@livekit/agents 1.0.47 → 1.1.0-dev.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 (444) hide show
  1. package/dist/beta/index.cjs +29 -0
  2. package/dist/beta/index.cjs.map +1 -0
  3. package/dist/beta/index.d.cts +2 -0
  4. package/dist/beta/index.d.ts +2 -0
  5. package/dist/beta/index.d.ts.map +1 -0
  6. package/dist/beta/index.js +7 -0
  7. package/dist/beta/index.js.map +1 -0
  8. package/dist/beta/workflows/index.cjs +29 -0
  9. package/dist/beta/workflows/index.cjs.map +1 -0
  10. package/dist/beta/workflows/index.d.cts +2 -0
  11. package/dist/beta/workflows/index.d.ts +2 -0
  12. package/dist/beta/workflows/index.d.ts.map +1 -0
  13. package/dist/beta/workflows/index.js +7 -0
  14. package/dist/beta/workflows/index.js.map +1 -0
  15. package/dist/beta/workflows/task_group.cjs +162 -0
  16. package/dist/beta/workflows/task_group.cjs.map +1 -0
  17. package/dist/beta/workflows/task_group.d.cts +32 -0
  18. package/dist/beta/workflows/task_group.d.ts +32 -0
  19. package/dist/beta/workflows/task_group.d.ts.map +1 -0
  20. package/dist/beta/workflows/task_group.js +138 -0
  21. package/dist/beta/workflows/task_group.js.map +1 -0
  22. package/dist/constants.cjs +27 -0
  23. package/dist/constants.cjs.map +1 -1
  24. package/dist/constants.d.cts +9 -0
  25. package/dist/constants.d.ts +9 -0
  26. package/dist/constants.d.ts.map +1 -1
  27. package/dist/constants.js +18 -0
  28. package/dist/constants.js.map +1 -1
  29. package/dist/index.cjs +3 -0
  30. package/dist/index.cjs.map +1 -1
  31. package/dist/index.d.cts +2 -1
  32. package/dist/index.d.ts +2 -1
  33. package/dist/index.d.ts.map +1 -1
  34. package/dist/index.js +2 -0
  35. package/dist/index.js.map +1 -1
  36. package/dist/inference/api_protos.d.cts +12 -12
  37. package/dist/inference/api_protos.d.ts +12 -12
  38. package/dist/inference/interruption/defaults.cjs +81 -0
  39. package/dist/inference/interruption/defaults.cjs.map +1 -0
  40. package/dist/inference/interruption/defaults.d.cts +19 -0
  41. package/dist/inference/interruption/defaults.d.ts +19 -0
  42. package/dist/inference/interruption/defaults.d.ts.map +1 -0
  43. package/dist/inference/interruption/defaults.js +46 -0
  44. package/dist/inference/interruption/defaults.js.map +1 -0
  45. package/dist/inference/interruption/errors.cjs +44 -0
  46. package/dist/inference/interruption/errors.cjs.map +1 -0
  47. package/dist/inference/interruption/errors.d.cts +12 -0
  48. package/dist/inference/interruption/errors.d.ts +12 -0
  49. package/dist/inference/interruption/errors.d.ts.map +1 -0
  50. package/dist/inference/interruption/errors.js +20 -0
  51. package/dist/inference/interruption/errors.js.map +1 -0
  52. package/dist/inference/interruption/http_transport.cjs +147 -0
  53. package/dist/inference/interruption/http_transport.cjs.map +1 -0
  54. package/dist/inference/interruption/http_transport.d.cts +63 -0
  55. package/dist/inference/interruption/http_transport.d.ts +63 -0
  56. package/dist/inference/interruption/http_transport.d.ts.map +1 -0
  57. package/dist/inference/interruption/http_transport.js +121 -0
  58. package/dist/inference/interruption/http_transport.js.map +1 -0
  59. package/dist/inference/interruption/interruption_cache_entry.cjs +58 -0
  60. package/dist/inference/interruption/interruption_cache_entry.cjs.map +1 -0
  61. package/dist/inference/interruption/interruption_cache_entry.d.cts +30 -0
  62. package/dist/inference/interruption/interruption_cache_entry.d.ts +30 -0
  63. package/dist/inference/interruption/interruption_cache_entry.d.ts.map +1 -0
  64. package/dist/inference/interruption/interruption_cache_entry.js +34 -0
  65. package/dist/inference/interruption/interruption_cache_entry.js.map +1 -0
  66. package/dist/inference/interruption/interruption_detector.cjs +181 -0
  67. package/dist/inference/interruption/interruption_detector.cjs.map +1 -0
  68. package/dist/inference/interruption/interruption_detector.d.cts +59 -0
  69. package/dist/inference/interruption/interruption_detector.d.ts +59 -0
  70. package/dist/inference/interruption/interruption_detector.d.ts.map +1 -0
  71. package/dist/inference/interruption/interruption_detector.js +147 -0
  72. package/dist/inference/interruption/interruption_detector.js.map +1 -0
  73. package/dist/inference/interruption/interruption_stream.cjs +368 -0
  74. package/dist/inference/interruption/interruption_stream.cjs.map +1 -0
  75. package/dist/inference/interruption/interruption_stream.d.cts +46 -0
  76. package/dist/inference/interruption/interruption_stream.d.ts +46 -0
  77. package/dist/inference/interruption/interruption_stream.d.ts.map +1 -0
  78. package/dist/inference/interruption/interruption_stream.js +344 -0
  79. package/dist/inference/interruption/interruption_stream.js.map +1 -0
  80. package/dist/inference/interruption/types.cjs +17 -0
  81. package/dist/inference/interruption/types.cjs.map +1 -0
  82. package/dist/inference/interruption/types.d.cts +66 -0
  83. package/dist/inference/interruption/types.d.ts +66 -0
  84. package/dist/inference/interruption/types.d.ts.map +1 -0
  85. package/dist/inference/interruption/types.js +1 -0
  86. package/dist/inference/interruption/types.js.map +1 -0
  87. package/dist/inference/interruption/utils.cjs +130 -0
  88. package/dist/inference/interruption/utils.cjs.map +1 -0
  89. package/dist/inference/interruption/utils.d.cts +41 -0
  90. package/dist/inference/interruption/utils.d.ts +41 -0
  91. package/dist/inference/interruption/utils.d.ts.map +1 -0
  92. package/dist/inference/interruption/utils.js +105 -0
  93. package/dist/inference/interruption/utils.js.map +1 -0
  94. package/dist/inference/interruption/utils.test.cjs +105 -0
  95. package/dist/inference/interruption/utils.test.cjs.map +1 -0
  96. package/dist/inference/interruption/utils.test.js +104 -0
  97. package/dist/inference/interruption/utils.test.js.map +1 -0
  98. package/dist/inference/interruption/ws_transport.cjs +329 -0
  99. package/dist/inference/interruption/ws_transport.cjs.map +1 -0
  100. package/dist/inference/interruption/ws_transport.d.cts +33 -0
  101. package/dist/inference/interruption/ws_transport.d.ts +33 -0
  102. package/dist/inference/interruption/ws_transport.d.ts.map +1 -0
  103. package/dist/inference/interruption/ws_transport.js +295 -0
  104. package/dist/inference/interruption/ws_transport.js.map +1 -0
  105. package/dist/inference/llm.cjs +14 -10
  106. package/dist/inference/llm.cjs.map +1 -1
  107. package/dist/inference/llm.d.cts +2 -1
  108. package/dist/inference/llm.d.ts +2 -1
  109. package/dist/inference/llm.d.ts.map +1 -1
  110. package/dist/inference/llm.js +8 -10
  111. package/dist/inference/llm.js.map +1 -1
  112. package/dist/inference/stt.cjs +7 -2
  113. package/dist/inference/stt.cjs.map +1 -1
  114. package/dist/inference/stt.d.cts +2 -0
  115. package/dist/inference/stt.d.ts +2 -0
  116. package/dist/inference/stt.d.ts.map +1 -1
  117. package/dist/inference/stt.js +8 -3
  118. package/dist/inference/stt.js.map +1 -1
  119. package/dist/inference/tts.cjs +7 -2
  120. package/dist/inference/tts.cjs.map +1 -1
  121. package/dist/inference/tts.d.cts +2 -0
  122. package/dist/inference/tts.d.ts +2 -0
  123. package/dist/inference/tts.d.ts.map +1 -1
  124. package/dist/inference/tts.js +8 -3
  125. package/dist/inference/tts.js.map +1 -1
  126. package/dist/inference/utils.cjs +26 -7
  127. package/dist/inference/utils.cjs.map +1 -1
  128. package/dist/inference/utils.d.cts +13 -0
  129. package/dist/inference/utils.d.ts +13 -0
  130. package/dist/inference/utils.d.ts.map +1 -1
  131. package/dist/inference/utils.js +18 -2
  132. package/dist/inference/utils.js.map +1 -1
  133. package/dist/llm/chat_context.cjs +108 -2
  134. package/dist/llm/chat_context.cjs.map +1 -1
  135. package/dist/llm/chat_context.d.cts +28 -1
  136. package/dist/llm/chat_context.d.ts +28 -1
  137. package/dist/llm/chat_context.d.ts.map +1 -1
  138. package/dist/llm/chat_context.js +108 -2
  139. package/dist/llm/chat_context.js.map +1 -1
  140. package/dist/llm/chat_context.test.cjs +43 -0
  141. package/dist/llm/chat_context.test.cjs.map +1 -1
  142. package/dist/llm/chat_context.test.js +43 -0
  143. package/dist/llm/chat_context.test.js.map +1 -1
  144. package/dist/llm/index.cjs +2 -0
  145. package/dist/llm/index.cjs.map +1 -1
  146. package/dist/llm/index.d.cts +2 -2
  147. package/dist/llm/index.d.ts +2 -2
  148. package/dist/llm/index.d.ts.map +1 -1
  149. package/dist/llm/index.js +3 -1
  150. package/dist/llm/index.js.map +1 -1
  151. package/dist/llm/llm.cjs +16 -1
  152. package/dist/llm/llm.cjs.map +1 -1
  153. package/dist/llm/llm.d.cts +9 -0
  154. package/dist/llm/llm.d.ts +9 -0
  155. package/dist/llm/llm.d.ts.map +1 -1
  156. package/dist/llm/llm.js +16 -1
  157. package/dist/llm/llm.js.map +1 -1
  158. package/dist/llm/provider_format/index.d.cts +1 -1
  159. package/dist/llm/provider_format/index.d.ts +1 -1
  160. package/dist/llm/realtime.cjs +3 -0
  161. package/dist/llm/realtime.cjs.map +1 -1
  162. package/dist/llm/realtime.d.cts +1 -0
  163. package/dist/llm/realtime.d.ts +1 -0
  164. package/dist/llm/realtime.d.ts.map +1 -1
  165. package/dist/llm/realtime.js +3 -0
  166. package/dist/llm/realtime.js.map +1 -1
  167. package/dist/llm/tool_context.cjs +7 -0
  168. package/dist/llm/tool_context.cjs.map +1 -1
  169. package/dist/llm/tool_context.d.cts +10 -2
  170. package/dist/llm/tool_context.d.ts +10 -2
  171. package/dist/llm/tool_context.d.ts.map +1 -1
  172. package/dist/llm/tool_context.js +6 -0
  173. package/dist/llm/tool_context.js.map +1 -1
  174. package/dist/metrics/base.cjs.map +1 -1
  175. package/dist/metrics/base.d.cts +45 -1
  176. package/dist/metrics/base.d.ts +45 -1
  177. package/dist/metrics/base.d.ts.map +1 -1
  178. package/dist/metrics/index.cjs +5 -0
  179. package/dist/metrics/index.cjs.map +1 -1
  180. package/dist/metrics/index.d.cts +2 -1
  181. package/dist/metrics/index.d.ts +2 -1
  182. package/dist/metrics/index.d.ts.map +1 -1
  183. package/dist/metrics/index.js +6 -0
  184. package/dist/metrics/index.js.map +1 -1
  185. package/dist/metrics/model_usage.cjs +189 -0
  186. package/dist/metrics/model_usage.cjs.map +1 -0
  187. package/dist/metrics/model_usage.d.cts +92 -0
  188. package/dist/metrics/model_usage.d.ts +92 -0
  189. package/dist/metrics/model_usage.d.ts.map +1 -0
  190. package/dist/metrics/model_usage.js +164 -0
  191. package/dist/metrics/model_usage.js.map +1 -0
  192. package/dist/metrics/model_usage.test.cjs +474 -0
  193. package/dist/metrics/model_usage.test.cjs.map +1 -0
  194. package/dist/metrics/model_usage.test.js +476 -0
  195. package/dist/metrics/model_usage.test.js.map +1 -0
  196. package/dist/metrics/usage_collector.cjs +3 -0
  197. package/dist/metrics/usage_collector.cjs.map +1 -1
  198. package/dist/metrics/usage_collector.d.cts +9 -0
  199. package/dist/metrics/usage_collector.d.ts +9 -0
  200. package/dist/metrics/usage_collector.d.ts.map +1 -1
  201. package/dist/metrics/usage_collector.js +3 -0
  202. package/dist/metrics/usage_collector.js.map +1 -1
  203. package/dist/metrics/utils.cjs +9 -0
  204. package/dist/metrics/utils.cjs.map +1 -1
  205. package/dist/metrics/utils.d.ts.map +1 -1
  206. package/dist/metrics/utils.js +9 -0
  207. package/dist/metrics/utils.js.map +1 -1
  208. package/dist/stream/multi_input_stream.test.cjs +4 -0
  209. package/dist/stream/multi_input_stream.test.cjs.map +1 -1
  210. package/dist/stream/multi_input_stream.test.js +5 -1
  211. package/dist/stream/multi_input_stream.test.js.map +1 -1
  212. package/dist/stream/stream_channel.cjs +31 -0
  213. package/dist/stream/stream_channel.cjs.map +1 -1
  214. package/dist/stream/stream_channel.d.cts +4 -2
  215. package/dist/stream/stream_channel.d.ts +4 -2
  216. package/dist/stream/stream_channel.d.ts.map +1 -1
  217. package/dist/stream/stream_channel.js +31 -0
  218. package/dist/stream/stream_channel.js.map +1 -1
  219. package/dist/stt/stt.cjs +34 -2
  220. package/dist/stt/stt.cjs.map +1 -1
  221. package/dist/stt/stt.d.cts +22 -0
  222. package/dist/stt/stt.d.ts +22 -0
  223. package/dist/stt/stt.d.ts.map +1 -1
  224. package/dist/stt/stt.js +34 -2
  225. package/dist/stt/stt.js.map +1 -1
  226. package/dist/telemetry/otel_http_exporter.cjs +24 -5
  227. package/dist/telemetry/otel_http_exporter.cjs.map +1 -1
  228. package/dist/telemetry/otel_http_exporter.d.cts +1 -0
  229. package/dist/telemetry/otel_http_exporter.d.ts +1 -0
  230. package/dist/telemetry/otel_http_exporter.d.ts.map +1 -1
  231. package/dist/telemetry/otel_http_exporter.js +24 -5
  232. package/dist/telemetry/otel_http_exporter.js.map +1 -1
  233. package/dist/telemetry/trace_types.cjs +5 -5
  234. package/dist/telemetry/trace_types.cjs.map +1 -1
  235. package/dist/telemetry/trace_types.d.cts +9 -5
  236. package/dist/telemetry/trace_types.d.ts +9 -5
  237. package/dist/telemetry/trace_types.d.ts.map +1 -1
  238. package/dist/telemetry/trace_types.js +5 -5
  239. package/dist/telemetry/trace_types.js.map +1 -1
  240. package/dist/telemetry/traces.cjs +47 -8
  241. package/dist/telemetry/traces.cjs.map +1 -1
  242. package/dist/telemetry/traces.d.ts.map +1 -1
  243. package/dist/telemetry/traces.js +47 -8
  244. package/dist/telemetry/traces.js.map +1 -1
  245. package/dist/tts/tts.cjs +64 -2
  246. package/dist/tts/tts.cjs.map +1 -1
  247. package/dist/tts/tts.d.cts +34 -0
  248. package/dist/tts/tts.d.ts +34 -0
  249. package/dist/tts/tts.d.ts.map +1 -1
  250. package/dist/tts/tts.js +64 -2
  251. package/dist/tts/tts.js.map +1 -1
  252. package/dist/utils.cjs +1 -0
  253. package/dist/utils.cjs.map +1 -1
  254. package/dist/utils.d.ts.map +1 -1
  255. package/dist/utils.js +1 -0
  256. package/dist/utils.js.map +1 -1
  257. package/dist/version.cjs +1 -1
  258. package/dist/version.js +1 -1
  259. package/dist/voice/agent.cjs +34 -4
  260. package/dist/voice/agent.cjs.map +1 -1
  261. package/dist/voice/agent.d.cts +11 -2
  262. package/dist/voice/agent.d.ts +11 -2
  263. package/dist/voice/agent.d.ts.map +1 -1
  264. package/dist/voice/agent.js +34 -4
  265. package/dist/voice/agent.js.map +1 -1
  266. package/dist/voice/agent_activity.cjs +292 -44
  267. package/dist/voice/agent_activity.cjs.map +1 -1
  268. package/dist/voice/agent_activity.d.cts +27 -6
  269. package/dist/voice/agent_activity.d.ts +27 -6
  270. package/dist/voice/agent_activity.d.ts.map +1 -1
  271. package/dist/voice/agent_activity.js +293 -45
  272. package/dist/voice/agent_activity.js.map +1 -1
  273. package/dist/voice/agent_session.cjs +105 -48
  274. package/dist/voice/agent_session.cjs.map +1 -1
  275. package/dist/voice/agent_session.d.cts +90 -20
  276. package/dist/voice/agent_session.d.ts +90 -20
  277. package/dist/voice/agent_session.d.ts.map +1 -1
  278. package/dist/voice/agent_session.js +105 -46
  279. package/dist/voice/agent_session.js.map +1 -1
  280. package/dist/voice/audio_recognition.cjs +287 -6
  281. package/dist/voice/audio_recognition.cjs.map +1 -1
  282. package/dist/voice/audio_recognition.d.cts +42 -3
  283. package/dist/voice/audio_recognition.d.ts +42 -3
  284. package/dist/voice/audio_recognition.d.ts.map +1 -1
  285. package/dist/voice/audio_recognition.js +289 -7
  286. package/dist/voice/audio_recognition.js.map +1 -1
  287. package/dist/voice/client_events.cjs +554 -0
  288. package/dist/voice/client_events.cjs.map +1 -0
  289. package/dist/voice/client_events.d.cts +195 -0
  290. package/dist/voice/client_events.d.ts +195 -0
  291. package/dist/voice/client_events.d.ts.map +1 -0
  292. package/dist/voice/client_events.js +548 -0
  293. package/dist/voice/client_events.js.map +1 -0
  294. package/dist/voice/events.cjs +1 -0
  295. package/dist/voice/events.cjs.map +1 -1
  296. package/dist/voice/events.d.cts +8 -5
  297. package/dist/voice/events.d.ts +8 -5
  298. package/dist/voice/events.d.ts.map +1 -1
  299. package/dist/voice/events.js +1 -0
  300. package/dist/voice/events.js.map +1 -1
  301. package/dist/voice/generation.cjs +43 -8
  302. package/dist/voice/generation.cjs.map +1 -1
  303. package/dist/voice/generation.d.cts +3 -3
  304. package/dist/voice/generation.d.ts +3 -3
  305. package/dist/voice/generation.d.ts.map +1 -1
  306. package/dist/voice/generation.js +43 -8
  307. package/dist/voice/generation.js.map +1 -1
  308. package/dist/voice/index.cjs +1 -0
  309. package/dist/voice/index.cjs.map +1 -1
  310. package/dist/voice/index.d.cts +1 -0
  311. package/dist/voice/index.d.ts +1 -0
  312. package/dist/voice/index.d.ts.map +1 -1
  313. package/dist/voice/index.js +1 -0
  314. package/dist/voice/index.js.map +1 -1
  315. package/dist/voice/report.cjs +20 -8
  316. package/dist/voice/report.cjs.map +1 -1
  317. package/dist/voice/report.d.cts +5 -0
  318. package/dist/voice/report.d.ts +5 -0
  319. package/dist/voice/report.d.ts.map +1 -1
  320. package/dist/voice/report.js +20 -8
  321. package/dist/voice/report.js.map +1 -1
  322. package/dist/voice/report.test.cjs +106 -0
  323. package/dist/voice/report.test.cjs.map +1 -0
  324. package/dist/voice/report.test.js +105 -0
  325. package/dist/voice/report.test.js.map +1 -0
  326. package/dist/voice/room_io/room_io.cjs +16 -41
  327. package/dist/voice/room_io/room_io.cjs.map +1 -1
  328. package/dist/voice/room_io/room_io.d.cts +4 -9
  329. package/dist/voice/room_io/room_io.d.ts +4 -9
  330. package/dist/voice/room_io/room_io.d.ts.map +1 -1
  331. package/dist/voice/room_io/room_io.js +17 -43
