.cts +4 -0
  809. package/dist/voice/utils.d.ts +4 -0
  810. package/dist/voice/utils.d.ts.map +1 -0
  811. package/dist/voice/utils.js +23 -0
  812. package/dist/voice/utils.js.map +1 -0
  813. package/dist/worker.cjs +44 -52
  814. package/dist/worker.cjs.map +1 -1
  815. package/dist/worker.d.cts +18 -8
  816. package/dist/worker.d.ts +18 -8
  817. package/dist/worker.d.ts.map +1 -1
  818. package/dist/worker.js +43 -43
  819. package/dist/worker.js.map +1 -1
  820. package/package.json +35 -13
  821. package/resources/NOTICE +2 -0
  822. package/resources/keyboard-typing.ogg +0 -0
  823. package/resources/keyboard-typing2.ogg +0 -0
  824. package/resources/office-ambience.ogg +0 -0
  825. package/src/_exceptions.ts +5 -0
  826. package/src/audio.ts +132 -1
  827. package/src/beta/index.ts +9 -0
  828. package/src/beta/workflows/index.ts +9 -0
  829. package/src/beta/workflows/task_group.ts +203 -0
  830. package/src/cli.ts +57 -66
  831. package/src/connection_pool.test.ts +346 -0
  832. package/src/connection_pool.ts +307 -0
  833. package/src/constants.ts +14 -0
  834. package/src/cpu.test.ts +239 -0
  835. package/src/cpu.ts +173 -0
  836. package/src/http_server.ts +18 -6
  837. package/src/index.ts +15 -13
  838. package/src/inference/api_protos.ts +85 -2
  839. package/src/inference/index.ts +32 -4
  840. package/src/inference/interruption/defaults.ts +51 -0
  841. package/src/inference/interruption/errors.ts +25 -0
  842. package/src/inference/interruption/http_transport.ts +207 -0
  843. package/src/inference/interruption/interruption_cache_entry.ts +50 -0
  844. package/src/inference/interruption/interruption_detector.ts +204 -0
  845. package/src/inference/interruption/interruption_stream.ts +467 -0
  846. package/src/inference/interruption/types.ts +84 -0
  847. package/src/inference/interruption/utils.test.ts +132 -0
  848. package/src/inference/interruption/utils.ts +137 -0
  849. package/src/inference/interruption/ws_transport.ts +416 -0
  850. package/src/inference/llm.ts +214 -163
  851. package/src/inference/stt.test.ts +253 -0
  852. package/src/inference/stt.ts +449 -208
  853. package/src/inference/tts.test.ts +354 -0
  854. package/src/inference/tts.ts +417 -115
  855. package/src/inference/utils.ts +30 -2
  856. package/src/ipc/inference_proc_executor.ts +11 -3
  857. package/src/ipc/inference_proc_lazy_main.ts +13 -1
  858. package/src/ipc/job_proc_executor.ts +11 -1
  859. package/src/ipc/job_proc_lazy_main.ts +86 -20
  860. package/src/ipc/supervised_proc.test.ts +153 -0
  861. package/src/ipc/supervised_proc.ts +39 -10
  862. package/src/job.ts +120 -1
  863. package/src/language.test.ts +62 -0
  864. package/src/language.ts +380 -0
  865. package/src/llm/__snapshots__/zod-utils.test.ts.snap +559 -0
  866. package/src/llm/chat_context.test.ts +787 -0
  867. package/src/llm/chat_context.ts +493 -2
  868. package/src/llm/fallback_adapter.test.ts +238 -0
  869. package/src/llm/fallback_adapter.ts +394 -0
  870. package/src/llm/index.ts +13 -0
  871. package/src/llm/llm.ts +77 -12
  872. package/src/llm/provider_format/google.test.ts +72 -1
  873. package/src/llm/provider_format/google.ts +10 -6
  874. package/src/llm/provider_format/index.ts +7 -2
  875. package/src/llm/provider_format/openai.test.ts +480 -2
  876. package/src/llm/provider_format/openai.ts +152 -21
  877. package/src/llm/provider_format/utils.ts +11 -5
  878. package/src/llm/realtime.ts +23 -2