  332. package/dist/voice/room_io/room_io.js.map +1 -1
  333. package/dist/voice/testing/fake_llm.cjs +127 -0
  334. package/dist/voice/testing/fake_llm.cjs.map +1 -0
  335. package/dist/voice/testing/fake_llm.d.cts +30 -0
  336. package/dist/voice/testing/fake_llm.d.ts +30 -0
  337. package/dist/voice/testing/fake_llm.d.ts.map +1 -0
  338. package/dist/voice/testing/fake_llm.js +103 -0
  339. package/dist/voice/testing/fake_llm.js.map +1 -0
  340. package/dist/voice/testing/index.cjs +3 -0
  341. package/dist/voice/testing/index.cjs.map +1 -1
  342. package/dist/voice/testing/index.d.cts +1 -0
  343. package/dist/voice/testing/index.d.ts +1 -0
  344. package/dist/voice/testing/index.d.ts.map +1 -1
  345. package/dist/voice/testing/index.js +2 -0
  346. package/dist/voice/testing/index.js.map +1 -1
  347. package/dist/voice/turn_config/endpointing.cjs +33 -0
  348. package/dist/voice/turn_config/endpointing.cjs.map +1 -0
  349. package/dist/voice/turn_config/endpointing.d.cts +30 -0
  350. package/dist/voice/turn_config/endpointing.d.ts +30 -0
  351. package/dist/voice/turn_config/endpointing.d.ts.map +1 -0
  352. package/dist/voice/turn_config/endpointing.js +9 -0
  353. package/dist/voice/turn_config/endpointing.js.map +1 -0
  354. package/dist/voice/turn_config/interruption.cjs +37 -0
  355. package/dist/voice/turn_config/interruption.cjs.map +1 -0
  356. package/dist/voice/turn_config/interruption.d.cts +53 -0
  357. package/dist/voice/turn_config/interruption.d.ts +53 -0
  358. package/dist/voice/turn_config/interruption.d.ts.map +1 -0
  359. package/dist/voice/turn_config/interruption.js +13 -0
  360. package/dist/voice/turn_config/interruption.js.map +1 -0
  361. package/dist/voice/turn_config/turn_handling.cjs +35 -0
  362. package/dist/voice/turn_config/turn_handling.cjs.map +1 -0
  363. package/dist/voice/turn_config/turn_handling.d.cts +36 -0
  364. package/dist/voice/turn_config/turn_handling.d.ts +36 -0
  365. package/dist/voice/turn_config/turn_handling.d.ts.map +1 -0
  366. package/dist/voice/turn_config/turn_handling.js +11 -0
  367. package/dist/voice/turn_config/turn_handling.js.map +1 -0
  368. package/dist/voice/turn_config/utils.cjs +97 -0
  369. package/dist/voice/turn_config/utils.cjs.map +1 -0
  370. package/dist/voice/turn_config/utils.d.cts +25 -0
  371. package/dist/voice/turn_config/utils.d.ts +25 -0
  372. package/dist/voice/turn_config/utils.d.ts.map +1 -0
  373. package/dist/voice/turn_config/utils.js +73 -0
  374. package/dist/voice/turn_config/utils.js.map +1 -0
  375. package/dist/voice/turn_config/utils.test.cjs +86 -0
  376. package/dist/voice/turn_config/utils.test.cjs.map +1 -0
  377. package/dist/voice/turn_config/utils.test.js +85 -0
  378. package/dist/voice/turn_config/utils.test.js.map +1 -0
  379. package/dist/voice/wire_format.cjs +798 -0
  380. package/dist/voice/wire_format.cjs.map +1 -0
  381. package/dist/voice/wire_format.d.cts +5503 -0
  382. package/dist/voice/wire_format.d.ts +5503 -0
  383. package/dist/voice/wire_format.d.ts.map +1 -0
  384. package/dist/voice/wire_format.js +728 -0
  385. package/dist/voice/wire_format.js.map +1 -0
  386. package/package.json +2 -1
  387. package/src/beta/index.ts +9 -0
  388. package/src/beta/workflows/index.ts +9 -0
  389. package/src/beta/workflows/task_group.ts +194 -0
  390. package/src/constants.ts +13 -0
  391. package/src/index.ts +2 -1
  392. package/src/inference/interruption/defaults.ts +51 -0
  393. package/src/inference/interruption/errors.ts +25 -0
  394. package/src/inference/interruption/http_transport.ts +187 -0
  395. package/src/inference/interruption/interruption_cache_entry.ts +50 -0
  396. package/src/inference/interruption/interruption_detector.ts +188 -0
  397. package/src/inference/interruption/interruption_stream.ts +467 -0
  398. package/src/inference/interruption/types.ts +84 -0
  399. package/src/inference/interruption/utils.test.ts +132 -0
  400. package/src/inference/interruption/utils.ts +137 -0
  401. package/src/inference/interruption/ws_transport.ts +402 -0
  402. package/src/inference/llm.ts +9 -12
  403. package/src/inference/stt.ts +10 -3
  404. package/src/inference/tts.ts +10 -3
  405. package/src/inference/utils.ts +29 -1
  406. package/src/llm/chat_context.test.ts +48 -0
  407. package/src/llm/chat_context.ts +161 -0
  408. package/src/llm/index.ts +2 -0
  409. package/src/llm/llm.ts +16 -0
  410. package/src/llm/realtime.ts +4 -0
  411. package/src/llm/tool_context.ts +14 -0
  412. package/src/metrics/base.ts +48 -1
  413. package/src/metrics/index.ts +11 -0
  414. package/src/metrics/model_usage.test.ts +545 -0
  415. package/src/metrics/model_usage.ts +262 -0
  416. package/src/metrics/usage_collector.ts +11 -0
  417. package/src/metrics/utils.ts +11 -0
  418. package/src/stream/multi_input_stream.test.ts +6 -1
  419. package/src/stream/stream_channel.ts +34 -2
  420. package/src/stt/stt.ts +38 -0
  421. package/src/telemetry/otel_http_exporter.ts +28 -5
  422. package/src/telemetry/trace_types.ts +11 -8
  423. package/src/telemetry/traces.ts +111 -54
  424. package/src/tts/tts.ts +69 -1
  425. package/src/utils.ts +5 -0
  426. package/src/voice/agent.ts +41 -3
  427. package/src/voice/agent_activity.ts +371 -34
  428. package/src/voice/agent_session.ts +207 -59
  429. package/src/voice/audio_recognition.ts +385 -9
  430. package/src/voice/client_events.ts +838 -0
  431. package/src/voice/events.ts +14 -4
  432. package/src/voice/generation.ts +52 -9
  433. package/src/voice/index.ts +1 -0
  434. package/src/voice/report.test.ts +117 -0
  435. package/src/voice/report.ts +29 -6
  436. package/src/voice/room_io/room_io.ts +21 -64
  437. package/src/voice/testing/fake_llm.ts +138 -0
  438. package/src/voice/testing/index.ts +2 -0
  439. package/src/voice/turn_config/endpointing.ts +33 -0
  440. package/src/voice/turn_config/interruption.ts +56 -0
  441. package/src/voice/turn_config/turn_handling.ts +45 -0
  442. package/src/voice/turn_config/utils.test.ts +100 -0
  443. package/src/voice/turn_config/utils.ts +103 -0
  444. package/src/voice/wire_format.ts +827 -0
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/llm.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport OpenAI from 'openai';\nimport {\n APIConnectionError,\n APIStatusError,\n APITimeoutError,\n DEFAULT_API_CONNECT_OPTIONS,\n type Expand,\n toError,\n} from '../index.js';\nimport * as llm from '../llm/index.js';\nimport type { APIConnectOptions } from '../types.js';\nimport { type AnyString, createAccessToken } from './utils.js';\n\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\n\nexport type OpenAIModels =\n | 'openai/gpt-5.2'\n | 'openai/gpt-5.2-chat-latest'\n | 'openai/gpt-5.1'\n | 'openai/gpt-5.1-chat-latest'\n | 'openai/gpt-5'\n | 'openai/gpt-5-mini'\n | 'openai/gpt-5-nano'\n | 'openai/gpt-4.1'\n | 'openai/gpt-4.1-mini'\n | 'openai/gpt-4.1-nano'\n | 'openai/gpt-4o'\n | 'openai/gpt-4o-mini'\n | 'openai/gpt-oss-120b';\n\nexport type GoogleModels =\n | 'google/gemini-3-pro'\n | 'google/gemini-3-flash'\n | 'google/gemini-2.5-pro'\n | 'google/gemini-2.5-flash'\n | 'google/gemini-2.5-flash-lite'\n | 'google/gemini-2.0-flash'\n | 'google/gemini-2.0-flash-lite';\n\nexport type MoonshotModels = 'moonshotai/kimi-k2-instruct';\n\nexport type DeepSeekModels = 'deepseek-ai/deepseek-v3' | 'deepseek-ai/deepseek-v3.2';\n\ntype ChatCompletionPredictionContentParam =\n Expand<OpenAI.Chat.Completions.ChatCompletionPredictionContent>;\ntype WebSearchOptions = Expand<OpenAI.Chat.Completions.ChatCompletionCreateParams.WebSearchOptions>;\ntype ToolChoice = Expand<OpenAI.Chat.Completions.ChatCompletionCreateParams['tool_choice']>;\ntype Verbosity = 'low' | 'medium' | 'high';\n\nexport interface ChatCompletionOptions extends Record<string, unknown> {\n frequency_penalty?: number;\n logit_bias?: Record<string, number>;\n logprobs?: boolean;\n max_completion_tokens?: number;\n max_tokens?: number;\n metadata?: Record<string, string>;\n modalities?: Array<'text' | 'audio'>;\n n?: number;\n parallel_tool_calls?: boolean;\n prediction?: ChatCompletionPredictionContentParam | null;\n presence_penalty?: number;\n prompt_cache_key?: string;\n reasoning_effort?: 'minimal' | 'low' | 'medium' | 'high';\n safety_identifier?: string;\n seed?: number;\n service_tier?: 'auto' | 'default' | 'flex' | 'scale' | 'priority';\n stop?: string | string[];\n store?: boolean;\n temperature?: number;\n top_logprobs?: number;\n top_p?: number;\n user?: string;\n verbosity?: Verbosity;\n web_search_options?: WebSearchOptions;\n\n // livekit-typed arguments\n tool_choice?: ToolChoice;\n // TODO(brian): support response format\n // response_format?: OpenAI.Chat.Completions.ChatCompletionCreateParams['response_format']\n}\n\nexport type LLMModels = OpenAIModels | GoogleModels | MoonshotModels | DeepSeekModels | AnyString;\n\nexport interface InferenceLLMOptions {\n model: LLMModels;\n provider?: string;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: ChatCompletionOptions;\n strictToolSchema?: boolean;\n}\n\nexport interface GatewayOptions {\n apiKey: string;\n apiSecret: string;\n}\n\n/**\n * Livekit Cloud Inference LLM\n */\nexport class LLM extends llm.LLM {\n private client: OpenAI;\n private opts: InferenceLLMOptions;\n\n constructor(opts: {\n model: LLMModels;\n provider?: string;\n baseURL?: string;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: InferenceLLMOptions['modelOptions'];\n strictToolSchema?: boolean;\n }) {\n super();\n\n const {\n model,\n provider,\n baseURL,\n apiKey,\n apiSecret,\n modelOptions,\n strictToolSchema = false,\n } = opts;\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n this.opts = {\n model,\n provider,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions: modelOptions || {},\n strictToolSchema,\n };\n\n this.client = new OpenAI({\n baseURL: this.opts.baseURL,\n apiKey: '', // leave a temporary empty string to avoid OpenAI complain about missing key\n });\n }\n\n label(): string {\n return 'inference.LLM';\n }\n\n get model(): string {\n return this.opts.model;\n }\n\n static fromModelString(modelString: string): LLM {\n return new LLM({ model: modelString });\n }\n\n chat({\n chatCtx,\n toolCtx,\n connOptions = DEFAULT_API_CONNECT_OPTIONS,\n parallelToolCalls,\n toolChoice,\n // TODO(AJS-270): Add response_format parameter support\n extraKwargs,\n }: {\n chatCtx: llm.ChatContext;\n toolCtx?: llm.ToolContext;\n connOptions?: APIConnectOptions;\n parallelToolCalls?: boolean;\n toolChoice?: llm.ToolChoice;\n // TODO(AJS-270): Add responseFormat parameter\n extraKwargs?: Record<string, unknown>;\n }): LLMStream {\n let modelOptions: Record<string, unknown> = { ...(extraKwargs || {}) };\n\n parallelToolCalls =\n parallelToolCalls !== undefined\n ? parallelToolCalls\n : this.opts.modelOptions.parallel_tool_calls;\n\n if (toolCtx && Object.keys(toolCtx).length > 0 && parallelToolCalls !== undefined) {\n modelOptions.parallel_tool_calls = parallelToolCalls;\n }\n\n toolChoice =\n toolChoice !== undefined\n ? toolChoice\n : (this.opts.modelOptions.tool_choice as llm.ToolChoice | undefined);\n\n if (toolChoice) {\n modelOptions.tool_choice = toolChoice as ToolChoice;\n }\n\n // TODO(AJS-270): Add response_format support here\n\n modelOptions = { ...modelOptions, ...this.opts.modelOptions };\n\n return new LLMStream(this, {\n model: this.opts.model,\n provider: this.opts.provider,\n client: this.client,\n chatCtx,\n toolCtx,\n connOptions,\n modelOptions,\n strictToolSchema: this.opts.strictToolSchema ?? false, // default to false if not set\n gatewayOptions: {\n apiKey: this.opts.apiKey,\n apiSecret: this.opts.apiSecret,\n },\n });\n }\n}\n\nexport class LLMStream extends llm.LLMStream {\n private model: LLMModels;\n private provider?: string;\n private providerFmt: llm.ProviderFormat;\n private client: OpenAI;\n private modelOptions: Record<string, unknown>;\n private strictToolSchema: boolean;\n\n private gatewayOptions?: GatewayOptions;\n private toolCallId?: string;\n private toolIndex?: number;\n private fncName?: string;\n private fncRawArguments?: string;\n private toolExtra?: Record<string, unknown>;\n\n constructor(\n llm: LLM,\n {\n model,\n provider,\n client,\n chatCtx,\n toolCtx,\n gatewayOptions,\n connOptions,\n modelOptions,\n providerFmt,\n strictToolSchema,\n }: {\n model: LLMModels;\n provider?: string;\n client: OpenAI;\n chatCtx: llm.ChatContext;\n toolCtx?: llm.ToolContext;\n gatewayOptions?: GatewayOptions;\n connOptions: APIConnectOptions;\n modelOptions: Record<string, unknown>;\n providerFmt?: llm.ProviderFormat;\n strictToolSchema: boolean;\n },\n ) {\n super(llm, { chatCtx, toolCtx, connOptions });\n this.client = client;\n this.gatewayOptions = gatewayOptions;\n this.provider = provider;\n this.providerFmt = providerFmt || 'openai';\n this.modelOptions = modelOptions;\n this.model = model;\n this.strictToolSchema = strictToolSchema;\n }\n\n protected async run(): Promise<void> {\n // current function call that we're waiting for full completion (args are streamed)\n // (defined inside the run method to make sure the state is reset for each run/attempt)\n let retryable = true;\n this.toolCallId = this.fncName = this.fncRawArguments = this.toolIndex = undefined;\n this.toolExtra = undefined;\n\n try {\n const messages = (await this.chatCtx.toProviderFormat(\n this.providerFmt,\n )) as OpenAI.ChatCompletionMessageParam[];\n\n const tools = this.toolCtx\n ? Object.entries(this.toolCtx).map(([name, func]) => {\n const oaiParams = {\n type: 'function' as const,\n function: {\n name,\n description: func.description,\n parameters: llm.toJsonSchema(\n func.parameters,\n true,\n this.strictToolSchema,\n ) as unknown as OpenAI.Chat.Completions.ChatCompletionFunctionTool['function']['parameters'],\n } as OpenAI.Chat.Completions.ChatCompletionFunctionTool['function'],\n };\n\n if (this.strictToolSchema) {\n oaiParams.function.strict = true;\n }\n\n return oaiParams;\n })\n : undefined;\n\n const requestOptions: Record<string, unknown> = { ...this.modelOptions };\n if (!tools) {\n delete requestOptions.tool_choice;\n }\n\n // Dynamically set the access token for the LiveKit Agent Gateway API\n if (this.gatewayOptions) {\n this.client.apiKey = await createAccessToken(\n this.gatewayOptions.apiKey,\n this.gatewayOptions.apiSecret,\n );\n }\n\n if (this.provider) {\n const extraHeaders = requestOptions.extra_headers\n ? (requestOptions.extra_headers as Record<string, string>)\n : {};\n extraHeaders['X-LiveKit-Inference-Provider'] = this.provider;\n requestOptions.extra_headers = extraHeaders;\n }\n\n const stream = await this.client.chat.completions.create(\n {\n model: this.model,\n messages,\n tools,\n stream: true,\n stream_options: { include_usage: true },\n ...requestOptions,\n },\n {\n timeout: this.connOptions.timeoutMs,\n },\n );\n\n for await (const chunk of stream) {\n for (const choice of chunk.choices) {\n if (this.abortController.signal.aborted) {\n break;\n }\n const chatChunk = this.parseChoice(chunk.id, choice);\n if (chatChunk) {\n retryable = false;\n this.queue.put(chatChunk);\n }\n }\n\n if (chunk.usage) {\n const usage = chunk.usage;\n retryable = false;\n this.queue.put({\n id: chunk.id,\n usage: {\n completionTokens: usage.completion_tokens,\n promptTokens: usage.prompt_tokens,\n promptCachedTokens: usage.prompt_tokens_details?.cached_tokens || 0,\n totalTokens: usage.total_tokens,\n },\n });\n }\n }\n } catch (error) {\n if (error instanceof OpenAI.APIConnectionTimeoutError) {\n throw new APITimeoutError({ options: { retryable } });\n } else if (error instanceof OpenAI.APIError) {\n throw new APIStatusError({\n message: error.message,\n options: {\n statusCode: error.status,\n body: error.error,\n requestId: error.requestID,\n retryable,\n },\n });\n } else {\n throw new APIConnectionError({\n message: toError(error).message,\n options: { retryable },\n });\n }\n }\n }\n\n private parseChoice(\n id: string,\n choice: OpenAI.ChatCompletionChunk.Choice,\n ): llm.ChatChunk | undefined {\n const delta = choice.delta;\n\n // https://github.com/livekit/agents/issues/688\n // the delta can be None when using Azure OpenAI (content filtering)\n if (delta === undefined) return undefined;\n\n if (delta.tool_calls) {\n // check if we have functions to calls\n for (const tool of delta.tool_calls) {\n if (!tool.function) {\n continue; // oai may add other tools in the future\n }\n\n /**\n * The way OpenAI streams tool calls is a bit tricky.\n *\n * For any new tool call, it first emits a delta tool call with id, and function name,\n * the rest of the delta chunks will only stream the remaining arguments string,\n * until a new tool call is started or the tool call is finished.\n * See below for an example.\n *\n * Choice(delta=ChoiceDelta(content=None, function_call=None, refusal=None, role='assistant', tool_calls=None), finish_reason=None, index=0, logprobs=None)\n * [ChoiceDeltaToolCall(index=0, id='call_LaVeHWUHpef9K1sd5UO8TtLg', function=ChoiceDeltaToolCallFunction(arguments='', name='get_weather'), type='function')]\n * [ChoiceDeltaToolCall(index=0, id=None, function=ChoiceDeltaToolCallFunction(arguments='\\{\"location\": \"P', name=None), type=None)]\n * [ChoiceDeltaToolCall(index=0, id=None, function=ChoiceDeltaToolCallFunction(arguments='aris\\}', name=None), type=None)]\n * [ChoiceDeltaToolCall(index=1, id='call_ThU4OmMdQXnnVmpXGOCknXIB', function=ChoiceDeltaToolCallFunction(arguments='', name='get_weather'), type='function')]\n * [ChoiceDeltaToolCall(index=1, id=None, function=ChoiceDeltaToolCallFunction(arguments='\\{\"location\": \"T', name=None), type=None)]\n * [ChoiceDeltaToolCall(index=1, id=None, function=ChoiceDeltaToolCallFunction(arguments='okyo', name=None), type=None)]\n * Choice(delta=ChoiceDelta(content=None, function_call=None, refusal=None, role=None, tool_calls=None), finish_reason='tool_calls', index=0, logprobs=None)\n */\n let callChunk: llm.ChatChunk | undefined;\n // If we have a previous tool call and this is a new one, emit the previous\n if (this.toolCallId && tool.id && tool.index !== this.toolIndex) {\n callChunk = this.createRunningToolCallChunk(id, delta);\n this.toolCallId = this.fncName = this.fncRawArguments = undefined;\n // Note: We intentionally do NOT reset toolExtra here.\n // For Gemini 3+, the thought_signature is only provided on the first tool call\n // in a parallel batch, but must be applied to ALL tool calls in the batch.\n // We preserve toolExtra so subsequent tool calls inherit the thought_signature.\n }\n\n // Start or continue building the current tool call\n if (tool.function.name) {\n this.toolIndex = tool.index;\n this.toolCallId = tool.id;\n this.fncName = tool.function.name;\n this.fncRawArguments = tool.function.arguments || '';\n // Extract extra from tool call (e.g., Google thought signatures)\n // Only update toolExtra if this tool call has extra_content.\n // Otherwise, inherit from previous tool call (for parallel Gemini tool calls).\n const newToolExtra =\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n ((tool as any).extra_content as Record<string, unknown> | undefined) ?? undefined;\n if (newToolExtra) {\n this.toolExtra = newToolExtra;\n }\n } else if (tool.function.arguments) {\n this.fncRawArguments = (this.fncRawArguments || '') + tool.function.arguments;\n }\n\n if (callChunk) {\n return callChunk;\n }\n }\n }\n\n // If we're done with tool calls, emit the final one\n if (\n choice.finish_reason &&\n ['tool_calls', 'stop'].includes(choice.finish_reason) &&\n this.toolCallId !== undefined\n ) {\n const callChunk = this.createRunningToolCallChunk(id, delta);\n this.toolCallId = this.fncName = this.fncRawArguments = undefined;\n // Reset toolExtra at the end of the response (not between parallel tool calls)\n this.toolExtra = undefined;\n return callChunk;\n }\n\n // Extract extra from delta (e.g., Google thought signatures on text parts)\n const deltaExtra =\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n ((delta as any).extra_content as Record<string, unknown> | undefined) ?? undefined;\n\n // Regular content message\n if (!delta.content && !deltaExtra) {\n return undefined;\n }\n\n return {\n id,\n delta: {\n role: 'assistant',\n content: delta.content || undefined,\n extra: deltaExtra,\n },\n };\n }\n\n private createRunningToolCallChunk(\n id: string,\n delta: OpenAI.Chat.Completions.ChatCompletionChunk.Choice.Delta,\n ): llm.ChatChunk {\n const toolExtra = this.toolExtra ? { ...this.toolExtra } : {};\n const thoughtSignature = this.extractThoughtSignature(toolExtra);\n const deltaExtra =\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n ((delta as any).extra_content as Record<string, unknown> | undefined) ?? undefined;\n\n return {\n id,\n delta: {\n role: 'assistant',\n content: delta.content || undefined,\n extra: deltaExtra,\n toolCalls: [\n llm.FunctionCall.create({\n callId: this.toolCallId || '',\n name: this.fncName || '',\n args: this.fncRawArguments || '',\n extra: toolExtra,\n thoughtSignature,\n }),\n ],\n },\n };\n }\n\n private extractThoughtSignature(extra?: Record<string, unknown>): string | undefined {\n const googleExtra = extra?.google;\n if (googleExtra && typeof googleExtra === 'object') {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n return (googleExtra as any).thoughtSignature || (googleExtra as any).thought_signature;\n }\n return undefined;\n }\n}\n"],"mappings":"AAGA,OAAO,YAAY;AACnB;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,OACK;AACP,YAAY,SAAS;AAErB,SAAyB,yBAAyB;AAElD,MAAM,mBAAmB;AAwFlB,MAAM,YAAY,IAAI,IAAI;AAAA,EACvB;AAAA,EACA;AAAA,EAER,YAAY,MAQT;AACD,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,mBAAmB;AAAA,IACrB,IAAI;AAEJ,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAEA,SAAK,OAAO;AAAA,MACV;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX,cAAc,gBAAgB,CAAC;AAAA,MAC/B;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,SAAS,KAAK,KAAK;AAAA,MACnB,QAAQ;AAAA;AAAA,IACV,CAAC;AAAA,EACH;AAAA,EAEA,QAAgB;AACd,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,OAAO,gBAAgB,aAA0B;AAC/C,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,KAAK;AAAA,IACH;AAAA,IACA;AAAA,IACA,cAAc;AAAA,IACd;AAAA,IACA;AAAA;AAAA,IAEA;AAAA,EACF,GAQc;AACZ,QAAI,eAAwC,EAAE,GAAI,eAAe,CAAC,EAAG;AAErE,wBACE,sBAAsB,SAClB,oBACA,KAAK,KAAK,aAAa;AAE7B,QAAI,WAAW,OAAO,KAAK,OAAO,EAAE,SAAS,KAAK,sBAAsB,QAAW;AACjF,mBAAa,sBAAsB;AAAA,IACrC;AAEA,iBACE,eAAe,SACX,aACC,KAAK,KAAK,aAAa;AAE9B,QAAI,YAAY;AACd,mBAAa,cAAc;AAAA,IAC7B;AAIA,mBAAe,EAAE,GAAG,cAAc,GAAG,KAAK,KAAK,aAAa;AAE5D,WAAO,IAAI,UAAU,MAAM;AAAA,MACzB,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU,KAAK,KAAK;AAAA,MACpB,QAAQ,KAAK;AAAA,MACb;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,kBAAkB,KAAK,KAAK,oBAAoB;AAAA;AAAA,MAChD,gBAAgB;AAAA,QACd,QAAQ,KAAK,KAAK;AAAA,QAClB,WAAW,KAAK,KAAK;AAAA,MACvB;AAAA,IACF,CAAC;AAAA,EACH;AACF;AAEO,MAAM,kBAAkB,IAAI,UAAU;AAAA,EACnC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YACEA,MACA;AAAA,IACE;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAYA;AACA,UAAMA,MAAK,EAAE,SAAS,SAAS,YAAY,CAAC;AAC5C,SAAK,SAAS;AACd,SAAK,iBAAiB;AACtB,SAAK,WAAW;AAChB,SAAK,cAAc,eAAe;AAClC,SAAK,eAAe;AACpB,SAAK,QAAQ;AACb,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AAtRvC;AAyRI,QAAI,YAAY;AAChB,SAAK,aAAa,KAAK,UAAU,KAAK,kBAAkB,KAAK,YAAY;AACzE,SAAK,YAAY;AAEjB,QAAI;AACF,YAAM,WAAY,MAAM,KAAK,QAAQ;AAAA,QACnC,KAAK;AAAA,MACP;AAEA,YAAM,QAAQ,KAAK,UACf,OAAO,QAAQ,KAAK,OAAO,EAAE,IAAI,CAAC,CAAC,MAAM,IAAI,MAAM;AACjD,cAAM,YAAY;AAAA,UAChB,MAAM;AAAA,UACN,UAAU;AAAA,YACR;AAAA,YACA,aAAa,KAAK;AAAA,YAClB,YAAY,IAAI;AAAA,cACd,KAAK;AAAA,cACL;AAAA,cACA,KAAK;AAAA,YACP;AAAA,UACF;AAAA,QACF;AAEA,YAAI,KAAK,kBAAkB;AACzB,oBAAU,SAAS,SAAS;AAAA,QAC9B;AAEA,eAAO;AAAA,MACT,CAAC,IACD;AAEJ,YAAM,iBAA0C,EAAE,GAAG,KAAK,aAAa;AACvE,UAAI,CAAC,OAAO;AACV,eAAO,eAAe;AAAA,MACxB;AAGA,UAAI,KAAK,gBAAgB;AACvB,aAAK,OAAO,SAAS,MAAM;AAAA,UACzB,KAAK,eAAe;AAAA,UACpB,KAAK,eAAe;AAAA,QACtB;AAAA,MACF;AAEA,UAAI,KAAK,UAAU;AACjB,cAAM,eAAe,eAAe,gBAC/B,eAAe,gBAChB,CAAC;AACL,qBAAa,8BAA8B,IAAI,KAAK;AACpD,uBAAe,gBAAgB;AAAA,MACjC;AAEA,YAAM,SAAS,MAAM,KAAK,OAAO,KAAK,YAAY;AAAA,QAChD;AAAA,UACE,OAAO,KAAK;AAAA,UACZ;AAAA,UACA;AAAA,UACA,QAAQ;AAAA,UACR,gBAAgB,EAAE,eAAe,KAAK;AAAA,UACtC,GAAG;AAAA,QACL;AAAA,QACA;AAAA,UACE,SAAS,KAAK,YAAY;AAAA,QAC5B;AAAA,MACF;AAEA,uBAAiB,SAAS,QAAQ;AAChC,mBAAW,UAAU,MAAM,SAAS;AAClC,cAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,UACF;AACA,gBAAM,YAAY,KAAK,YAAY,MAAM,IAAI,MAAM;AACnD,cAAI,WAAW;AACb,wBAAY;AACZ,iBAAK,MAAM,IAAI,SAAS;AAAA,UAC1B;AAAA,QACF;AAEA,YAAI,MAAM,OAAO;AACf,gBAAM,QAAQ,MAAM;AACpB,sBAAY;AACZ,eAAK,MAAM,IAAI;AAAA,YACb,IAAI,MAAM;AAAA,YACV,OAAO;AAAA,cACL,kBAAkB,MAAM;AAAA,cACxB,cAAc,MAAM;AAAA,cACpB,sBAAoB,WAAM,0BAAN,mBAA6B,kBAAiB;AAAA,cAClE,aAAa,MAAM;AAAA,YACrB;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF,SAAS,OAAO;AACd,UAAI,iBAAiB,OAAO,2BAA2B;AACrD,cAAM,IAAI,gBAAgB,EAAE,SAAS,EAAE,UAAU,EAAE,CAAC;AAAA,MACtD,WAAW,iBAAiB,OAAO,UAAU;AAC3C,cAAM,IAAI,eAAe;AAAA,UACvB,SAAS,MAAM;AAAA,UACf,SAAS;AAAA,YACP,YAAY,MAAM;AAAA,YAClB,MAAM,MAAM;AAAA,YACZ,WAAW,MAAM;AAAA,YACjB;AAAA,UACF;AAAA,QACF,CAAC;AAAA,MACH,OAAO;AACL,cAAM,IAAI,mBAAmB;AAAA,UAC3B,SAAS,QAAQ,KAAK,EAAE;AAAA,UACxB,SAAS,EAAE,UAAU;AAAA,QACvB,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,YACN,IACA,QAC2B;AAC3B,UAAM,QAAQ,OAAO;AAIrB,QAAI,UAAU,OAAW,QAAO;AAEhC,QAAI,MAAM,YAAY;AAEpB,iBAAW,QAAQ,MAAM,YAAY;AACnC,YAAI,CAAC,KAAK,UAAU;AAClB;AAAA,QACF;AAmBA,YAAI;AAEJ,YAAI,KAAK,cAAc,KAAK,MAAM,KAAK,UAAU,KAAK,WAAW;AAC/D,sBAAY,KAAK,2BAA2B,IAAI,KAAK;AACrD,eAAK,aAAa,KAAK,UAAU,KAAK,kBAAkB;AAAA,QAK1D;AAGA,YAAI,KAAK,SAAS,MAAM;AACtB,eAAK,YAAY,KAAK;AACtB,eAAK,aAAa,KAAK;AACvB,eAAK,UAAU,KAAK,SAAS;AAC7B,eAAK,kBAAkB,KAAK,SAAS,aAAa;AAIlD,gBAAM;AAAA;AAAA,YAEF,KAAa,iBAAyD;AAAA;AAC1E,cAAI,cAAc;AAChB,iBAAK,YAAY;AAAA,UACnB;AAAA,QACF,WAAW,KAAK,SAAS,WAAW;AAClC,eAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,SAAS;AAAA,QACtE;AAEA,YAAI,WAAW;AACb,iBAAO;AAAA,QACT;AAAA,MACF;AAAA,IACF;AAGA,QACE,OAAO,iBACP,CAAC,cAAc,MAAM,EAAE,SAAS,OAAO,aAAa,KACpD,KAAK,eAAe,QACpB;AACA,YAAM,YAAY,KAAK,2BAA2B,IAAI,KAAK;AAC3D,WAAK,aAAa,KAAK,UAAU,KAAK,kBAAkB;AAExD,WAAK,YAAY;AACjB,aAAO;AAAA,IACT;AAGA,UAAM;AAAA;AAAA,MAEF,MAAc,iBAAyD;AAAA;AAG3E,QAAI,CAAC,MAAM,WAAW,CAAC,YAAY;AACjC,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,MACL;AAAA,MACA,OAAO;AAAA,QACL,MAAM;AAAA,QACN,SAAS,MAAM,WAAW;AAAA,QAC1B,OAAO;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,2BACN,IACA,OACe;AACf,UAAM,YAAY,KAAK,YAAY,EAAE,GAAG,KAAK,UAAU,IAAI,CAAC;AAC5D,UAAM,mBAAmB,KAAK,wBAAwB,SAAS;AAC/D,UAAM;AAAA;AAAA,MAEF,MAAc,iBAAyD;AAAA;AAE3E,WAAO;AAAA,MACL;AAAA,MACA,OAAO;AAAA,QACL,MAAM;AAAA,QACN,SAAS,MAAM,WAAW;AAAA,QAC1B,OAAO;AAAA,QACP,WAAW;AAAA,UACT,IAAI,aAAa,OAAO;AAAA,YACtB,QAAQ,KAAK,cAAc;AAAA,YAC3B,MAAM,KAAK,WAAW;AAAA,YACtB,MAAM,KAAK,mBAAmB;AAAA,YAC9B,OAAO;AAAA,YACP;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,wBAAwB,OAAqD;AACnF,UAAM,cAAc,+BAAO;AAC3B,QAAI,eAAe,OAAO,gBAAgB,UAAU;AAElD,aAAQ,YAAoB,oBAAqB,YAAoB;AAAA,IACvE;AACA,WAAO;AAAA,EACT;AACF;","names":["llm"]}
1
+ {"version":3,"sources":["../../src/inference/llm.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport OpenAI from 'openai';\nimport { APIConnectionError, APIStatusError, APITimeoutError } from '../_exceptions.js';\nimport * as llm from '../llm/index.js';\nimport { DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport type { APIConnectOptions } from '../types.js';\nimport { type Expand, toError } from '../utils.js';\nimport { type AnyString, createAccessToken, getDefaultInferenceUrl } from './utils.js';\n\nexport type OpenAIModels =\n | 'openai/gpt-5.2'\n | 'openai/gpt-5.2-chat-latest'\n | 'openai/gpt-5.1'\n | 'openai/gpt-5.1-chat-latest'\n | 'openai/gpt-5'\n | 'openai/gpt-5-mini'\n | 'openai/gpt-5-nano'\n | 'openai/gpt-4.1'\n | 'openai/gpt-4.1-mini'\n | 'openai/gpt-4.1-nano'\n | 'openai/gpt-4o'\n | 'openai/gpt-4o-mini'\n | 'openai/gpt-oss-120b';\n\nexport type GoogleModels =\n | 'google/gemini-3-pro'\n | 'google/gemini-3-flash'\n | 'google/gemini-2.5-pro'\n | 'google/gemini-2.5-flash'\n | 'google/gemini-2.5-flash-lite'\n | 'google/gemini-2.0-flash'\n | 'google/gemini-2.0-flash-lite';\n\nexport type MoonshotModels = 'moonshotai/kimi-k2-instruct';\n\nexport type DeepSeekModels = 'deepseek-ai/deepseek-v3' | 'deepseek-ai/deepseek-v3.2';\n\ntype ChatCompletionPredictionContentParam =\n Expand<OpenAI.Chat.Completions.ChatCompletionPredictionContent>;\ntype WebSearchOptions = Expand<OpenAI.Chat.Completions.ChatCompletionCreateParams.WebSearchOptions>;\ntype ToolChoice = Expand<OpenAI.Chat.Completions.ChatCompletionCreateParams['tool_choice']>;\ntype Verbosity = 'low' | 'medium' | 'high';\n\nexport interface ChatCompletionOptions extends Record<string, unknown> {\n frequency_penalty?: number;\n logit_bias?: Record<string, number>;\n logprobs?: boolean;\n max_completion_tokens?: number;\n max_tokens?: number;\n metadata?: Record<string, string>;\n modalities?: Array<'text' | 'audio'>;\n n?: number;\n parallel_tool_calls?: boolean;\n prediction?: ChatCompletionPredictionContentParam | null;\n presence_penalty?: number;\n prompt_cache_key?: string;\n reasoning_effort?: 'minimal' | 'low' | 'medium' | 'high';\n safety_identifier?: string;\n seed?: number;\n service_tier?: 'auto' | 'default' | 'flex' | 'scale' | 'priority';\n stop?: string | string[];\n store?: boolean;\n temperature?: number;\n top_logprobs?: number;\n top_p?: number;\n user?: string;\n verbosity?: Verbosity;\n web_search_options?: WebSearchOptions;\n\n // livekit-typed arguments\n tool_choice?: ToolChoice;\n // TODO(brian): support response format\n // response_format?: OpenAI.Chat.Completions.ChatCompletionCreateParams['response_format']\n}\n\nexport type LLMModels = OpenAIModels | GoogleModels | MoonshotModels | DeepSeekModels | AnyString;\n\nexport interface InferenceLLMOptions {\n model: LLMModels;\n provider?: string;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: ChatCompletionOptions;\n strictToolSchema?: boolean;\n}\n\nexport interface GatewayOptions {\n apiKey: string;\n apiSecret: string;\n}\n\n/**\n * Livekit Cloud Inference LLM\n */\nexport class LLM extends llm.LLM {\n private client: OpenAI;\n private opts: InferenceLLMOptions;\n\n constructor(opts: {\n model: LLMModels;\n provider?: string;\n baseURL?: string;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: InferenceLLMOptions['modelOptions'];\n strictToolSchema?: boolean;\n }) {\n super();\n\n const {\n model,\n provider,\n baseURL,\n apiKey,\n apiSecret,\n modelOptions,\n strictToolSchema = false,\n } = opts;\n\n const lkBaseURL = baseURL || getDefaultInferenceUrl();\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n this.opts = {\n model,\n provider,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions: modelOptions || {},\n strictToolSchema,\n };\n\n this.client = new OpenAI({\n baseURL: this.opts.baseURL,\n apiKey: '', // leave a temporary empty string to avoid OpenAI complain about missing key\n });\n }\n\n label(): string {\n return 'inference.LLM';\n }\n\n get model(): string {\n return this.opts.model;\n }\n\n get provider(): string {\n return 'livekit';\n }\n\n static fromModelString(modelString: string): LLM {\n return new LLM({ model: modelString });\n }\n\n chat({\n chatCtx,\n toolCtx,\n connOptions = DEFAULT_API_CONNECT_OPTIONS,\n parallelToolCalls,\n toolChoice,\n // TODO(AJS-270): Add response_format parameter support\n extraKwargs,\n }: {\n chatCtx: llm.ChatContext;\n toolCtx?: llm.ToolContext;\n connOptions?: APIConnectOptions;\n parallelToolCalls?: boolean;\n toolChoice?: llm.ToolChoice;\n // TODO(AJS-270): Add responseFormat parameter\n extraKwargs?: Record<string, unknown>;\n }): LLMStream {\n let modelOptions: Record<string, unknown> = { ...(extraKwargs || {}) };\n\n parallelToolCalls =\n parallelToolCalls !== undefined\n ? parallelToolCalls\n : this.opts.modelOptions.parallel_tool_calls;\n\n if (toolCtx && Object.keys(toolCtx).length > 0 && parallelToolCalls !== undefined) {\n modelOptions.parallel_tool_calls = parallelToolCalls;\n }\n\n toolChoice =\n toolChoice !== undefined\n ? toolChoice\n : (this.opts.modelOptions.tool_choice as llm.ToolChoice | undefined);\n\n if (toolChoice) {\n modelOptions.tool_choice = toolChoice as ToolChoice;\n }\n\n // TODO(AJS-270): Add response_format support here\n\n modelOptions = { ...modelOptions, ...this.opts.modelOptions };\n\n return new LLMStream(this, {\n model: this.opts.model,\n provider: this.opts.provider,\n client: this.client,\n chatCtx,\n toolCtx,\n connOptions,\n modelOptions,\n strictToolSchema: this.opts.strictToolSchema ?? false, // default to false if not set\n gatewayOptions: {\n apiKey: this.opts.apiKey,\n apiSecret: this.opts.apiSecret,\n },\n });\n }\n}\n\nexport class LLMStream extends llm.LLMStream {\n private model: LLMModels;\n private provider?: string;\n private providerFmt: llm.ProviderFormat;\n private client: OpenAI;\n private modelOptions: Record<string, unknown>;\n private strictToolSchema: boolean;\n\n private gatewayOptions?: GatewayOptions;\n private toolCallId?: string;\n private toolIndex?: number;\n private fncName?: string;\n private fncRawArguments?: string;\n private toolExtra?: Record<string, unknown>;\n\n constructor(\n llm: LLM,\n {\n model,\n provider,\n client,\n chatCtx,\n toolCtx,\n gatewayOptions,\n connOptions,\n modelOptions,\n providerFmt,\n strictToolSchema,\n }: {\n model: LLMModels;\n provider?: string;\n client: OpenAI;\n chatCtx: llm.ChatContext;\n toolCtx?: llm.ToolContext;\n gatewayOptions?: GatewayOptions;\n connOptions: APIConnectOptions;\n modelOptions: Record<string, unknown>;\n providerFmt?: llm.ProviderFormat;\n strictToolSchema: boolean;\n },\n ) {\n super(llm, { chatCtx, toolCtx, connOptions });\n this.client = client;\n this.gatewayOptions = gatewayOptions;\n this.provider = provider;\n this.providerFmt = providerFmt || 'openai';\n this.modelOptions = modelOptions;\n this.model = model;\n this.strictToolSchema = strictToolSchema;\n }\n\n protected async run(): Promise<void> {\n // current function call that we're waiting for full completion (args are streamed)\n // (defined inside the run method to make sure the state is reset for each run/attempt)\n let retryable = true;\n this.toolCallId = this.fncName = this.fncRawArguments = this.toolIndex = undefined;\n this.toolExtra = undefined;\n\n try {\n const messages = (await this.chatCtx.toProviderFormat(\n this.providerFmt,\n )) as OpenAI.ChatCompletionMessageParam[];\n\n const tools = this.toolCtx\n ? Object.entries(this.toolCtx).map(([name, func]) => {\n const oaiParams = {\n type: 'function' as const,\n function: {\n name,\n description: func.description,\n parameters: llm.toJsonSchema(\n func.parameters,\n true,\n this.strictToolSchema,\n ) as unknown as OpenAI.Chat.Completions.ChatCompletionFunctionTool['function']['parameters'],\n } as OpenAI.Chat.Completions.ChatCompletionFunctionTool['function'],\n };\n\n if (this.strictToolSchema) {\n oaiParams.function.strict = true;\n }\n\n return oaiParams;\n })\n : undefined;\n\n const requestOptions: Record<string, unknown> = { ...this.modelOptions };\n if (!tools) {\n delete requestOptions.tool_choice;\n }\n\n // Dynamically set the access token for the LiveKit Agent Gateway API\n if (this.gatewayOptions) {\n this.client.apiKey = await createAccessToken(\n this.gatewayOptions.apiKey,\n this.gatewayOptions.apiSecret,\n );\n }\n\n if (this.provider) {\n const extraHeaders = requestOptions.extra_headers\n ? (requestOptions.extra_headers as Record<string, string>)\n : {};\n extraHeaders['X-LiveKit-Inference-Provider'] = this.provider;\n requestOptions.extra_headers = extraHeaders;\n }\n\n const stream = await this.client.chat.completions.create(\n {\n model: this.model,\n messages,\n tools,\n stream: true,\n stream_options: { include_usage: true },\n ...requestOptions,\n },\n {\n timeout: this.connOptions.timeoutMs,\n },\n );\n\n for await (const chunk of stream) {\n for (const choice of chunk.choices) {\n if (this.abortController.signal.aborted) {\n break;\n }\n const chatChunk = this.parseChoice(chunk.id, choice);\n if (chatChunk) {\n retryable = false;\n this.queue.put(chatChunk);\n }\n }\n\n if (chunk.usage) {\n const usage = chunk.usage;\n retryable = false;\n this.queue.put({\n id: chunk.id,\n usage: {\n completionTokens: usage.completion_tokens,\n promptTokens: usage.prompt_tokens,\n promptCachedTokens: usage.prompt_tokens_details?.cached_tokens || 0,\n totalTokens: usage.total_tokens,\n },\n });\n }\n }\n } catch (error) {\n if (error instanceof OpenAI.APIConnectionTimeoutError) {\n throw new APITimeoutError({ options: { retryable } });\n } else if (error instanceof OpenAI.APIError) {\n throw new APIStatusError({\n message: error.message,\n options: {\n statusCode: error.status,\n body: error.error,\n requestId: error.requestID,\n retryable,\n },\n });\n } else {\n throw new APIConnectionError({\n message: toError(error).message,\n options: { retryable },\n });\n }\n }\n }\n\n private parseChoice(\n id: string,\n choice: OpenAI.ChatCompletionChunk.Choice,\n ): llm.ChatChunk | undefined {\n const delta = choice.delta;\n\n // https://github.com/livekit/agents/issues/688\n // the delta can be None when using Azure OpenAI (content filtering)\n if (delta === undefined) return undefined;\n\n if (delta.tool_calls) {\n // check if we have functions to calls\n for (const tool of delta.tool_calls) {\n if (!tool.function) {\n continue; // oai may add other tools in the future\n }\n\n /**\n * The way OpenAI streams tool calls is a bit tricky.