  879. package/src/llm/remote_chat_context.ts +2 -2
  880. package/src/llm/tool_context.test.ts +210 -1
  881. package/src/llm/tool_context.ts +115 -17
  882. package/src/llm/utils.test.ts +103 -2
  883. package/src/llm/utils.ts +152 -16
  884. package/src/llm/zod-utils.test.ts +577 -0
  885. package/src/llm/zod-utils.ts +153 -0
  886. package/src/log.ts +71 -19
  887. package/src/metrics/base.ts +78 -19
  888. package/src/metrics/index.ts +12 -0
  889. package/src/metrics/model_usage.test.ts +545 -0
  890. package/src/metrics/model_usage.ts +262 -0
  891. package/src/metrics/usage_collector.ts +14 -3
  892. package/src/metrics/utils.ts +27 -7
  893. package/src/stream/deferred_stream.test.ts +3 -3
  894. package/src/stream/deferred_stream.ts +43 -11
  895. package/src/stream/index.ts +1 -0
  896. package/src/stream/multi_input_stream.test.ts +545 -0
  897. package/src/stream/multi_input_stream.ts +172 -0
  898. package/src/stream/stream_channel.test.ts +37 -0
  899. package/src/stream/stream_channel.ts +43 -3
  900. package/src/stt/stream_adapter.ts +30 -9
  901. package/src/stt/stt.ts +140 -23
  902. package/src/telemetry/index.ts +28 -0
  903. package/src/telemetry/logging.ts +55 -0
  904. package/src/telemetry/otel_http_exporter.ts +218 -0
  905. package/src/telemetry/pino_otel_transport.ts +265 -0
  906. package/src/telemetry/trace_types.ts +109 -0
  907. package/src/telemetry/traces.ts +673 -0
  908. package/src/telemetry/utils.ts +61 -0
  909. package/src/tokenize/basic/sentence.ts +3 -3
  910. package/src/tokenize/tokenizer.test.ts +4 -0
  911. package/src/transcription.ts +6 -0
  912. package/src/tts/fallback_adapter.ts +586 -0
  913. package/src/tts/index.ts +1 -0
  914. package/src/tts/stream_adapter.ts +38 -8
  915. package/src/tts/tts.ts +245 -62
  916. package/src/typed_promise.ts +67 -0
  917. package/src/types.ts +62 -33
  918. package/src/utils.test.ts +90 -10
  919. package/src/utils.ts +178 -33
  920. package/src/vad.ts +42 -18
  921. package/src/version.ts +1 -1
  922. package/src/voice/agent.test.ts +347 -2
  923. package/src/voice/agent.ts +346 -44
  924. package/src/voice/agent_activity.test.ts +194 -0
  925. package/src/voice/agent_activity.ts +1457 -388
  926. package/src/voice/agent_session.ts +817 -112
  927. package/src/voice/audio_recognition.ts +845 -70
  928. package/src/voice/audio_recognition_span.test.ts +341 -0
  929. package/src/voice/avatar/datastream_io.ts +9 -1
  930. package/src/voice/background_audio.ts +494 -0
  931. package/src/voice/events.ts +27 -7
  932. package/src/voice/generation.ts +310 -56
  933. package/src/voice/generation_tools.test.ts +268 -0
  934. package/src/voice/index.ts +17 -3
  935. package/src/voice/interruption_detection.test.ts +151 -0
  936. package/src/voice/io.ts +115 -12
  937. package/src/voice/recorder_io/index.ts +4 -0
  938. package/src/voice/recorder_io/recorder_io.ts +783 -0
  939. package/src/voice/remote_session.ts +1083 -0
  940. package/src/voice/report.test.ts +136 -0
  941. package/src/voice/report.ts +140 -0
  942. package/src/voice/room_io/_input.ts +45 -10
  943. package/src/voice/room_io/_output.ts +26 -14
  944. package/src/voice/room_io/room_io.ts +67 -22
  945. package/src/voice/speech_handle.ts +38 -6
  946. package/src/voice/testing/fake_llm.ts +138 -0
  947. package/src/voice/testing/index.ts +52 -0
  948. package/src/voice/testing/run_result.ts +995 -0
  949. package/src/voice/testing/types.ts +118 -0