\n *\n * For any new tool call, it first emits a delta tool call with id, and function name,\n * the rest of the delta chunks will only stream the remaining arguments string,\n * until a new tool call is started or the tool call is finished.\n * See below for an example.\n *\n * Choice(delta=ChoiceDelta(content=None, function_call=None, refusal=None, role='assistant', tool_calls=None), finish_reason=None, index=0, logprobs=None)\n * [ChoiceDeltaToolCall(index=0, id='call_LaVeHWUHpef9K1sd5UO8TtLg', function=ChoiceDeltaToolCallFunction(arguments='', name='get_weather'), type='function')]\n * [ChoiceDeltaToolCall(index=0, id=None, function=ChoiceDeltaToolCallFunction(arguments='\\{\"location\": \"P', name=None), type=None)]\n * [ChoiceDeltaToolCall(index=0, id=None, function=ChoiceDeltaToolCallFunction(arguments='aris\\}', name=None), type=None)]\n * [ChoiceDeltaToolCall(index=1, id='call_ThU4OmMdQXnnVmpXGOCknXIB', function=ChoiceDeltaToolCallFunction(arguments='', name='get_weather'), type='function')]\n * [ChoiceDeltaToolCall(index=1, id=None, function=ChoiceDeltaToolCallFunction(arguments='\\{\"location\": \"T', name=None), type=None)]\n * [ChoiceDeltaToolCall(index=1, id=None, function=ChoiceDeltaToolCallFunction(arguments='okyo', name=None), type=None)]\n * Choice(delta=ChoiceDelta(content=None, function_call=None, refusal=None, role=None, tool_calls=None), finish_reason='tool_calls', index=0, logprobs=None)\n */\n let callChunk: llm.ChatChunk | undefined;\n // If we have a previous tool call and this is a new one, emit the previous\n if (this.toolCallId && tool.id && tool.index !== this.toolIndex) {\n callChunk = this.createRunningToolCallChunk(id, delta);\n this.toolCallId = this.fncName = this.fncRawArguments = undefined;\n // Note: We intentionally do NOT reset toolExtra here.\n // For Gemini 3+, the thought_signature is only provided on the first tool call\n // in a parallel batch, but must be applied to ALL tool calls in the batch.\n // We preserve toolExtra so subsequent tool calls inherit the thought_signature.\n }\n\n // Start or continue building the current tool call\n if (tool.function.name) {\n this.toolIndex = tool.index;\n this.toolCallId = tool.id;\n this.fncName = tool.function.name;\n this.fncRawArguments = tool.function.arguments || '';\n // Extract extra from tool call (e.g., Google thought signatures)\n // Only update toolExtra if this tool call has extra_content.\n // Otherwise, inherit from previous tool call (for parallel Gemini tool calls).\n const newToolExtra =\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n ((tool as any).extra_content as Record<string, unknown> | undefined) ?? undefined;\n if (newToolExtra) {\n this.toolExtra = newToolExtra;\n }\n } else if (tool.function.arguments) {\n this.fncRawArguments = (this.fncRawArguments || '') + tool.function.arguments;\n }\n\n if (callChunk) {\n return callChunk;\n }\n }\n }\n\n // If we're done with tool calls, emit the final one\n if (\n choice.finish_reason &&\n ['tool_calls', 'stop'].includes(choice.finish_reason) &&\n this.toolCallId !== undefined\n ) {\n const callChunk = this.createRunningToolCallChunk(id, delta);\n this.toolCallId = this.fncName = this.fncRawArguments = undefined;\n // Reset toolExtra at the end of the response (not between parallel tool calls)\n this.toolExtra = undefined;\n return callChunk;\n }\n\n // Extract extra from delta (e.g., Google thought signatures on text parts)\n const deltaExtra =\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n ((delta as any).extra_content as Record<string, unknown> | undefined) ?? undefined;\n\n // Regular content message\n if (!delta.content && !deltaExtra) {\n return undefined;\n }\n\n return {\n id,\n delta: {\n role: 'assistant',\n content: delta.content || undefined,\n extra: deltaExtra,\n },\n };\n }\n\n private createRunningToolCallChunk(\n id: string,\n delta: OpenAI.Chat.Completions.ChatCompletionChunk.Choice.Delta,\n ): llm.ChatChunk {\n const toolExtra = this.toolExtra ? { ...this.toolExtra } : {};\n const thoughtSignature = this.extractThoughtSignature(toolExtra);\n const deltaExtra =\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n ((delta as any).extra_content as Record<string, unknown> | undefined) ?? undefined;\n\n return {\n id,\n delta: {\n role: 'assistant',\n content: delta.content || undefined,\n extra: deltaExtra,\n toolCalls: [\n llm.FunctionCall.create({\n callId: this.toolCallId || '',\n name: this.fncName || '',\n args: this.fncRawArguments || '',\n extra: toolExtra,\n thoughtSignature,\n }),\n ],\n },\n };\n }\n\n private extractThoughtSignature(extra?: Record<string, unknown>): string | undefined {\n const googleExtra = extra?.google;\n if (googleExtra && typeof googleExtra === 'object') {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n return (googleExtra as any).thoughtSignature || (googleExtra as any).thought_signature;\n }\n return undefined;\n }\n}\n"],"mappings":"AAGA,OAAO,YAAY;AACnB,SAAS,oBAAoB,gBAAgB,uBAAuB;AACpE,YAAY,SAAS;AACrB,SAAS,mCAAmC;AAE5C,SAAsB,eAAe;AACrC,SAAyB,mBAAmB,8BAA8B;AAwFnE,MAAM,YAAY,IAAI,IAAI;AAAA,EACvB;AAAA,EACA;AAAA,EAER,YAAY,MAQT;AACD,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,mBAAmB;AAAA,IACrB,IAAI;AAEJ,UAAM,YAAY,WAAW,uBAAuB;AACpD,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAEA,SAAK,OAAO;AAAA,MACV;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX,cAAc,gBAAgB,CAAC;AAAA,MAC/B;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,SAAS,KAAK,KAAK;AAAA,MACnB,QAAQ;AAAA;AAAA,IACV,CAAC;AAAA,EACH;AAAA,EAEA,QAAgB;AACd,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAA0B;AAC/C,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,KAAK;AAAA,IACH;AAAA,IACA;AAAA,IACA,cAAc;AAAA,IACd;AAAA,IACA;AAAA;AAAA,IAEA;AAAA,EACF,GAQc;AACZ,QAAI,eAAwC,EAAE,GAAI,eAAe,CAAC,EAAG;AAErE,wBACE,sBAAsB,SAClB,oBACA,KAAK,KAAK,aAAa;AAE7B,QAAI,WAAW,OAAO,KAAK,OAAO,EAAE,SAAS,KAAK,sBAAsB,QAAW;AACjF,mBAAa,sBAAsB;AAAA,IACrC;AAEA,iBACE,eAAe,SACX,aACC,KAAK,KAAK,aAAa;AAE9B,QAAI,YAAY;AACd,mBAAa,cAAc;AAAA,IAC7B;AAIA,mBAAe,EAAE,GAAG,cAAc,GAAG,KAAK,KAAK,aAAa;AAE5D,WAAO,IAAI,UAAU,MAAM;AAAA,MACzB,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU,KAAK,KAAK;AAAA,MACpB,QAAQ,KAAK;AAAA,MACb;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,kBAAkB,KAAK,KAAK,oBAAoB;AAAA;AAAA,MAChD,gBAAgB;AAAA,QACd,QAAQ,KAAK,KAAK;AAAA,QAClB,WAAW,KAAK,KAAK;AAAA,MACvB;AAAA,IACF,CAAC;AAAA,EACH;AACF;AAEO,MAAM,kBAAkB,IAAI,UAAU;AAAA,EACnC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YACEA,MACA;AAAA,IACE;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAYA;AACA,UAAMA,MAAK,EAAE,SAAS,SAAS,YAAY,CAAC;AAC5C,SAAK,SAAS;AACd,SAAK,iBAAiB;AACtB,SAAK,WAAW;AAChB,SAAK,cAAc,eAAe;AAClC,SAAK,eAAe;AACpB,SAAK,QAAQ;AACb,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AAnRvC;AAsRI,QAAI,YAAY;AAChB,SAAK,aAAa,KAAK,UAAU,KAAK,kBAAkB,KAAK,YAAY;AACzE,SAAK,YAAY;AAEjB,QAAI;AACF,YAAM,WAAY,MAAM,KAAK,QAAQ;AAAA,QACnC,KAAK;AAAA,MACP;AAEA,YAAM,QAAQ,KAAK,UACf,OAAO,QAAQ,KAAK,OAAO,EAAE,IAAI,CAAC,CAAC,MAAM,IAAI,MAAM;AACjD,cAAM,YAAY;AAAA,UAChB,MAAM;AAAA,UACN,UAAU;AAAA,YACR;AAAA,YACA,aAAa,KAAK;AAAA,YAClB,YAAY,IAAI;AAAA,cACd,KAAK;AAAA,cACL;AAAA,cACA,KAAK;AAAA,YACP;AAAA,UACF;AAAA,QACF;AAEA,YAAI,KAAK,kBAAkB;AACzB,oBAAU,SAAS,SAAS;AAAA,QAC9B;AAEA,eAAO;AAAA,MACT,CAAC,IACD;AAEJ,YAAM,iBAA0C,EAAE,GAAG,KAAK,aAAa;AACvE,UAAI,CAAC,OAAO;AACV,eAAO,eAAe;AAAA,MACxB;AAGA,UAAI,KAAK,gBAAgB;AACvB,aAAK,OAAO,SAAS,MAAM;AAAA,UACzB,KAAK,eAAe;AAAA,UACpB,KAAK,eAAe;AAAA,QACtB;AAAA,MACF;AAEA,UAAI,KAAK,UAAU;AACjB,cAAM,eAAe,eAAe,gBAC/B,eAAe,gBAChB,CAAC;AACL,qBAAa,8BAA8B,IAAI,KAAK;AACpD,uBAAe,gBAAgB;AAAA,MACjC;AAEA,YAAM,SAAS,MAAM,KAAK,OAAO,KAAK,YAAY;AAAA,QAChD;AAAA,UACE,OAAO,KAAK;AAAA,UACZ;AAAA,UACA;AAAA,UACA,QAAQ;AAAA,UACR,gBAAgB,EAAE,eAAe,KAAK;AAAA,UACtC,GAAG;AAAA,QACL;AAAA,QACA;AAAA,UACE,SAAS,KAAK,YAAY;AAAA,QAC5B;AAAA,MACF;AAEA,uBAAiB,SAAS,QAAQ;AAChC,mBAAW,UAAU,MAAM,SAAS;AAClC,cAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,UACF;AACA,gBAAM,YAAY,KAAK,YAAY,MAAM,IAAI,MAAM;AACnD,cAAI,WAAW;AACb,wBAAY;AACZ,iBAAK,MAAM,IAAI,SAAS;AAAA,UAC1B;AAAA,QACF;AAEA,YAAI,MAAM,OAAO;AACf,gBAAM,QAAQ,MAAM;AACpB,sBAAY;AACZ,eAAK,MAAM,IAAI;AAAA,YACb,IAAI,MAAM;AAAA,YACV,OAAO;AAAA,cACL,kBAAkB,MAAM;AAAA,cACxB,cAAc,MAAM;AAAA,cACpB,sBAAoB,WAAM,0BAAN,mBAA6B,kBAAiB;AAAA,cAClE,aAAa,MAAM;AAAA,YACrB;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF,SAAS,OAAO;AACd,UAAI,iBAAiB,OAAO,2BAA2B;AACrD,cAAM,IAAI,gBAAgB,EAAE,SAAS,EAAE,UAAU,EAAE,CAAC;AAAA,MACtD,WAAW,iBAAiB,OAAO,UAAU;AAC3C,cAAM,IAAI,eAAe;AAAA,UACvB,SAAS,MAAM;AAAA,UACf,SAAS;AAAA,YACP,YAAY,MAAM;AAAA,YAClB,MAAM,MAAM;AAAA,YACZ,WAAW,MAAM;AAAA,YACjB;AAAA,UACF;AAAA,QACF,CAAC;AAAA,MACH,OAAO;AACL,cAAM,IAAI,mBAAmB;AAAA,UAC3B,SAAS,QAAQ,KAAK,EAAE;AAAA,UACxB,SAAS,EAAE,UAAU;AAAA,QACvB,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,YACN,IACA,QAC2B;AAC3B,UAAM,QAAQ,OAAO;AAIrB,QAAI,UAAU,OAAW,QAAO;AAEhC,QAAI,MAAM,YAAY;AAEpB,iBAAW,QAAQ,MAAM,YAAY;AACnC,YAAI,CAAC,KAAK,UAAU;AAClB;AAAA,QACF;AAmBA,YAAI;AAEJ,YAAI,KAAK,cAAc,KAAK,MAAM,KAAK,UAAU,KAAK,WAAW;AAC/D,sBAAY,KAAK,2BAA2B,IAAI,KAAK;AACrD,eAAK,aAAa,KAAK,UAAU,KAAK,kBAAkB;AAAA,QAK1D;AAGA,YAAI,KAAK,SAAS,MAAM;AACtB,eAAK,YAAY,KAAK;AACtB,eAAK,aAAa,KAAK;AACvB,eAAK,UAAU,KAAK,SAAS;AAC7B,eAAK,kBAAkB,KAAK,SAAS,aAAa;AAIlD,gBAAM;AAAA;AAAA,YAEF,KAAa,iBAAyD;AAAA;AAC1E,cAAI,cAAc;AAChB,iBAAK,YAAY;AAAA,UACnB;AAAA,QACF,WAAW,KAAK,SAAS,WAAW;AAClC,eAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,SAAS;AAAA,QACtE;AAEA,YAAI,WAAW;AACb,iBAAO;AAAA,QACT;AAAA,MACF;AAAA,IACF;AAGA,QACE,OAAO,iBACP,CAAC,cAAc,MAAM,EAAE,SAAS,OAAO,aAAa,KACpD,KAAK,eAAe,QACpB;AACA,YAAM,YAAY,KAAK,2BAA2B,IAAI,KAAK;AAC3D,WAAK,aAAa,KAAK,UAAU,KAAK,kBAAkB;AAExD,WAAK,YAAY;AACjB,aAAO;AAAA,IACT;AAGA,UAAM;AAAA;AAAA,MAEF,MAAc,iBAAyD;AAAA;AAG3E,QAAI,CAAC,MAAM,WAAW,CAAC,YAAY;AACjC,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,MACL;AAAA,MACA,OAAO;AAAA,QACL,MAAM;AAAA,QACN,SAAS,MAAM,WAAW;AAAA,QAC1B,OAAO;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,2BACN,IACA,OACe;AACf,UAAM,YAAY,KAAK,YAAY,EAAE,GAAG,KAAK,UAAU,IAAI,CAAC;AAC5D,UAAM,mBAAmB,KAAK,wBAAwB,SAAS;AAC/D,UAAM;AAAA;AAAA,MAEF,MAAc,iBAAyD;AAAA;AAE3E,WAAO;AAAA,MACL;AAAA,MACA,OAAO;AAAA,QACL,MAAM;AAAA,QACN,SAAS,MAAM,WAAW;AAAA,QAC1B,OAAO;AAAA,QACP,WAAW;AAAA,UACT,IAAI,aAAa,OAAO;AAAA,YACtB,QAAQ,KAAK,cAAc;AAAA,YAC3B,MAAM,KAAK,WAAW;AAAA,YACtB,MAAM,KAAK,mBAAmB;AAAA,YAC9B,OAAO;AAAA,YACP;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,wBAAwB,OAAqD;AACnF,UAAM,cAAc,+BAAO;AAC3B,QAAI,eAAe,OAAO,gBAAgB,UAAU;AAElD,aAAQ,YAAoB,oBAAqB,YAAoB;AAAA,IACvE;AACA,WAAO;AAAA,EACT;AACF;","names":["llm"]}
@@ -57,7 +57,6 @@ function normalizeSTTFallback(fallback) {
57
57
  }
58
58
  const DEFAULT_ENCODING = "pcm_s16le";
59
59
  const DEFAULT_SAMPLE_RATE = 16e3;
60
- const DEFAULT_BASE_URL = "wss://agent-gateway.livekit.cloud/v1";
61
60
  const DEFAULT_CANCEL_TIMEOUT = 5e3;
62
61
  class STT extends import_stt.STT {
63
62
  opts;
@@ -77,7 +76,7 @@ class STT extends import_stt.STT {
77
76
  fallback,
78
77
  connOptions
79
78
  } = opts || {};
80
- const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;
79
+ const lkBaseURL = baseURL || (0, import_utils2.getDefaultInferenceUrl)();
81
80
  const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;
82
81
  if (!lkApiKey) {
83
82
  throw new Error("apiKey is required: pass apiKey or set LIVEKIT_API_KEY");
@@ -120,6 +119,12 @@ class STT extends import_stt.STT {
120
119
  get label() {
121
120
  return "inference.STT";
122
121
  }
122
+ get model() {
123
+ return this.opts.model ?? "auto";
124
+ }
125
+ get provider() {
126
+ return "livekit";
127
+ }
123
128
  static fromModelString(modelString) {
124
129
  const [model, language] = parseSTTModelString(modelString);
125
130
  return new STT({ model, language });
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\n STT as BaseSTT,\n SpeechStream as BaseSpeechStream,\n type SpeechData,\n type SpeechEvent,\n SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type TimedString, createTimedString } from '../voice/io.js';\nimport {\n type SttServerEvent,\n type SttTranscriptEvent,\n sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type DeepgramModels =\n | 'deepgram/flux-general'\n | 'deepgram/nova-3'\n | 'deepgram/nova-3-medical'\n | 'deepgram/nova-2'\n | 'deepgram/nova-2-medical'\n | 'deepgram/nova-2-conversationalai'\n | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels =\n | 'assemblyai/universal-streaming'\n | 'assemblyai/universal-streaming-multilingual';\n\nexport type ElevenlabsSTTModels = 'elevenlabs/scribe_v2_realtime';\n\nexport interface CartesiaOptions {\n /** Minimum volume threshold. Default: not specified. */\n min_volume?: number;\n /** Maximum silence duration in seconds. Default: not specified. */\n max_silence_duration_secs?: number;\n}\n\nexport interface DeepgramOptions {\n /** Enable filler words. Default: true. */\n filler_words?: boolean;\n /** Enable interim results. Default: true. */\n interim_results?: boolean;\n /** Endpointing timeout in milliseconds. Default: 25. */\n endpointing?: number;\n /** Enable punctuation. Default: false. */\n punctuate?: boolean;\n /** Enable smart formatting. */\n smart_format?: boolean;\n /** Keywords with boost values. */\n keywords?: Array<[string, number]>;\n /** Key terms for recognition. */\n keyterms?: string[];\n /** Enable profanity filter. */\n profanity_filter?: boolean;\n /** Convert spoken numbers to numerals. */\n numerals?: boolean;\n /** Opt out of model improvement program. */\n mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n /** Enable turn formatting. Default: false. */\n format_turns?: boolean;\n /** End of turn confidence threshold. Default: 0.01. */\n end_of_turn_confidence_threshold?: number;\n /** Minimum silence duration in milliseconds when confident about end of turn. Default: 0. */\n min_end_of_turn_silence_when_confident?: number;\n /** Maximum turn silence in milliseconds. Default: not specified. */\n max_turn_silence?: number;\n /** Key terms prompt for recognition. Default: not specified. */\n keyterms_prompt?: string[];\n}\n\nexport type STTLanguages =\n | 'multi'\n | 'en'\n | 'de'\n | 'es'\n | 'fr'\n | 'ja'\n | 'pt'\n | 'zh'\n | 'hi'\n | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels | ElevenlabsSTTModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n ? DeepgramOptions\n : TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends AssemblyaiModels\n ? AssemblyAIOptions\n : Record<string, unknown>;\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface STTFallbackModel {\n /** Model name (e.g. \"deepgram/nova-3\", \"assemblyai/universal-streaming\", \"cartesia/ink-whisper\"). */\n model: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type STTFallbackModelType = STTFallbackModel | string;\n\n/** Parse a model string into [model, language]. Language is undefined if not specified. */\nexport function parseSTTModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** Normalize a single or list of FallbackModelType into STTFallbackModel[]. */\nexport function normalizeSTTFallback(\n fallback: STTFallbackModelType | STTFallbackModelType[],\n): STTFallbackModel[] {\n const makeFallback = (model: STTFallbackModelType): STTFallbackModel => {\n if (typeof model === 'string') {\n const [name] = parseSTTModelString(model);\n return { model: name };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'wss://agent-gateway.livekit.cloud/v1';\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n model?: TModel;\n language?: STTLanguages;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\n fallback?: STTFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n private opts: InferenceSTTOptions<TModel>;\n private streams: Set<SpeechStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts?: {\n model?: ModelWithLanguage;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\n fallback?: STTFallbackModelType | STTFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n super({ streaming: true, interimResults: true, alignedTranscript: 'word' });\n\n const {\n model,\n language,\n baseURL,\n encoding = DEFAULT_ENCODING,\n sampleRate = DEFAULT_SAMPLE_RATE,\n apiKey,\n apiSecret,\n modelOptions = {} as STTOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // Parse language from model string if provided: \"provider/model:language\"\n let nextModel = model;\n let nextLanguage = language;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const languageFromModel = nextModel.slice(idx + 1) as STTLanguages;\n if (nextLanguage && nextLanguage !== languageFromModel) {\n this.#logger.warn(\n '`language` is provided via both argument and model, using the one from the argument',\n { language: nextLanguage, model: nextModel },\n );\n } else {\n nextLanguage = languageFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel as TModel,\n language: nextLanguage,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n const [model, language] = parseSTTModelString(modelString);\n return new STT({ model, language });\n }\n\n protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n stream(options?: {\n language?: STTLanguages | string;\n connOptions?: APIConnectOptions;\n }): SpeechStream<TModel> {\n const { language, connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } =\n options || {};\n const streamOpts = {\n ...this.opts,\n language: language ?? this.opts.language,\n } as InferenceSTTOptions<TModel>;\n\n const stream = new SpeechStream(this, streamOpts, connOptions);\n this.streams.add(stream);\n\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n const params = {\n settings: {\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n },\n } as Record<string, unknown>;\n\n if (this.opts.model && this.opts.model !== 'auto') {\n params.model = this.opts.model;\n }\n\n if (this.opts.language) {\n (params.settings as Record<string, unknown>).language = this.opts.language;\n }\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/stt`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const socket = await connectWs(url, headers, timeout);\n const msg = { ...params, type: 'session.create' };\n socket.send(JSON.stringify(msg));\n\n return socket;\n }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n private opts: InferenceSTTOptions<TModel>;\n private requestId = shortuuid('stt_request_');\n private speaking = false;\n private speechDuration = 0;\n private reconnectEvent = new Event();\n private stt: STT<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(\n sttImpl: STT<TModel>,\n opts: InferenceSTTOptions<TModel>,\n connOptions: APIConnectOptions,\n ) {\n super(sttImpl, opts.sampleRate, connOptions);\n this.opts = opts;\n this.stt = sttImpl;\n this.connOptions = connOptions;\n }\n\n get label(): string {\n return 'inference.SpeechStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n this.reconnectEvent.set();\n }\n\n protected async run(): Promise<void> {\n while (true) {\n // Create fresh resources for each connection attempt\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n\n const eventChannel = createStreamChannel<SttServerEvent>();\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const createWsListener = async (ws: WebSocket, signal: AbortSignal) => {\n return new Promise<void>((resolve, reject) => {\n const onAbort = () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n };\n\n signal.addEventListener('abort', onAbort, { once: true });\n\n ws.on('message', (data) => {\n const json = JSON.parse(data.toString()) as SttServerEvent;\n eventChannel.write(json);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', (code: number) => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'LiveKit STT connection closed unexpectedly',\n options: { statusCode: code },\n }),\n );\n });\n });\n };\n\n const send = async (socket: WebSocket, signal: AbortSignal) => {\n const audioStream = new AudioByteStream(\n this.opts.sampleRate,\n 1,\n Math.floor(this.opts.sampleRate / 20), // 50ms\n );\n\n // Create abort promise once to avoid memory leak\n const abortPromise = new Promise<never>((_, reject) => {\n if (signal.aborted) {\n return reject(new Error('Send aborted'));\n }\n const onAbort = () => reject(new Error('Send aborted'));\n signal.addEventListener('abort', onAbort, { once: true });\n });\n\n // Manual iteration to support cancellation\n const iterator = this.input[Symbol.asyncIterator]();\n try {\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result.done) break;\n const ev = result.value;\n\n let frames: AudioFrame[];\n if (ev === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else {\n const frame = ev as AudioFrame;\n frames = audioStream.write(new Int16Array(frame.data).buffer);\n }\n\n for (const frame of frames) {\n this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n const base64 = Buffer.from(frame.data.buffer).toString('base64');\n const msg = { type: 'input_audio', audio: base64 };\n socket.send(JSON.stringify(msg));\n }\n }\n\n closing = true;\n socket.send(JSON.stringify({ type: 'session.finalize' }));\n } catch (e) {\n if ((e as Error).message === 'Send aborted') {\n // Expected abort, don't log\n return;\n }\n throw e;\n }\n };\n\n const recv = async (signal: AbortSignal) => {\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n // Parse and validate with Zod schema\n const parseResult = await sttServerEventSchema.safeParseAsync(result.value);\n if (!parseResult.success) {\n this.#logger.warn(\n { error: parseResult.error, rawData: result.value },\n 'Failed to parse STT server event',\n );\n continue;\n }\n\n const event: SttServerEvent = parseResult.data;\n\n switch (event.type) {\n case 'session.created':\n case 'session.finalized':\n break;\n case 'session.closed':\n finalReceived = true;\n resourceCleanup();\n break;\n case 'interim_transcript':\n this.processTranscript(event, false);\n break;\n case 'final_transcript':\n this.processTranscript(event, true);\n break;\n case 'error':\n this.#logger.error({ error: event }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.stt.connectWs(this.connOptions.timeoutMs);\n\n const controller = this.abortController; // Use base class abortController for proper cancellation\n const sendTask = Task.from(({ signal }) => send(ws!, signal), controller);\n const wsListenerTask = Task.from(({ signal }) => createWsListener(ws!, signal), controller);\n const recvTask = Task.from(({ signal }) => recv(signal), controller);\n const waitReconnectTask = Task.from(\n ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),\n controller,\n );\n\n try {\n await Promise.race([\n Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),\n waitReconnectTask.result,\n ]);\n\n // If reconnect didn't trigger, tasks finished - exit loop\n if (!waitReconnectTask.done) break;\n\n // Reconnect triggered - clear event and continue loop\n this.reconnectEvent.clear();\n } finally {\n // Cancel all tasks to ensure cleanup\n await cancelAndWait(\n [sendTask, wsListenerTask, recvTask, waitReconnectTask],\n DEFAULT_CANCEL_TIMEOUT,\n );\n resourceCleanup();\n }\n } finally {\n // Ensure cleanup even if connectWs throws\n resourceCleanup();\n }\n }\n }\n\n private processTranscript(data: SttTranscriptEvent, isFinal: boolean) {\n // Check if queue is closed to avoid race condition during disconnect\n if (this.queue.closed) return;\n\n const requestId = data.session_id || this.requestId;\n const text = data.transcript;\n const language = data.language || this.opts.language || 'en';\n\n if (!text && !isFinal) return;\n\n try {\n // We'll have a more accurate way of detecting when speech started when we have VAD\n if (!this.speaking) {\n this.speaking = true;\n this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n }\n\n const speechData: SpeechData = {\n language,\n startTime: this.startTimeOffset + data.start,\n endTime: this.startTimeOffset + data.start + data.duration,\n confidence: data.confidence,\n text,\n words: data.words.map(\n (word): TimedString =>\n createTimedString({\n text: word.word,\n startTime: word.start + this.startTimeOffset,\n endTime: word.end + this.startTimeOffset,\n startTimeOffset: this.startTimeOffset,\n confidence: word.confidence,\n }),\n ),\n };\n\n if (isFinal) {\n if (this.speechDuration > 0) {\n this.queue.put({\n type: SpeechEventType.RECOGNITION_USAGE,\n requestId,\n recognitionUsage: { audioDuration: this.speechDuration },\n });\n this.speechDuration = 0;\n }\n\n this.queue.put({\n type: SpeechEventType.FINAL_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n\n if (this.speaking) {\n this.speaking = false;\n this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n }\n } else {\n this.queue.put({\n type: SpeechEventType.INTERIM_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n }\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n // Expected behavior on disconnect, log as warning\n this.#logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err: e }, 'Error putting transcript to queue');\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgC;AAEhC,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,iBAMO;AACP,mBAAoE;AACpE,mBAAsF;AACtF,gBAAoD;AACpD,wBAIO;AACP,IAAAA,gBAA6D;AAmGtD,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAGO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,IAAI,IAAI,oBAAoB,KAAK;AACxC,aAAO,EAAE,OAAO,KAAK;AAAA,IACvB;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAkBxB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,cAAU,gBAAI;AAAA,EAEd,YAAY,MAWT;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,mBAAmB,OAAO,CAAC;AAE1E,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,eAAe;AACnB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,oBAAoB,UAAU,MAAM,MAAM,CAAC;AACjD,YAAI,gBAAgB,iBAAiB,mBAAmB;AACtD,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,UAAU,cAAc,OAAO,UAAU;AAAA,UAC7C;AAAA,QACF,OAAO;AACL,yBAAe;AAAA,QACjB;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AACA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,UAAU;AAAA,MACV;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,WAAW;AACzD,WAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,EACpC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,KAAK,KAAK,eAAe,yCAA4B,IACnF,WAAW,CAAC;AACd,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AAjSvD;AAkSI,UAAM,SAAS;AAAA,MACb,UAAU;AAAA,QACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,QACxC,UAAU,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,QAAI,KAAK,KAAK,UAAU;AACtB,MAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,IACpE;AAEA,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,UAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,WAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,WAAAC,aAAiB;AAAA,EACnE;AAAA,EACA,gBAAY,wBAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,mBAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,YAAM,mBAAe,2CAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOC,KAAe,WAAwB;AACrE,eAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,gBAAM,UAAU,MAAM;AACpB,4BAAgB;AAChB,mBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,UAClD;AAEA,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAExD,UAAAA,IAAG,GAAG,WAAW,CAAC,SAAS;AACzB,kBAAM,OAAO,KAAK,MAAM,KAAK,SAAS,CAAC;AACvC,yBAAa,MAAM,IAAI;AAAA,UACzB,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,iBAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,4BAAgB;AAChB,mBAAO,CAAC;AAAA,UACV,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,SAAiB;AAC/B,4BAAgB;AAEhB,gBAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,gBAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,cACE,IAAI,iCAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS,EAAE,YAAY,KAAK;AAAA,cAC9B,CAAC;AAAA,YACH;AAAA,UACF,CAAC;AAAA,QACH,CAAC;AAAA,MACH;AAEA,YAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,cAAM,cAAc,IAAI;AAAA,UACtB,KAAK,KAAK;AAAA,UACV;AAAA,UACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,QACtC;AAGA,cAAM,eAAe,IAAI,QAAe,CAAC,GAAG,WAAW;AACrD,cAAI,OAAO,SAAS;AAClB,mBAAO,OAAO,IAAI,MAAM,cAAc,CAAC;AAAA,UACzC;AACA,gBAAM,UAAU,MAAM,OAAO,IAAI,MAAM,cAAc,CAAC;AACtD,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,QAC1D,CAAC;AAGD,cAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,YAAI;AACF,iBAAO,MAAM;AACX,kBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,gBAAI,OAAO,KAAM;AACjB,kBAAM,KAAK,OAAO;AAElB,gBAAI;AACJ,gBAAI,OAAO,aAAa,gBAAgB;AACtC,uBAAS,YAAY,MAAM;AAAA,YAC7B,OAAO;AACL,oBAAM,QAAQ;AACd,uBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,YAC9D;AAEA,uBAAW,SAAS,QAAQ;AAC1B,mBAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,oBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,oBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,qBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,YACjC;AAAA,UACF;AAEA,oBAAU;AACV,iBAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,QAC1D,SAAS,GAAG;AACV,cAAK,EAAY,YAAY,gBAAgB;AAE3C;AAAA,UACF;AACA,gBAAM;AAAA,QACR;AAAA,MACF;AAEA,YAAM,OAAO,OAAO,WAAwB;AAC1C,cAAM,oBAAoB,aAAa,OAAO;AAC9C,cAAM,SAAS,kBAAkB,UAAU;AAE3C,YAAI;AACF,iBAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,kBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,gBAAI,OAAO,QAAS;AACpB,gBAAI,OAAO,KAAM;AAGjB,kBAAM,cAAc,MAAM,uCAAqB,eAAe,OAAO,KAAK;AAC1E,gBAAI,CAAC,YAAY,SAAS;AACxB,mBAAK,QAAQ;AAAA,gBACX,EAAE,OAAO,YAAY,OAAO,SAAS,OAAO,MAAM;AAAA,gBAClD;AAAA,cACF;AACA;AAAA,YACF;AAEA,kBAAM,QAAwB,YAAY;AAE1C,oBAAQ,MAAM,MAAM;AAAA,cAClB,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,KAAK;AACnC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,IAAI;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,MAAM,GAAG,iCAAiC;AACtE,gCAAgB;AAChB,sBAAM,IAAI,2BAAS,+BAA+B,KAAK,UAAU,KAAK,CAAC,EAAE;AAAA,YAC7E;AAAA,UACF;AAAA,QACF,UAAE;AACA,iBAAO,YAAY;AACnB,cAAI;AACF,kBAAM,kBAAkB,OAAO;AAAA,UACjC,SAAS,GAAG;AACV,iBAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,UACxF;AAAA,QACF;AAAA,MACF;AAEA,UAAI;AACF,aAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,cAAM,aAAa,KAAK;AACxB,cAAM,WAAW,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,kBAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,OAAG,2BAAa,MAAM,CAAC,CAAC;AAAA,UAC/E;AAAA,QACF;AAEA,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,CAAC,SAAS,QAAQ,eAAe,QAAQ,SAAS,MAAM,CAAC;AAAA,YACrE,kBAAkB;AAAA,UACpB,CAAC;AAGD,cAAI,CAAC,kBAAkB,KAAM;AAG7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AAEA,oBAAM;AAAA,YACJ,CAAC,UAAU,gBAAgB,UAAU,iBAAiB;AAAA,YACtD;AAAA,UACF;AACA,0BAAgB;AAAA,QAClB;AAAA,MACF,UAAE;AAEA,wBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA0B,SAAkB;AAEpE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK;AAClB,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,gBAAgB,CAAC;AAAA,MAC1D;AAEA,YAAM,aAAyB;AAAA,QAC7B;AAAA,QACA,WAAW,KAAK,kBAAkB,KAAK;AAAA,QACvC,SAAS,KAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,QAClD,YAAY,KAAK;AAAA,QACjB;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,UAChB,CAAC,aACC,6BAAkB;AAAA,YAChB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB,CAAC;AAAA,QACL;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,2BAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAAA,MACH;AAAA,IACF,SAAS,GAAG;AACV,UAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAE/D,aAAK,QAAQ;AAAA,UACX,EAAE,KAAK,EAAE;AAAA,UACT;AAAA,QACF;AAAA,MACF,OAAO;AACL,aAAK,QAAQ,MAAM,EAAE,KAAK,EAAE,GAAG,mCAAmC;AAAA,MACpE;AAAA,IACF;AAAA,EACF;AACF;","names":["import_utils","BaseSTT","BaseSpeechStream","ws"]}
1