  950. package/src/voice/transcription/synchronizer.test.ts +206 -0
  951. package/src/voice/transcription/synchronizer.ts +204 -19
  952. package/src/voice/turn_config/endpointing.ts +33 -0
  953. package/src/voice/turn_config/interruption.ts +56 -0
  954. package/src/voice/turn_config/turn_handling.ts +45 -0
  955. package/src/voice/turn_config/utils.test.ts +148 -0
  956. package/src/voice/turn_config/utils.ts +167 -0
  957. package/src/voice/utils.ts +29 -0
  958. package/src/worker.ts +92 -78
  959. package/src/llm/__snapshots__/utils.test.ts.snap +0 -65
@@ -1,8 +1,9 @@
1
1
  import {} from "@livekit/rtc-node";
2
- import { WebSocket } from "ws";
3
2
  import { APIError, APIStatusError } from "../_exceptions.js";
4
3
  import { AudioByteStream } from "../audio.js";
4
+ import { areLanguagesEquivalent, normalizeLanguage } from "../language.js";
5
5
  import { log } from "../log.js";
6
+ import { createStreamChannel } from "../stream/stream_channel.js";
6
7
  import {
7
8
  STT as BaseSTT,
8
9
  SpeechStream as BaseSpeechStream,
@@ -10,16 +11,40 @@ import {
10
11
  } from "../stt/index.js";
11
12
  import { DEFAULT_API_CONNECT_OPTIONS } from "../types.js";
12
13
  import { Event, Task, cancelAndWait, shortuuid, waitForAbort } from "../utils.js";
13
- import { connectWs, createAccessToken } from "./utils.js";
14
+ import { createTimedString } from "../voice/io.js";
15
+ import {
16
+ sttServerEventSchema
17
+ } from "./api_protos.js";
18
+ import { connectWs, createAccessToken, getDefaultInferenceUrl } from "./utils.js";
19
+ function parseSTTModelString(model) {
20
+ const idx = model.lastIndexOf(":");
21
+ if (idx !== -1) {
22
+ return [model.slice(0, idx), normalizeLanguage(model.slice(idx + 1))];
23
+ }
24
+ return [model, void 0];
25
+ }
26
+ function normalizeSTTFallback(fallback) {
27
+ const makeFallback = (model) => {
28
+ if (typeof model === "string") {
29
+ const [name] = parseSTTModelString(model);
30
+ return { model: name };
31
+ }
32
+ return model;
33
+ };
34
+ if (Array.isArray(fallback)) {
35
+ return fallback.map(makeFallback);
36
+ }
37
+ return [makeFallback(fallback)];
38
+ }
14
39
  const DEFAULT_ENCODING = "pcm_s16le";
15
40
  const DEFAULT_SAMPLE_RATE = 16e3;
16
- const DEFAULT_BASE_URL = "wss://agent-gateway.livekit.cloud/v1";
17
41
  const DEFAULT_CANCEL_TIMEOUT = 5e3;
18
42
  class STT extends BaseSTT {
19
43
  opts;
20
44
  streams = /* @__PURE__ */ new Set();
45
+ #logger = log();
21
46
  constructor(opts) {
22
- super({ streaming: true, interimResults: true });
47
+ super({ streaming: true, interimResults: true, alignedTranscript: "word" });
23
48
  const {
24
49
  model,
25
50
  language,
@@ -28,9 +53,11 @@ class STT extends BaseSTT {
28
53
  sampleRate = DEFAULT_SAMPLE_RATE,
29
54
  apiKey,
30
55
  apiSecret,
31
- extraKwargs = {}
56
+ modelOptions = {},
57
+ fallback,
58
+ connOptions
32
59
  } = opts || {};
33
- const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;
60
+ const lkBaseURL = baseURL || getDefaultInferenceUrl();
34
61
  const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;
35
62
  if (!lkApiKey) {
36
63
  throw new Error("apiKey is required: pass apiKey or set LIVEKIT_API_KEY");
@@ -39,39 +66,113 @@ class STT extends BaseSTT {
39
66
  if (!lkApiSecret) {
40
67
  throw new Error("apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET");
41
68
  }
69
+ let nextModel = model;
70
+ let nextLanguage = language;
71
+ if (typeof nextModel === "string") {
72
+ const [parsedModel, parsedLanguage] = parseSTTModelString(nextModel);
73
+ if (parsedLanguage !== void 0) {
74
+ if (nextLanguage && !areLanguagesEquivalent(nextLanguage, parsedLanguage)) {
75
+ this.#logger.warn(
76
+ "`language` is provided via both argument and model, using the one from the argument",
77
+ { language: nextLanguage, model: nextModel }
78
+ );
79
+ } else {
80
+ nextLanguage = parsedLanguage;
81
+ }
82
+ nextModel = parsedModel;
83
+ }
84
+ }
85
+ const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : void 0;
42
86
  this.opts = {
43
- model,
44
- language,
87
+ model: nextModel,
88
+ language: nextLanguage ? normalizeLanguage(nextLanguage) : void 0,
45
89
  encoding,
46
90
  sampleRate,
47
91
  baseURL: lkBaseURL,
48
92
  apiKey: lkApiKey,
49
93
  apiSecret: lkApiSecret,
50
- extraKwargs
94
+ modelOptions,
95
+ fallback: normalizedFallback,
96
+ connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS
51
97
  };
52
98
  }
53
99
  get label() {
54
100
  return "inference.STT";
55
101
  }
102
+ get model() {
103
+ return this.opts.model ?? "auto";
104
+ }
105
+ get provider() {
106
+ return "livekit";
107
+ }
108
+ static fromModelString(modelString) {
109
+ const [model, language] = parseSTTModelString(modelString);
110
+ return new STT({ model, language });
111
+ }
56
112
  async _recognize(_) {
57
113
  throw new Error("LiveKit STT does not support batch recognition, use stream() instead");
58
114
  }
59
115
  updateOptions(opts) {
60
- this.opts = { ...this.opts, ...opts };
116
+ this.opts = {
117
+ ...this.opts,
118
+ ...opts,
119
+ language: opts.language !== void 0 ? normalizeLanguage(opts.language) : this.opts.language
120
+ };
61
121
  for (const stream of this.streams) {
62
122
  stream.updateOptions(opts);
63
123
  }
64
124
  }
65
125
  stream(options) {
66
- const { language, connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};
126
+ const { language, connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } = options || {};
67
127
  const streamOpts = {
68
128
  ...this.opts,
69
- language: language ?? this.opts.language
129
+ language: language !== void 0 ? normalizeLanguage(language) : this.opts.language
70
130
  };
71
131
  const stream = new SpeechStream(this, streamOpts, connOptions);
72
132
  this.streams.add(stream);
73
133
  return stream;
74
134
  }
135
+ async connectWs(timeout) {
136
+ var _a;
137
+ const params = {
138
+ settings: {
139
+ sample_rate: String(this.opts.sampleRate),
140
+ encoding: this.opts.encoding,
141
+ extra: this.opts.modelOptions
142
+ }
143
+ };
144
+ if (this.opts.model && this.opts.model !== "auto") {
145
+ params.model = this.opts.model;
146
+ }
147
+ if (this.opts.language) {
148
+ params.settings.language = this.opts.language;
149
+ }
150
+ if ((_a = this.opts.fallback) == null ? void 0 : _a.length) {
151
+ params.fallback = {
152
+ models: this.opts.fallback.map((m) => ({
153
+ model: m.model,
154
+ extra: m.extraKwargs ?? {}
155
+ }))
156
+ };
157
+ }
158
+ if (this.opts.connOptions) {
159
+ params.connection = {
160
+ timeout: this.opts.connOptions.timeoutMs / 1e3,
161
+ retries: this.opts.connOptions.maxRetry
162
+ };
163
+ }
164
+ let baseURL = this.opts.baseURL;
165
+ if (baseURL.startsWith("http://") || baseURL.startsWith("https://")) {
166
+ baseURL = baseURL.replace("http", "ws");
167
+ }
168
+ const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);
169