+ {"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\n STT as BaseSTT,\n SpeechStream as BaseSpeechStream,\n type SpeechData,\n type SpeechEvent,\n SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type TimedString, createTimedString } from '../voice/io.js';\nimport {\n type SttServerEvent,\n type SttTranscriptEvent,\n sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken, getDefaultInferenceUrl } from './utils.js';\n\nexport type DeepgramModels =\n | 'deepgram/flux-general'\n | 'deepgram/nova-3'\n | 'deepgram/nova-3-medical'\n | 'deepgram/nova-2'\n | 'deepgram/nova-2-medical'\n | 'deepgram/nova-2-conversationalai'\n | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels =\n | 'assemblyai/universal-streaming'\n | 'assemblyai/universal-streaming-multilingual';\n\nexport type ElevenlabsSTTModels = 'elevenlabs/scribe_v2_realtime';\n\nexport interface CartesiaOptions {\n /** Minimum volume threshold. Default: not specified. */\n min_volume?: number;\n /** Maximum silence duration in seconds. Default: not specified. */\n max_silence_duration_secs?: number;\n}\n\nexport interface DeepgramOptions {\n /** Enable filler words. Default: true. */\n filler_words?: boolean;\n /** Enable interim results. Default: true. */\n interim_results?: boolean;\n /** Endpointing timeout in milliseconds. Default: 25. */\n endpointing?: number;\n /** Enable punctuation. Default: false. */\n punctuate?: boolean;\n /** Enable smart formatting. */\n smart_format?: boolean;\n /** Keywords with boost values. */\n keywords?: Array<[string, number]>;\n /** Key terms for recognition. */\n keyterms?: string[];\n /** Enable profanity filter. */\n profanity_filter?: boolean;\n /** Convert spoken numbers to numerals. */\n numerals?: boolean;\n /** Opt out of model improvement program. */\n mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n /** Enable turn formatting. Default: false. */\n format_turns?: boolean;\n /** End of turn confidence threshold. Default: 0.01. */\n end_of_turn_confidence_threshold?: number;\n /** Minimum silence duration in milliseconds when confident about end of turn. Default: 0. */\n min_end_of_turn_silence_when_confident?: number;\n /** Maximum turn silence in milliseconds. Default: not specified. */\n max_turn_silence?: number;\n /** Key terms prompt for recognition. Default: not specified. */\n keyterms_prompt?: string[];\n}\n\nexport type STTLanguages =\n | 'multi'\n | 'en'\n | 'de'\n | 'es'\n | 'fr'\n | 'ja'\n | 'pt'\n | 'zh'\n | 'hi'\n | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels | ElevenlabsSTTModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n ? DeepgramOptions\n : TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends AssemblyaiModels\n ? AssemblyAIOptions\n : Record<string, unknown>;\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface STTFallbackModel {\n /** Model name (e.g. \"deepgram/nova-3\", \"assemblyai/universal-streaming\", \"cartesia/ink-whisper\"). */\n model: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type STTFallbackModelType = STTFallbackModel | string;\n\n/** Parse a model string into [model, language]. Language is undefined if not specified. */\nexport function parseSTTModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** Normalize a single or list of FallbackModelType into STTFallbackModel[]. */\nexport function normalizeSTTFallback(\n fallback: STTFallbackModelType | STTFallbackModelType[],\n): STTFallbackModel[] {\n const makeFallback = (model: STTFallbackModelType): STTFallbackModel => {\n if (typeof model === 'string') {\n const [name] = parseSTTModelString(model);\n return { model: name };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n model?: TModel;\n language?: STTLanguages;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\n fallback?: STTFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n private opts: InferenceSTTOptions<TModel>;\n private streams: Set<SpeechStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts?: {\n model?: ModelWithLanguage;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\n fallback?: STTFallbackModelType | STTFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n super({ streaming: true, interimResults: true, alignedTranscript: 'word' });\n\n const {\n model,\n language,\n baseURL,\n encoding = DEFAULT_ENCODING,\n sampleRate = DEFAULT_SAMPLE_RATE,\n apiKey,\n apiSecret,\n modelOptions = {} as STTOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || getDefaultInferenceUrl();\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // Parse language from model string if provided: \"provider/model:language\"\n let nextModel = model;\n let nextLanguage = language;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const languageFromModel = nextModel.slice(idx + 1) as STTLanguages;\n if (nextLanguage && nextLanguage !== languageFromModel) {\n this.#logger.warn(\n '`language` is provided via both argument and model, using the one from the argument',\n { language: nextLanguage, model: nextModel },\n );\n } else {\n nextLanguage = languageFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel as TModel,\n language: nextLanguage,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n get model(): string {\n return this.opts.model ?? 'auto';\n }\n\n get provider(): string {\n return 'livekit';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n const [model, language] = parseSTTModelString(modelString);\n return new STT({ model, language });\n }\n\n protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n stream(options?: {\n language?: STTLanguages | string;\n connOptions?: APIConnectOptions;\n }): SpeechStream<TModel> {\n const { language, connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } =\n options || {};\n const streamOpts = {\n ...this.opts,\n language: language ?? this.opts.language,\n } as InferenceSTTOptions<TModel>;\n\n const stream = new SpeechStream(this, streamOpts, connOptions);\n this.streams.add(stream);\n\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n const params = {\n settings: {\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n },\n } as Record<string, unknown>;\n\n if (this.opts.model && this.opts.model !== 'auto') {\n params.model = this.opts.model;\n }\n\n if (this.opts.language) {\n (params.settings as Record<string, unknown>).language = this.opts.language;\n }\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/stt`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const socket = await connectWs(url, headers, timeout);\n const msg = { ...params, type: 'session.create' };\n socket.send(JSON.stringify(msg));\n\n return socket;\n }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n private opts: InferenceSTTOptions<TModel>;\n private requestId = shortuuid('stt_request_');\n private speaking = false;\n private speechDuration = 0;\n private reconnectEvent = new Event();\n private stt: STT<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(\n sttImpl: STT<TModel>,\n opts: InferenceSTTOptions<TModel>,\n connOptions: APIConnectOptions,\n ) {\n super(sttImpl, opts.sampleRate, connOptions);\n this.opts = opts;\n this.stt = sttImpl;\n this.connOptions = connOptions;\n }\n\n get label(): string {\n return 'inference.SpeechStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n this.reconnectEvent.set();\n }\n\n protected async run(): Promise<void> {\n while (true) {\n // Create fresh resources for each connection attempt\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n\n const eventChannel = createStreamChannel<SttServerEvent>();\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const createWsListener = async (ws: WebSocket, signal: AbortSignal) => {\n return new Promise<void>((resolve, reject) => {\n const onAbort = () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n };\n\n signal.addEventListener('abort', onAbort, { once: true });\n\n ws.on('message', (data) => {\n const json = JSON.parse(data.toString()) as SttServerEvent;\n eventChannel.write(json);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', (code: number) => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'LiveKit STT connection closed unexpectedly',\n options: { statusCode: code },\n }),\n );\n });\n });\n };\n\n const send = async (socket: WebSocket, signal: AbortSignal) => {\n const audioStream = new AudioByteStream(\n this.opts.sampleRate,\n 1,\n Math.floor(this.opts.sampleRate / 20), // 50ms\n );\n\n // Create abort promise once to avoid memory leak\n const abortPromise = new Promise<never>((_, reject) => {\n if (signal.aborted) {\n return reject(new Error('Send aborted'));\n }\n const onAbort = () => reject(new Error('Send aborted'));\n signal.addEventListener('abort', onAbort, { once: true });\n });\n\n // Manual iteration to support cancellation\n const iterator = this.input[Symbol.asyncIterator]();\n try {\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result.done) break;\n const ev = result.value;\n\n let frames: AudioFrame[];\n if (ev === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else {\n const frame = ev as AudioFrame;\n frames = audioStream.write(new Int16Array(frame.data).buffer);\n }\n\n for (const frame of frames) {\n this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n const base64 = Buffer.from(frame.data.buffer).toString('base64');\n const msg = { type: 'input_audio', audio: base64 };\n socket.send(JSON.stringify(msg));\n }\n }\n\n closing = true;\n socket.send(JSON.stringify({ type: 'session.finalize' }));\n } catch (e) {\n if ((e as Error).message === 'Send aborted') {\n // Expected abort, don't log\n return;\n }\n throw e;\n }\n };\n\n const recv = async (signal: AbortSignal) => {\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n // Parse and validate with Zod schema\n const parseResult = await sttServerEventSchema.safeParseAsync(result.value);\n if (!parseResult.success) {\n this.#logger.warn(\n { error: parseResult.error, rawData: result.value },\n 'Failed to parse STT server event',\n );\n continue;\n }\n\n const event: SttServerEvent = parseResult.data;\n\n switch (event.type) {\n case 'session.created':\n case 'session.finalized':\n break;\n case 'session.closed':\n finalReceived = true;\n resourceCleanup();\n break;\n case 'interim_transcript':\n this.processTranscript(event, false);\n break;\n case 'final_transcript':\n this.processTranscript(event, true);\n break;\n case 'error':\n this.#logger.error({ error: event }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.stt.connectWs(this.connOptions.timeoutMs);\n\n const controller = this.abortController; // Use base class abortController for proper cancellation\n const sendTask = Task.from(({ signal }) => send(ws!, signal), controller);\n const wsListenerTask = Task.from(({ signal }) => createWsListener(ws!, signal), controller);\n const recvTask = Task.from(({ signal }) => recv(signal), controller);\n const waitReconnectTask = Task.from(\n ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),\n controller,\n );\n\n try {\n await Promise.race([\n Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),\n waitReconnectTask.result,\n ]);\n\n // If reconnect didn't trigger, tasks finished - exit loop\n if (!waitReconnectTask.done) break;\n\n // Reconnect triggered - clear event and continue loop\n this.reconnectEvent.clear();\n } finally {\n // Cancel all tasks to ensure cleanup\n await cancelAndWait(\n [sendTask, wsListenerTask, recvTask, waitReconnectTask],\n DEFAULT_CANCEL_TIMEOUT,\n );\n resourceCleanup();\n }\n } finally {\n // Ensure cleanup even if connectWs throws\n resourceCleanup();\n }\n }\n }\n\n private processTranscript(data: SttTranscriptEvent, isFinal: boolean) {\n // Check if queue is closed to avoid race condition during disconnect\n if (this.queue.closed) return;\n\n const requestId = data.session_id || this.requestId;\n const text = data.transcript;\n const language = data.language || this.opts.language || 'en';\n\n if (!text && !isFinal) return;\n\n try {\n // We'll have a more accurate way of detecting when speech started when we have VAD\n if (!this.speaking) {\n this.speaking = true;\n this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n }\n\n const speechData: SpeechData = {\n language,\n startTime: this.startTimeOffset + data.start,\n endTime: this.startTimeOffset + data.start + data.duration,\n confidence: data.confidence,\n text,\n words: data.words.map(\n (word): TimedString =>\n createTimedString({\n text: word.word,\n startTime: word.start + this.startTimeOffset,\n endTime: word.end + this.startTimeOffset,\n startTimeOffset: this.startTimeOffset,\n confidence: word.confidence,\n }),\n ),\n };\n\n if (isFinal) {\n if (this.speechDuration > 0) {\n this.queue.put({\n type: SpeechEventType.RECOGNITION_USAGE,\n requestId,\n recognitionUsage: { audioDuration: this.speechDuration },\n });\n this.speechDuration = 0;\n }\n\n this.queue.put({\n type: SpeechEventType.FINAL_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n\n if (this.speaking) {\n this.speaking = false;\n this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n }\n } else {\n this.queue.put({\n type: SpeechEventType.INTERIM_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n }\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n // Expected behavior on disconnect, log as warning\n this.#logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err: e }, 'Error putting transcript to queue');\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgC;AAEhC,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,iBAMO;AACP,mBAAoE;AACpE,mBAAsF;AACtF,gBAAoD;AACpD,wBAIO;AACP,IAAAA,gBAAqF;AAmG9E,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAGO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,IAAI,IAAI,oBAAoB,KAAK;AACxC,aAAO,EAAE,OAAO,KAAK;AAAA,IACvB;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,yBAAyB;AAkBxB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,cAAU,gBAAI;AAAA,EAEd,YAAY,MAWT;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,mBAAmB,OAAO,CAAC;AAE1E,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,eAAW,sCAAuB;AACpD,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,eAAe;AACnB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,oBAAoB,UAAU,MAAM,MAAM,CAAC;AACjD,YAAI,gBAAgB,iBAAiB,mBAAmB;AACtD,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,UAAU,cAAc,OAAO,UAAU;AAAA,UAC7C;AAAA,QACF,OAAO;AACL,yBAAe;AAAA,QACjB;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AACA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,UAAU;AAAA,MACV;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK,SAAS;AAAA,EAC5B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,WAAW;AACzD,WAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,EACpC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,KAAK,KAAK,eAAe,yCAA4B,IACnF,WAAW,CAAC;AACd,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AAxSvD;AAySI,UAAM,SAAS;AAAA,MACb,UAAU;AAAA,QACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,QACxC,UAAU,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,QAAI,KAAK,KAAK,UAAU;AACtB,MAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,IACpE;AAEA,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,UAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,WAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,WAAAC,aAAiB;AAAA,EACnE;AAAA,EACA,gBAAY,wBAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,mBAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,YAAM,mBAAe,2CAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOC,KAAe,WAAwB;AACrE,eAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,gBAAM,UAAU,MAAM;AACpB,4BAAgB;AAChB,mBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,UAClD;AAEA,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAExD,UAAAA,IAAG,GAAG,WAAW,CAAC,SAAS;AACzB,kBAAM,OAAO,KAAK,MAAM,KAAK,SAAS,CAAC;AACvC,yBAAa,MAAM,IAAI;AAAA,UACzB,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,iBAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,4BAAgB;AAChB,mBAAO,CAAC;AAAA,UACV,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,SAAiB;AAC/B,4BAAgB;AAEhB,gBAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,gBAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,cACE,IAAI,iCAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS,EAAE,YAAY,KAAK;AAAA,cAC9B,CAAC;AAAA,YACH;AAAA,UACF,CAAC;AAAA,QACH,CAAC;AAAA,MACH;AAEA,YAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,cAAM,cAAc,IAAI;AAAA,UACtB,KAAK,KAAK;AAAA,UACV;AAAA,UACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,QACtC;AAGA,cAAM,eAAe,IAAI,QAAe,CAAC,GAAG,WAAW;AACrD,cAAI,OAAO,SAAS;AAClB,mBAAO,OAAO,IAAI,MAAM,cAAc,CAAC;AAAA,UACzC;AACA,gBAAM,UAAU,MAAM,OAAO,IAAI,MAAM,cAAc,CAAC;AACtD,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,QAC1D,CAAC;AAGD,cAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,YAAI;AACF,iBAAO,MAAM;AACX,kBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,gBAAI,OAAO,KAAM;AACjB,kBAAM,KAAK,OAAO;AAElB,gBAAI;AACJ,gBAAI,OAAO,aAAa,gBAAgB;AACtC,uBAAS,YAAY,MAAM;AAAA,YAC7B,OAAO;AACL,oBAAM,QAAQ;AACd,uBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,YAC9D;AAEA,uBAAW,SAAS,QAAQ;AAC1B,mBAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,oBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,oBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,qBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,YACjC;AAAA,UACF;AAEA,oBAAU;AACV,iBAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,QAC1D,SAAS,GAAG;AACV,cAAK,EAAY,YAAY,gBAAgB;AAE3C;AAAA,UACF;AACA,gBAAM;AAAA,QACR;AAAA,MACF;AAEA,YAAM,OAAO,OAAO,WAAwB;AAC1C,cAAM,oBAAoB,aAAa,OAAO;AAC9C,cAAM,SAAS,kBAAkB,UAAU;AAE3C,YAAI;AACF,iBAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,kBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,gBAAI,OAAO,QAAS;AACpB,gBAAI,OAAO,KAAM;AAGjB,kBAAM,cAAc,MAAM,uCAAqB,eAAe,OAAO,KAAK;AAC1E,gBAAI,CAAC,YAAY,SAAS;AACxB,mBAAK,QAAQ;AAAA,gBACX,EAAE,OAAO,YAAY,OAAO,SAAS,OAAO,MAAM;AAAA,gBAClD;AAAA,cACF;AACA;AAAA,YACF;AAEA,kBAAM,QAAwB,YAAY;AAE1C,oBAAQ,MAAM,MAAM;AAAA,cAClB,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,KAAK;AACnC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,IAAI;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,MAAM,GAAG,iCAAiC;AACtE,gCAAgB;AAChB,sBAAM,IAAI,2BAAS,+BAA+B,KAAK,UAAU,KAAK,CAAC,EAAE;AAAA,YAC7E;AAAA,UACF;AAAA,QACF,UAAE;AACA,iBAAO,YAAY;AACnB,cAAI;AACF,kBAAM,kBAAkB,OAAO;AAAA,UACjC,SAAS,GAAG;AACV,iBAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,UACxF;AAAA,QACF;AAAA,MACF;AAEA,UAAI;AACF,aAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,cAAM,aAAa,KAAK;AACxB,cAAM,WAAW,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,kBAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,OAAG,2BAAa,MAAM,CAAC,CAAC;AAAA,UAC/E;AAAA,QACF;AAEA,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,CAAC,SAAS,QAAQ,eAAe,QAAQ,SAAS,MAAM,CAAC;AAAA,YACrE,kBAAkB;AAAA,UACpB,CAAC;AAGD,cAAI,CAAC,kBAAkB,KAAM;AAG7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AAEA,oBAAM;AAAA,YACJ,CAAC,UAAU,gBAAgB,UAAU,iBAAiB;AAAA,YACtD;AAAA,UACF;AACA,0BAAgB;AAAA,QAClB;AAAA,MACF,UAAE;AAEA,wBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA0B,SAAkB;AAEpE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK;AAClB,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,gBAAgB,CAAC;AAAA,MAC1D;AAEA,YAAM,aAAyB;AAAA,QAC7B;AAAA,QACA,WAAW,KAAK,kBAAkB,KAAK;AAAA,QACvC,SAAS,KAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,QAClD,YAAY,KAAK;AAAA,QACjB;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,UAChB,CAAC,aACC,6BAAkB;AAAA,YAChB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB,CAAC;AAAA,QACL;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,2BAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAAA,MACH;AAAA,IACF,SAAS,GAAG;AACV,UAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAE/D,aAAK,QAAQ;AAAA,UACX,EAAE,KAAK,EAAE;AAAA,UACT;AAAA,QACF;AAAA,MACF,OAAO;AACL,aAAK,QAAQ,MAAM,EAAE,KAAK,EAAE,GAAG,mCAAmC;AAAA,MACpE;AAAA,IACF;AAAA,EACF;AACF;","names":["import_utils","BaseSTT","BaseSpeechStream","ws"]}