+ const url = `${baseURL}/stt`;
170
+ const headers = { Authorization: `Bearer ${token}` };
171
+ const socket = await connectWs(url, headers, timeout);
172
+ const msg = { ...params, type: "session.create" };
173
+ socket.send(JSON.stringify(msg));
174
+ return socket;
175
+ }
75
176
  }
76
177
  class SpeechStream extends BaseSpeechStream {
77
178
  opts;
@@ -79,206 +180,254 @@ class SpeechStream extends BaseSpeechStream {
79
180
  speaking = false;
80
181
  speechDuration = 0;
81
182
  reconnectEvent = new Event();
183
+ stt;
184
+ connOptions;
82
185
  #logger = log();
83
186
  constructor(sttImpl, opts, connOptions) {
84
187
  super(sttImpl, opts.sampleRate, connOptions);
85
188
  this.opts = opts;
189
+ this.stt = sttImpl;
190
+ this.connOptions = connOptions;
86
191
  }
87
192
  get label() {
88
193
  return "inference.SpeechStream";
89
194
  }
90
195
  updateOptions(opts) {
91
- this.opts = { ...this.opts, ...opts };
196
+ this.opts = {
197
+ ...this.opts,
198
+ ...opts,
199
+ language: opts.language !== void 0 ? normalizeLanguage(opts.language) : this.opts.language
200
+ };
201
+ this.reconnectEvent.set();
92
202
  }
93
203
  async run() {
94
- let ws = null;
95
- let closingWs = false;
96
- this.reconnectEvent.set();
97
- const connect = async () => {
98
- const params = {
99
- settings: {
100
- sample_rate: String(this.opts.sampleRate),
101
- encoding: this.opts.encoding,
102
- extra: this.opts.extraKwargs
103
- }
204
+ while (true) {
205
+ let ws = null;
206
+ let closing = false;
207
+ let finalReceived = false;
208
+ const eventChannel = createStreamChannel();
209
+ const resourceCleanup = () => {
210
+ if (closing) return;
211
+ closing = true;
212
+ eventChannel.close();
213
+ ws == null ? void 0 : ws.removeAllListeners();
214
+ ws == null ? void 0 : ws.close();
104
215
  };
105
- if (this.opts.model) {
106
- params.model = this.opts.model;
107
- }
108
- if (this.opts.language) {
109
- params.settings.language = this.opts.language;
110
- }
111
- let baseURL = this.opts.baseURL;
112
- if (baseURL.startsWith("http://") || baseURL.startsWith("https://")) {
113
- baseURL = baseURL.replace("http", "ws");
114
- }
115
- const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);
116
- const url = `${baseURL}/stt`;
117
- const headers = { Authorization: `Bearer ${token}` };
118
- const socket = await connectWs(url, headers, 1e4);
119
- const msg = { ...params, type: "session.create" };
120
- socket.send(JSON.stringify(msg));
121
- return socket;
122
- };
123
- const send = async (socket, signal) => {
124
- const audioStream = new AudioByteStream(
125
- this.opts.sampleRate,
126
- 1,
127
- Math.floor(this.opts.sampleRate / 20)
128
- // 50ms
129
- );
130
- for await (const ev of this.input) {
131
- if (signal.aborted) break;
132
- let frames;
133
- if (ev === SpeechStream.FLUSH_SENTINEL) {
134
- frames = audioStream.flush();
135
- } else {
136
- const frame = ev;
137
- frames = audioStream.write(new Int16Array(frame.data).buffer);
138
- }
139
- for (const frame of frames) {
140
- this.speechDuration += frame.samplesPerChannel / frame.sampleRate;
141
- const base64 = Buffer.from(frame.data.buffer).toString("base64");
142
- const msg = { type: "input_audio", audio: base64 };
143
- socket.send(JSON.stringify(msg));
144
- }
145
- }
146
- closingWs = true;
147
- socket.send(JSON.stringify({ type: "session.finalize" }));
148
- };
149
- const recv = async (socket, signal) => {
150
- while (!this.closed && !signal.aborted) {
151
- const dataPromise = new Promise((resolve, reject) => {
152
- const messageHandler = (d) => {
153
- resolve(d.toString());
154
- removeListeners();
216
+ const createWsListener = async (ws2, signal) => {
217
+ return new Promise((resolve, reject) => {
218
+ const onAbort = () => {
219
+ resourceCleanup();
220
+ reject(new Error("WebSocket connection aborted"));
155
221
  };
156
- const errorHandler = (e) => {
222
+ signal.addEventListener("abort", onAbort, { once: true });
223
+ ws2.on("message", (data) => {
224
+ const json = JSON.parse(data.toString());
225
+ eventChannel.write(json);
226
+ });
227
+ ws2.on("error", (e) => {
228
+ this.#logger.error({ error: e }, "WebSocket error");
229
+ resourceCleanup();
157
230
  reject(e);
158
- removeListeners();
159
- };
160
- const closeHandler = (code) => {
161
- if (closingWs) {
162
- resolve("");
231
+ });
232
+ ws2.on("close", (code) => {
233
+ resourceCleanup();
234
+ if (!closing) return this.#logger.error("WebSocket closed unexpectedly");
235
+ if (finalReceived) return resolve();
236
+ reject(
237
+ new APIStatusError({
238
+ message: "LiveKit STT connection closed unexpectedly",
239
+ options: { statusCode: code }
240
+ })
241
+ );
242
+ });
243
+ });
244
+ };
245
+ const send = async (socket, signal) => {
246
+ const audioStream = new AudioByteStream(
247
+ this.opts.sampleRate,
248
+ 1,
249
+ Math.floor(this.opts.sampleRate / 20)
250
+ // 50ms
251
+ );
252
+ const abortPromise = new Promise((_, reject) => {
253
+ if (signal.aborted) {
254
+ return reject(new Error("Send aborted"));
255
+ }
256
+ const onAbort = () => reject(new Error("Send aborted"));
257
+ signal.addEventListener("abort", onAbort, { once: true });
258
+ });
259
+ const iterator = this.input[Symbol.asyncIterator]();
260
+ try {
261
+ while (true) {
262
+ const result = await Promise.race([iterator.next(), abortPromise]);
263
+ if (result.done) break;
264
+ const ev = result.value;
265
+ let frames;
266
+ if (ev === SpeechStream.FLUSH_SENTINEL) {
267
+ frames = audioStream.flush();
163
268
  } else {
164
- reject(
165
- new APIStatusError({
166
- message: "LiveKit STT connection closed unexpectedly",
167
- options: { statusCode: code }
168
- })
269
+ const frame = ev;
270
+ frames = audioStream.write(new Int16Array(frame.data).buffer);
271
+ }
272
+ for (const frame of frames) {
273
+ this.speechDuration += frame.samplesPerChannel / frame.sampleRate;
274
+ const base64 = Buffer.from(frame.data.buffer).toString("base64");
275
+ const msg = { type: "input_audio", audio: base64 };
276
+ socket.send(JSON.stringify(msg));
277
+ }
278
+ }
279
+ closing = true;
280
+ socket.send(JSON.stringify({ type: "session.finalize" }));
281
+ } catch (e) {
282
+ if (e.message === "Send aborted") {
283
+ return;
284
+ }
285
+ throw e;
286
+ }
287
+ };
288
+ const recv = async (signal) => {
289
+ const serverEventStream = eventChannel.stream();
290
+ const reader = serverEventStream.getReader();
291
+ try {
292
+ while (!this.closed && !signal.aborted) {
293
+ const result = await reader.read();
294
+ if (signal.aborted) return;
295
+ if (result.done) return;
296
+ const parseResult = await sttServerEventSchema.safeParseAsync(result.value);
297
+ if (!parseResult.success) {
298
+ this.#logger.warn(
299
+ { error: parseResult.error, rawData: result.value },
300
+ "Failed to parse STT server event"
169
301
  );
302
+ continue;
170
303
  }
171