@@ -97,6 +97,8 @@ export declare class STT<TModel extends STTModels> extends BaseSTT {
97
97
  connOptions?: APIConnectOptions;
98
98
  });
99
99
  get label(): string;
100
+ get model(): string;
101
+ get provider(): string;
100
102
  static fromModelString(modelString: string): STT<AnyString>;
101
103
  protected _recognize(_: AudioBuffer): Promise<SpeechEvent>;
102
104
  updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void;
@@ -97,6 +97,8 @@ export declare class STT<TModel extends STTModels> extends BaseSTT {
97
97
  connOptions?: APIConnectOptions;
98
98
  });
99
99
  get label(): string;
100
+ get model(): string;
101
+ get provider(): string;
100
102
  static fromModelString(modelString: string): STT<AnyString>;
101
103
  protected _recognize(_: AudioBuffer): Promise<SpeechEvent>;
102
104
  updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void;
@@ -1 +1 @@
1
- {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/inference/stt.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAKpC,OAAO,EACL,GAAG,IAAI,OAAO,EACd,YAAY,IAAI,gBAAgB,EAEhC,KAAK,WAAW,EAEjB,MAAM,iBAAiB,CAAC;AACzB,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,KAAK,WAAW,EAAuD,MAAM,aAAa,CAAC;AAOpG,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,uBAAuB,GACvB,iBAAiB,GACjB,yBAAyB,GACzB,iBAAiB,GACjB,yBAAyB,GACzB,kCAAkC,GAClC,2BAA2B,CAAC;AAEhC,MAAM,MAAM,cAAc,GAAG,sBAAsB,CAAC;AAEpD,MAAM,MAAM,gBAAgB,GACxB,gCAAgC,GAChC,6CAA6C,CAAC;AAElD,MAAM,MAAM,mBAAmB,GAAG,+BAA+B,CAAC;AAElE,MAAM,WAAW,eAAe;IAC9B,wDAAwD;IACxD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,mEAAmE;IACnE,yBAAyB,CAAC,EAAE,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,eAAe;IAC9B,0CAA0C;IAC1C,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,6CAA6C;IAC7C,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,wDAAwD;IACxD,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,0CAA0C;IAC1C,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,+BAA+B;IAC/B,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,kCAAkC;IAClC,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC;IACnC,iCAAiC;IACjC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;IACpB,+BAA+B;IAC/B,gBAAgB,CAAC,EAAE,OAAO,CAAC;IAC3B,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,4CAA4C;IAC5C,WAAW,CAAC,EAAE,OAAO,CAAC;CACvB;AAED,MAAM,WAAW,iBAAiB;IAChC,8CAA8C;IAC9C,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,uDAAuD;IACvD,gCAAgC,CAAC,EAAE,MAAM,CAAC;IAC1C,6FAA6F;IAC7F,sCAAsC,CAAC,EAAE,MAAM,CAAC;IAChD,oEAAoE;IACpE,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,gEAAgE;IAChE,eAAe,CAAC,EAAE,MAAM,EAAE,CAAC;CAC5B;AAED,MAAM,MAAM,YAAY,GACpB,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,SAAS,CAAC;AAEd,KAAK,UAAU,GAAG,cAAc,GAAG,cAAc,GAAG,gBAAgB,GAAG,mBAAmB,CAAC;AAE3F,MAAM,MAAM,SAAS,GAAG,UAAU,GAAG,MAAM,GAAG,SAAS,CAAC;AAExD,MAAM,MAAM,iBAAiB,GAAG,GAAG,UAAU,IAAI,YAAY,EAAE,GAAG,SAAS,CAAC;AAE5E,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,cAAc,GAC3B,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAEhC,2GAA2G;AAC3G,MAAM,WAAW,gBAAgB;IAC/B,qGAAqG;IACrG,KAAK,EAAE,MAAM,CAAC;IACd,yCAAyC;IACzC,WAAW,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACvC;AAED,MAAM,MAAM,oBAAoB,GAAG,gBAAgB,GAAG,MAAM,CAAC;AAE7D,2FAA2F;AAC3F,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,MAAM,GAAG,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC,CAM/E;AAED,+EAA+E;AAC/E,wBAAgB,oBAAoB,CAClC,QAAQ,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,GACtD,gBAAgB,EAAE,CAapB;AAED,MAAM,MAAM,WAAW,GAAG,WAAW,CAAC;AAOtC,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,YAAY,CAAC;IACxB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;IACjC,QAAQ,CAAC,EAAE,gBAAgB,EAAE,CAAC;IAC9B,WAAW,CAAC,EAAE,iBAAiB,CAAC;CACjC;AAED;;GAEG;AACH,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAAwC;gBAI3C,IAAI,CAAC,EAAE;QACjB,KAAK,CAAC,EAAE,iBAAiB,CAAC;QAC1B,QAAQ,CAAC,EAAE,YAAY,CAAC;QACxB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;QAClC,QAAQ,CAAC,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,CAAC;QACzD,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC;IA8DD,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;cAK3C,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAIhE,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;IAQ3F,MAAM,CAAC,OAAO,CAAC,EAAE;QACf,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;QACjC,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC,GAAG,YAAY,CAAC,MAAM,CAAC;IAclB,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;CAgDrD;AAED,qBAAa,YAAY,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,gBAAgB;;IAC1E,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,SAAS,CAA6B;IAC9C,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,cAAc,CAAK;IAC3B,OAAO,CAAC,cAAc,CAAe;IACrC,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAKrC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,EACpB,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EACjC,WAAW,EAAE,iBAAiB;IAQhC,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;cAK3E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAmMpC,OAAO,CAAC,iBAAiB;CA0E1B"}
1
+ {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/inference/stt.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAKpC,OAAO,EACL,GAAG,IAAI,OAAO,EACd,YAAY,IAAI,gBAAgB,EAEhC,KAAK,WAAW,EAEjB,MAAM,iBAAiB,CAAC;AACzB,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,KAAK,WAAW,EAAuD,MAAM,aAAa,CAAC;AAOpG,OAAO,EAAE,KAAK,SAAS,EAAwD,MAAM,YAAY,CAAC;AAElG,MAAM,MAAM,cAAc,GACtB,uBAAuB,GACvB,iBAAiB,GACjB,yBAAyB,GACzB,iBAAiB,GACjB,yBAAyB,GACzB,kCAAkC,GAClC,2BAA2B,CAAC;AAEhC,MAAM,MAAM,cAAc,GAAG,sBAAsB,CAAC;AAEpD,MAAM,MAAM,gBAAgB,GACxB,gCAAgC,GAChC,6CAA6C,CAAC;AAElD,MAAM,MAAM,mBAAmB,GAAG,+BAA+B,CAAC;AAElE,MAAM,WAAW,eAAe;IAC9B,wDAAwD;IACxD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,mEAAmE;IACnE,yBAAyB,CAAC,EAAE,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,eAAe;IAC9B,0CAA0C;IAC1C,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,6CAA6C;IAC7C,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,wDAAwD;IACxD,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,0CAA0C;IAC1C,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,+BAA+B;IAC/B,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,kCAAkC;IAClC,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC;IACnC,iCAAiC;IACjC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;IACpB,+BAA+B;IAC/B,gBAAgB,CAAC,EAAE,OAAO,CAAC;IAC3B,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,4CAA4C;IAC5C,WAAW,CAAC,EAAE,OAAO,CAAC;CACvB;AAED,MAAM,WAAW,iBAAiB;IAChC,8CAA8C;IAC9C,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,uDAAuD;IACvD,gCAAgC,CAAC,EAAE,MAAM,CAAC;IAC1C,6FAA6F;IAC7F,sCAAsC,CAAC,EAAE,MAAM,CAAC;IAChD,oEAAoE;IACpE,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,gEAAgE;IAChE,eAAe,CAAC,EAAE,MAAM,EAAE,CAAC;CAC5B;AAED,MAAM,MAAM,YAAY,GACpB,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,SAAS,CAAC;AAEd,KAAK,UAAU,GAAG,cAAc,GAAG,cAAc,GAAG,gBAAgB,GAAG,mBAAmB,CAAC;AAE3F,MAAM,MAAM,SAAS,GAAG,UAAU,GAAG,MAAM,GAAG,SAAS,CAAC;AAExD,MAAM,MAAM,iBAAiB,GAAG,GAAG,UAAU,IAAI,YAAY,EAAE,GAAG,SAAS,CAAC;AAE5E,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,cAAc,GAC3B,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAEhC,2GAA2G;AAC3G,MAAM,WAAW,gBAAgB;IAC/B,qGAAqG;IACrG,KAAK,EAAE,MAAM,CAAC;IACd,yCAAyC;IACzC,WAAW,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACvC;AAED,MAAM,MAAM,oBAAoB,GAAG,gBAAgB,GAAG,MAAM,CAAC;AAE7D,2FAA2F;AAC3F,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,MAAM,GAAG,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC,CAM/E;AAED,+EAA+E;AAC/E,wBAAgB,oBAAoB,CAClC,QAAQ,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,GACtD,gBAAgB,EAAE,CAapB;AAED,MAAM,MAAM,WAAW,GAAG,WAAW,CAAC;AAMtC,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,YAAY,CAAC;IACxB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;IACjC,QAAQ,CAAC,EAAE,gBAAgB,EAAE,CAAC;IAC9B,WAAW,CAAC,EAAE,iBAAiB,CAAC;CACjC;AAED;;GAEG;AACH,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAAwC;gBAI3C,IAAI,CAAC,EAAE;QACjB,KAAK,CAAC,EAAE,iBAAiB,CAAC;QAC1B,QAAQ,CAAC,EAAE,YAAY,CAAC;QACxB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;QAClC,QAAQ,CAAC,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,CAAC;QACzD,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC;IA8DD,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;cAK3C,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAIhE,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;IAQ3F,MAAM,CAAC,OAAO,CAAC,EAAE;QACf,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;QACjC,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC,GAAG,YAAY,CAAC,MAAM,CAAC;IAclB,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;CAgDrD;AAED,qBAAa,YAAY,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,gBAAgB;;IAC1E,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,SAAS,CAA6B;IAC9C,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,cAAc,CAAK;IAC3B,OAAO,CAAC,cAAc,CAAe;IACrC,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAKrC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,EACpB,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EACjC,WAAW,EAAE,iBAAiB;IAQhC,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;cAK3E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAmMpC,OAAO,CAAC,iBAAiB;CA0E1B"}
@@ -14,7 +14,7 @@ import { createTimedString } from "../voice/io.js";
14
14
  import {
15
15
  sttServerEventSchema
16
16
  } from "./api_protos.js";
17
- import { connectWs, createAccessToken } from "./utils.js";
17
+ import { connectWs, createAccessToken, getDefaultInferenceUrl } from "./utils.js";
18
18
  function parseSTTModelString(model) {
19
19
  const idx = model.lastIndexOf(":");
20
20
  if (idx !== -1) {
@@ -37,7 +37,6 @@ function normalizeSTTFallback(fallback) {
37
37
  }
38
38
  const DEFAULT_ENCODING = "pcm_s16le";
39
39
  const DEFAULT_SAMPLE_RATE = 16e3;
40
- const DEFAULT_BASE_URL = "wss://agent-gateway.livekit.cloud/v1";
41
40
  const DEFAULT_CANCEL_TIMEOUT = 5e3;
42
41
  class STT extends BaseSTT {
43
42
  opts;
@@ -57,7 +56,7 @@ class STT extends BaseSTT {
57
56
  fallback,
58
57
  connOptions
59
58
  } = opts || {};
60
- const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;
59
+ const lkBaseURL = baseURL || getDefaultInferenceUrl();
61
60
  const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;
62
61
  if (!lkApiKey) {
63
62
  throw new Error("apiKey is required: pass apiKey or set LIVEKIT_API_KEY");
@@ -100,6 +99,12 @@ class STT extends BaseSTT {
100
99
  get label() {
101
100
  return "inference.STT";
102
101
  }
102
+ get model() {
103
+ return this.opts.model ?? "auto";
104
+ }
105
+ get provider() {
106
+ return "livekit";
107
+ }
103
108
  static fromModelString(modelString) {
104
109
  const [model, language] = parseSTTModelString(modelString);
105
110
  return new STT({ model, language });
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\n STT as BaseSTT,\n SpeechStream as BaseSpeechStream,\n type SpeechData,\n type SpeechEvent,\n SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type TimedString, createTimedString } from '../voice/io.js';\nimport {\n type SttServerEvent,\n type SttTranscriptEvent,\n sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type DeepgramModels =\n | 'deepgram/flux-general'\n | 'deepgram/nova-3'\n | 'deepgram/nova-3-medical'\n | 'deepgram/nova-2'\n | 'deepgram/nova-2-medical'\n | 'deepgram/nova-2-conversationalai'\n | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels =\n | 'assemblyai/universal-streaming'\n | 'assemblyai/universal-streaming-multilingual';\n\nexport type ElevenlabsSTTModels = 'elevenlabs/scribe_v2_realtime';\n\nexport interface CartesiaOptions {\n /** Minimum volume threshold. Default: not specified. */\n min_volume?: number;\n /** Maximum silence duration in seconds. Default: not specified. */\n max_silence_duration_secs?: number;\n}\n\nexport interface DeepgramOptions {\n /** Enable filler words. Default: true. */\n filler_words?: boolean;\n /** Enable interim results. Default: true. */\n interim_results?: boolean;\n /** Endpointing timeout in milliseconds. Default: 25. */\n endpointing?: number;\n /** Enable punctuation. Default: false. */\n punctuate?: boolean;\n /** Enable smart formatting. */\n smart_format?: boolean;\n /** Keywords with boost values. */\n keywords?: Array<[string, number]>;\n /** Key terms for recognition. */\n keyterms?: string[];\n /** Enable profanity filter. */\n profanity_filter?: boolean;\n /** Convert spoken numbers to numerals. */\n numerals?: boolean;\n /** Opt out of model improvement program. */\n mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n /** Enable turn formatting. Default: false. */\n format_turns?: boolean;\n /** End of turn confidence threshold. Default: 0.01. */\n end_of_turn_confidence_threshold?: number;\n /** Minimum silence duration in milliseconds when confident about end of turn. Default: 0. */\n min_end_of_turn_silence_when_confident?: number;\n /** Maximum turn silence in milliseconds. Default: not specified. */\n max_turn_silence?: number;\n /** Key terms prompt for recognition. Default: not specified. */\n keyterms_prompt?: string[];\n}\n\nexport type STTLanguages =\n | 'multi'\n | 'en'\n | 'de'\n | 'es'\n | 'fr'\n | 'ja'\n | 'pt'\n | 'zh'\n | 'hi'\n | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels | ElevenlabsSTTModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n ? DeepgramOptions\n : TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends AssemblyaiModels\n ? AssemblyAIOptions\n : Record<string, unknown>;\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface STTFallbackModel {\n /** Model name (e.g. \"deepgram/nova-3\", \"assemblyai/universal-streaming\", \"cartesia/ink-whisper\"). */\n model: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type STTFallbackModelType = STTFallbackModel | string;\n\n/** Parse a model string into [model, language]. Language is undefined if not specified. */\nexport function parseSTTModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** Normalize a single or list of FallbackModelType into STTFallbackModel[]. */\nexport function normalizeSTTFallback(\n fallback: STTFallbackModelType | STTFallbackModelType[],\n): STTFallbackModel[] {\n const makeFallback = (model: STTFallbackModelType): STTFallbackModel => {\n if (typeof model === 'string') {\n const [name] = parseSTTModelString(model);\n return { model: name };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'wss://agent-gateway.livekit.cloud/v1';\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n model?: TModel;\n language?: STTLanguages;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\n fallback?: STTFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n private opts: InferenceSTTOptions<TModel>;\n private streams: Set<SpeechStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts?: {\n model?: ModelWithLanguage;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\n fallback?: STTFallbackModelType | STTFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n super({ streaming: true, interimResults: true, alignedTranscript: 'word' });\n\n const {\n model,\n language,\n baseURL,\n encoding = DEFAULT_ENCODING,\n sampleRate = DEFAULT_SAMPLE_RATE,\n apiKey,\n apiSecret,\n modelOptions = {} as STTOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // Parse language from model string if provided: \"provider/model:language\"\n let nextModel = model;\n let nextLanguage = language;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const languageFromModel = nextModel.slice(idx + 1) as STTLanguages;\n if (nextLanguage && nextLanguage !== languageFromModel) {\n this.#logger.warn(\n '`language` is provided via both argument and model, using the one from the argument',\n { language: nextLanguage, model: nextModel },\n );\n } else {\n nextLanguage = languageFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel as TModel,\n language: nextLanguage,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n const [model, language] = parseSTTModelString(modelString);\n return new STT({ model, language });\n }\n\n protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n stream(options?: {\n language?: STTLanguages | string;\n connOptions?: APIConnectOptions;\n }): SpeechStream<TModel> {\n const { language, connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } =\n options || {};\n const streamOpts = {\n ...this.opts,\n language: language ?? this.opts.language,\n } as InferenceSTTOptions<TModel>;\n\n const stream = new SpeechStream(this, streamOpts, connOptions);\n this.streams.add(stream);\n\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n const params = {\n settings: {\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n },\n } as Record<string, unknown>;\n\n if (this.opts.model && this.opts.model !