- removeListeners();
172
- };
173
- const removeListeners = () => {
174
- socket.removeListener("message", messageHandler);
175
- socket.removeListener("error", errorHandler);
176
- socket.removeListener("close", closeHandler);
177
- };
178
- socket.once("message", messageHandler);
179
- socket.once("error", errorHandler);
180
- socket.once("close", closeHandler);
181
- });
182
- const data = await Promise.race([dataPromise, waitForAbort(signal)]);
183
- if (!data || signal.aborted) return;
184
- const json = JSON.parse(data);
185
- const type = json.type;
186
- switch (type) {
187
- case "session.created":
188
- case "session.finalized":
189
- case "session.closed":
190
- break;
191
- case "interim_transcript":
192
- this.processTranscript(json, false);
193
- break;
194
- case "final_transcript":
195
- this.processTranscript(json, true);
196
- break;
197
- case "error":
198
- this.#logger.error("received error from LiveKit STT: %o", json);
199
- throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);
200
- default:
201
- this.#logger.warn("received unexpected message from LiveKit STT: %o", json);
202
- break;
304
+ const event = parseResult.data;
305
+ switch (event.type) {
306
+ case "session.created":
307
+ case "session.finalized":
308
+ break;
309
+ case "session.closed":
310
+ finalReceived = true;
311
+ resourceCleanup();
312
+ break;
313
+ case "interim_transcript":
314
+ this.processTranscript(event, false);
315
+ break;
316
+ case "final_transcript":
317
+ this.processTranscript(event, true);
318
+ break;
319
+ case "error":
320
+ this.#logger.error({ error: event }, "Received error from LiveKit STT");
321
+ resourceCleanup();
322
+ throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);
323
+ }
324
+ }
325
+ } finally {
326
+ reader.releaseLock();
327
+ try {
328
+ await serverEventStream.cancel();
329
+ } catch (e) {
330
+ this.#logger.debug("Error cancelling serverEventStream (may already be cancelled):", e);
331
+ }
203
332
  }
204
- }
205
- };
206
- while (true) {
333
+ };
207
334
  try {
208
- ws = await connect();
209
- const sendTask = Task.from(async ({ signal }) => {
210
- await send(ws, signal);
211
- });
212
- const recvTask = Task.from(async ({ signal }) => {
213
- await recv(ws, signal);
214
- });
215
- const tasks = [sendTask, recvTask];
216
- const waitReconnectTask = Task.from(async ({ signal }) => {
217
- await Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]);
218
- });
335
+ ws = await this.stt.connectWs(this.connOptions.timeoutMs);
336
+ const controller = this.abortController;
337
+ const sendTask = Task.from(({ signal }) => send(ws, signal), controller);
338
+ const wsListenerTask = Task.from(({ signal }) => createWsListener(ws, signal), controller);
339
+ const recvTask = Task.from(({ signal }) => recv(signal), controller);
340
+ const waitReconnectTask = Task.from(
341
+ ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),
342
+ controller
343
+ );
219
344
  try {
220
345
  await Promise.race([
221
- Promise.all(tasks.map((task) => task.result)),
346
+ Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),
222
347
  waitReconnectTask.result
223
348
  ]);
224
349
  if (!waitReconnectTask.done) break;
225
350
  this.reconnectEvent.clear();
226
351
  } finally {
227
- await cancelAndWait([sendTask, recvTask, waitReconnectTask], DEFAULT_CANCEL_TIMEOUT);
352
+ await cancelAndWait(
353
+ [sendTask, wsListenerTask, recvTask, waitReconnectTask],
354
+ DEFAULT_CANCEL_TIMEOUT
355
+ );
356
+ resourceCleanup();
228
357
  }
229
358
  } finally {
230
- try {
231
- if (ws) ws.close();
232
- } catch {
233
- }
359
+ resourceCleanup();
234
360
  }
235
361
  }
236
362
  }
237
363
  processTranscript(data, isFinal) {
238
- const requestId = data.request_id ?? this.requestId;
239
- const text = data.transcript ?? "";
240
- const language = data.language ?? this.opts.language ?? "en";
364
+ if (this.queue.closed) return;
365
+ const requestId = data.session_id || this.requestId;
366
+ const text = data.transcript;
367
+ const language = normalizeLanguage(data.language || this.opts.language || "en");
241
368
  if (!text && !isFinal) return;
242
- if (!this.speaking) {
243
- this.speaking = true;
244
- this.queue.put({ type: SpeechEventType.START_OF_SPEECH });
245
- }
246
- const speechData = {
247
- language,
248
- startTime: data.start ?? 0,
249
- endTime: data.duration ?? 0,
250
- confidence: data.confidence ?? 1,
251
- text
252
- };
253
- if (isFinal) {
254
- if (this.speechDuration > 0) {
369
+ try {
370
+ if (!this.speaking) {
371
+ this.speaking = true;
372
+ this.queue.put({ type: SpeechEventType.START_OF_SPEECH });
373
+ }
374
+ const speechData = {
375
+ language,
376
+ startTime: this.startTimeOffset + data.start,
377
+ endTime: this.startTimeOffset + data.start + data.duration,
378
+ confidence: data.confidence,
379
+ text,
380
+ words: data.words.map(
381
+ (word) => createTimedString({
382
+ text: word.word,
383
+ startTime: word.start + this.startTimeOffset,
384
+ endTime: word.end + this.startTimeOffset,
385
+ startTimeOffset: this.startTimeOffset,
386
+ confidence: word.confidence
387
+ })
388
+ )
389
+ };
390
+ if (isFinal) {
391
+ if (this.speechDuration > 0) {
392
+ this.queue.put({
393
+ type: SpeechEventType.RECOGNITION_USAGE,
394
+ requestId,
395
+ recognitionUsage: { audioDuration: this.speechDuration }
396
+ });
397
+ this.speechDuration = 0;
398
+ }
399
+ this.queue.put({
400
+ type: SpeechEventType.FINAL_TRANSCRIPT,
401
+ requestId,
402
+ alternatives: [speechData]
403
+ });
404
+ if (this.speaking) {
405
+ this.speaking = false;
406
+ this.queue.put({ type: SpeechEventType.END_OF_SPEECH });
407
+ }
408
+ } else {
255
409
  this.queue.put({
256
- type: SpeechEventType.RECOGNITION_USAGE,
410
+ type: SpeechEventType.INTERIM_TRANSCRIPT,
257
411
  requestId,
258
- recognitionUsage: { audioDuration: this.speechDuration }
412
+ alternatives: [speechData]
259
413
  });
260
- this.speechDuration = 0;
261
414
  }
262
- this.queue.put({
263
- type: SpeechEventType.FINAL_TRANSCRIPT,
264
- requestId,
265
- alternatives: [speechData]
266
- });
267
- if (this.speaking) {
268
- this.speaking = false;
269
- this.queue.put({ type: SpeechEventType.END_OF_SPEECH });
415
+ } catch (e) {
416
+ if (e instanceof Error && e.message.includes("Queue is closed")) {
417
+ this.#logger.warn(
418
+ { err: e },
419
+ "Queue closed during transcript processing (expected during disconnect)"
420
+ );
421
+ } else {
422
+ this.#logger.error({ err: e }, "Error putting transcript to queue");
270
423
  }
271
- } else {
272
- this.queue.put({
273
- type: SpeechEventType.INTERIM_TRANSCRIPT,
274
- requestId,
275
- alternatives: [speechData]
276
- });
277
424
  }
278
425
  }
279
426
  }
280
427
  export {
281
428
  STT,
282
- SpeechStream
429
+ SpeechStream,
430
+ normalizeSTTFallback,
431
+ parseSTTModelString
283
432
  };
284
433
  //# sourceMappingURL=stt.js.map