== 'auto') {\n params.model = this.opts.model;\n }\n\n if (this.opts.language) {\n (params.settings as Record<string, unknown>).language = this.opts.language;\n }\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/stt`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const socket = await connectWs(url, headers, timeout);\n const msg = { ...params, type: 'session.create' };\n socket.send(JSON.stringify(msg));\n\n return socket;\n }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n private opts: InferenceSTTOptions<TModel>;\n private requestId = shortuuid('stt_request_');\n private speaking = false;\n private speechDuration = 0;\n private reconnectEvent = new Event();\n private stt: STT<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(\n sttImpl: STT<TModel>,\n opts: InferenceSTTOptions<TModel>,\n connOptions: APIConnectOptions,\n ) {\n super(sttImpl, opts.sampleRate, connOptions);\n this.opts = opts;\n this.stt = sttImpl;\n this.connOptions = connOptions;\n }\n\n get label(): string {\n return 'inference.SpeechStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n this.reconnectEvent.set();\n }\n\n protected async run(): Promise<void> {\n while (true) {\n // Create fresh resources for each connection attempt\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n\n const eventChannel = createStreamChannel<SttServerEvent>();\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const createWsListener = async (ws: WebSocket, signal: AbortSignal) => {\n return new Promise<void>((resolve, reject) => {\n const onAbort = () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n };\n\n signal.addEventListener('abort', onAbort, { once: true });\n\n ws.on('message', (data) => {\n const json = JSON.parse(data.toString()) as SttServerEvent;\n eventChannel.write(json);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', (code: number) => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'LiveKit STT connection closed unexpectedly',\n options: { statusCode: code },\n }),\n );\n });\n });\n };\n\n const send = async (socket: WebSocket, signal: AbortSignal) => {\n const audioStream = new AudioByteStream(\n this.opts.sampleRate,\n 1,\n Math.floor(this.opts.sampleRate / 20), // 50ms\n );\n\n // Create abort promise once to avoid memory leak\n const abortPromise = new Promise<never>((_, reject) => {\n if (signal.aborted) {\n return reject(new Error('Send aborted'));\n }\n const onAbort = () => reject(new Error('Send aborted'));\n signal.addEventListener('abort', onAbort, { once: true });\n });\n\n // Manual iteration to support cancellation\n const iterator = this.input[Symbol.asyncIterator]();\n try {\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result.done) break;\n const ev = result.value;\n\n let frames: AudioFrame[];\n if (ev === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else {\n const frame = ev as AudioFrame;\n frames = audioStream.write(new Int16Array(frame.data).buffer);\n }\n\n for (const frame of frames) {\n this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n const base64 = Buffer.from(frame.data.buffer).toString('base64');\n const msg = { type: 'input_audio', audio: base64 };\n socket.send(JSON.stringify(msg));\n }\n }\n\n closing = true;\n socket.send(JSON.stringify({ type: 'session.finalize' }));\n } catch (e) {\n if ((e as Error).message === 'Send aborted') {\n // Expected abort, don't log\n return;\n }\n throw e;\n }\n };\n\n const recv = async (signal: AbortSignal) => {\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n // Parse and validate with Zod schema\n const parseResult = await sttServerEventSchema.safeParseAsync(result.value);\n if (!parseResult.success) {\n this.#logger.warn(\n { error: parseResult.error, rawData: result.value },\n 'Failed to parse STT server event',\n );\n continue;\n }\n\n const event: SttServerEvent = parseResult.data;\n\n switch (event.type) {\n case 'session.created':\n case 'session.finalized':\n break;\n case 'session.closed':\n finalReceived = true;\n resourceCleanup();\n break;\n case 'interim_transcript':\n this.processTranscript(event, false);\n break;\n case 'final_transcript':\n this.processTranscript(event, true);\n break;\n case 'error':\n this.#logger.error({ error: event }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.stt.connectWs(this.connOptions.timeoutMs);\n\n const controller = this.abortController; // Use base class abortController for proper cancellation\n const sendTask = Task.from(({ signal }) => send(ws!, signal), controller);\n const wsListenerTask = Task.from(({ signal }) => createWsListener(ws!, signal), controller);\n const recvTask = Task.from(({ signal }) => recv(signal), controller);\n const waitReconnectTask = Task.from(\n ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),\n controller,\n );\n\n try {\n await Promise.race([\n Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),\n waitReconnectTask.result,\n ]);\n\n // If reconnect didn't trigger, tasks finished - exit loop\n if (!waitReconnectTask.done) break;\n\n // Reconnect triggered - clear event and continue loop\n this.reconnectEvent.clear();\n } finally {\n // Cancel all tasks to ensure cleanup\n await cancelAndWait(\n [sendTask, wsListenerTask, recvTask, waitReconnectTask],\n DEFAULT_CANCEL_TIMEOUT,\n );\n resourceCleanup();\n }\n } finally {\n // Ensure cleanup even if connectWs throws\n resourceCleanup();\n }\n }\n }\n\n private processTranscript(data: SttTranscriptEvent, isFinal: boolean) {\n // Check if queue is closed to avoid race condition during disconnect\n if (this.queue.closed) return;\n\n const requestId = data.session_id || this.requestId;\n const text = data.transcript;\n const language = data.language || this.opts.language || 'en';\n\n if (!text && !isFinal) return;\n\n try {\n // We'll have a more accurate way of detecting when speech started when we have VAD\n if (!this.speaking) {\n this.speaking = true;\n this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n }\n\n const speechData: SpeechData = {\n language,\n startTime: this.startTimeOffset + data.start,\n endTime: this.startTimeOffset + data.start + data.duration,\n confidence: data.confidence,\n text,\n words: data.words.map(\n (word): TimedString =>\n createTimedString({\n text: word.word,\n startTime: word.start + this.startTimeOffset,\n endTime: word.end + this.startTimeOffset,\n startTimeOffset: this.startTimeOffset,\n confidence: word.confidence,\n }),\n ),\n };\n\n if (isFinal) {\n if (this.speechDuration > 0) {\n this.queue.put({\n type: SpeechEventType.RECOGNITION_USAGE,\n requestId,\n recognitionUsage: { audioDuration: this.speechDuration },\n });\n this.speechDuration = 0;\n }\n\n this.queue.put({\n type: SpeechEventType.FINAL_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n\n if (this.speaking) {\n this.speaking = false;\n this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n }\n } else {\n this.queue.put({\n type: SpeechEventType.INTERIM_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n }\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n // Expected behavior on disconnect, log as warning\n this.#logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err: e }, 'Error putting transcript to queue');\n }\n }\n }\n}\n"],"mappings":"AAGA,eAAgC;AAEhC,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC;AAAA,EACE,OAAO;AAAA,EACP,gBAAgB;AAAA,EAGhB;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAA2B,OAAO,MAAM,eAAe,WAAW,oBAAoB;AACtF,SAA2B,yBAAyB;AACpD;AAAA,EAGE;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AAmGtD,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAGO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,IAAI,IAAI,oBAAoB,KAAK;AACxC,aAAO,EAAE,OAAO,KAAK;AAAA,IACvB;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAkBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,YAAY,MAWT;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,mBAAmB,OAAO,CAAC;AAE1E,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,eAAe;AACnB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,oBAAoB,UAAU,MAAM,MAAM,CAAC;AACjD,YAAI,gBAAgB,iBAAiB,mBAAmB;AACtD,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,UAAU,cAAc,OAAO,UAAU;AAAA,UAC7C;AAAA,QACF,OAAO;AACL,yBAAe;AAAA,QACjB;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AACA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,UAAU;AAAA,MACV;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,WAAW;AACzD,WAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,EACpC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,KAAK,KAAK,eAAe,4BAA4B,IACnF,WAAW,CAAC;AACd,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AAjSvD;AAkSI,UAAM,SAAS;AAAA,MACb,UAAU;AAAA,QACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,QACxC,UAAU,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,QAAI,KAAK,KAAK,UAAU;AACtB,MAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,IACpE;AAEA,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,UAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,WAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,iBAAiB;AAAA,EACnE;AAAA,EACA,YAAY,UAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,MAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,YAAM,eAAe,oBAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOA,KAAe,WAAwB;AACrE,eAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,gBAAM,UAAU,MAAM;AACpB,4BAAgB;AAChB,mBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,UAClD;AAEA,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAExD,UAAAA,IAAG,GAAG,WAAW,CAAC,SAAS;AACzB,kBAAM,OAAO,KAAK,MAAM,KAAK,SAAS,CAAC;AACvC,yBAAa,MAAM,IAAI;AAAA,UACzB,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,iBAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,4BAAgB;AAChB,mBAAO,CAAC;AAAA,UACV,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,SAAiB;AAC/B,4BAAgB;AAEhB,gBAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,gBAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,cACE,IAAI,eAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS,EAAE,YAAY,KAAK;AAAA,cAC9B,CAAC;AAAA,YACH;AAAA,UACF,CAAC;AAAA,QACH,CAAC;AAAA,MACH;AAEA,YAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,cAAM,cAAc,IAAI;AAAA,UACtB,KAAK,KAAK;AAAA,UACV;AAAA,UACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,QACtC;AAGA,cAAM,eAAe,IAAI,QAAe,CAAC,GAAG,WAAW;AACrD,cAAI,OAAO,SAAS;AAClB,mBAAO,OAAO,IAAI,MAAM,cAAc,CAAC;AAAA,UACzC;AACA,gBAAM,UAAU,MAAM,OAAO,IAAI,MAAM,cAAc,CAAC;AACtD,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,QAC1D,CAAC;AAGD,cAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,YAAI;AACF,iBAAO,MAAM;AACX,kBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,gBAAI,OAAO,KAAM;AACjB,kBAAM,KAAK,OAAO;AAElB,gBAAI;AACJ,gBAAI,OAAO,aAAa,gBAAgB;AACtC,uBAAS,YAAY,MAAM;AAAA,YAC7B,OAAO;AACL,oBAAM,QAAQ;AACd,uBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,YAC9D;AAEA,uBAAW,SAAS,QAAQ;AAC1B,mBAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,oBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,oBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,qBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,YACjC;AAAA,UACF;AAEA,oBAAU;AACV,iBAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,QAC1D,SAAS,GAAG;AACV,cAAK,EAAY,YAAY,gBAAgB;AAE3C;AAAA,UACF;AACA,gBAAM;AAAA,QACR;AAAA,MACF;AAEA,YAAM,OAAO,OAAO,WAAwB;AAC1C,cAAM,oBAAoB,aAAa,OAAO;AAC9C,cAAM,SAAS,kBAAkB,UAAU;AAE3C,YAAI;AACF,iBAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,kBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,gBAAI,OAAO,QAAS;AACpB,gBAAI,OAAO,KAAM;AAGjB,kBAAM,cAAc,MAAM,qBAAqB,eAAe,OAAO,KAAK;AAC1E,gBAAI,CAAC,YAAY,SAAS;AACxB,mBAAK,QAAQ;AAAA,gBACX,EAAE,OAAO,YAAY,OAAO,SAAS,OAAO,MAAM;AAAA,gBAClD;AAAA,cACF;AACA;AAAA,YACF;AAEA,kBAAM,QAAwB,YAAY;AAE1C,oBAAQ,MAAM,MAAM;AAAA,cAClB,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,KAAK;AACnC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,IAAI;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,MAAM,GAAG,iCAAiC;AACtE,gCAAgB;AAChB,sBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,KAAK,CAAC,EAAE;AAAA,YAC7E;AAAA,UACF;AAAA,QACF,UAAE;AACA,iBAAO,YAAY;AACnB,cAAI;AACF,kBAAM,kBAAkB,OAAO;AAAA,UACjC,SAAS,GAAG;AACV,iBAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,UACxF;AAAA,QACF;AAAA,MACF;AAEA,UAAI;AACF,aAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,cAAM,aAAa,KAAK;AACxB,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,KAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,UAC/E;AAAA,QACF;AAEA,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,CAAC,SAAS,QAAQ,eAAe,QAAQ,SAAS,MAAM,CAAC;AAAA,YACrE,kBAAkB;AAAA,UACpB,CAAC;AAGD,cAAI,CAAC,kBAAkB,KAAM;AAG7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AAEA,gBAAM;AAAA,YACJ,CAAC,UAAU,gBAAgB,UAAU,iBAAiB;AAAA,YACtD;AAAA,UACF;AACA,0BAAgB;AAAA,QAClB;AAAA,MACF,UAAE;AAEA,wBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA0B,SAAkB;AAEpE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK;AAClB,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AAAA,MAC1D;AAEA,YAAM,aAAyB;AAAA,QAC7B;AAAA,QACA,WAAW,KAAK,kBAAkB,KAAK;AAAA,QACvC,SAAS,KAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,QAClD,YAAY,KAAK;AAAA,QACjB;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,UAChB,CAAC,SACC,kBAAkB;AAAA,YAChB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB,CAAC;AAAA,QACL;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,gBAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAAA,MACH;AAAA,IACF,SAAS,GAAG;AACV,UAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAE/D,aAAK,QAAQ;AAAA,UACX,EAAE,KAAK,EAAE;AAAA,UACT;AAAA,QACF;AAAA,MACF,OAAO;AACL,aAAK,QAAQ,MAAM,EAAE,KAAK,EAAE,GAAG,mCAAmC;AAAA,MACpE;AAAA,IACF;AAAA,EACF;AACF;","names":["ws"]}
1
+ {"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\n STT as BaseSTT,\n SpeechStream as BaseSpeechStream,\n type SpeechData,\n type SpeechEvent,\n SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type TimedString, createTimedString } from '../voice/io.js';\nimport {\n type SttServerEvent,\n type SttTranscriptEvent,\n sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken, getDefaultInferenceUrl } from './utils.js';\n\nexport type DeepgramModels =\n | 'deepgram/flux-general'\n | 'deepgram/nova-3'\n | 'deepgram/nova-3-medical'\n | 'deepgram/nova-2'\n | 'deepgram/nova-2-medical'\n | 'deepgram/nova-2-conversationalai'\n | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels =\n | 'assemblyai/universal-streaming'\n | 'assemblyai/universal-streaming-multilingual';\n\nexport type ElevenlabsSTTModels = 'elevenlabs/scribe_v2_realtime';\n\nexport interface CartesiaOptions {\n /** Minimum volume threshold. Default: not specified. */\n min_volume?: number;\n /** Maximum silence duration in seconds. Default: not specified. */\n max_silence_duration_secs?: number;\n}\n\nexport interface DeepgramOptions {\n /** Enable filler words. Default: true. */\n filler_words?: boolean;\n /** Enable interim results. Default: true. */\n interim_results?: boolean;\n /** Endpointing timeout in milliseconds. Default: 25. */\n endpointing?: number;\n /** Enable punctuation. Default: false. */\n punctuate?: boolean;\n /** Enable smart formatting. */\n smart_format?: boolean;\n /** Keywords with boost values. */\n keywords?: Array<[string, number]>;\n /** Key terms for recognition. */\n keyterms?: string[];\n /** Enable profanity filter. */\n profanity_filter?: boolean;\n /** Convert spoken numbers to numerals. */\n numerals?: boolean;\n /** Opt out of model improvement program. */\n mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n /** Enable turn formatting. Default: false. */\n format_turns?: boolean;\n /** End of turn confidence threshold. Default: 0.01. */\n end_of_turn_confidence_threshold?: number;\n /** Minimum silence duration in milliseconds when confident about end of turn. Default: 0. */\n min_end_of_turn_silence_when_confident?: number;\n /** Maximum turn silence in milliseconds. Default: not specified. */\n max_turn_silence?: number;\n /** Key terms prompt for recognition. Default: not specified. */\n keyterms_prompt?: string[];\n}\n\nexport type STTLanguages =\n | 'multi'\n | 'en'\n | 'de'\n | 'es'\n | 'fr'\n | 'ja'\n | 'pt'\n | 'zh'\n | 'hi'\n | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels | ElevenlabsSTTModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n ? DeepgramOptions\n : TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends AssemblyaiModels\n ? AssemblyAIOptions\n : Record<string, unknown>;\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface STTFallbackModel {\n /** Model name (e.g. \"deepgram/nova-3\", \"assemblyai/universal-streaming\", \"cartesia/ink-whisper\"). */\n model: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type STTFallbackModelType = STTFallbackModel | string;\n\n/** Parse a model string into [model, language]. Language is undefined if not specified. */\nexport function parseSTTModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** Normalize a single or list of FallbackModelType into STTFallbackModel[]. */\nexport function normalizeSTTFallback(\n fallback: STTFallbackModelType | STTFallbackModelType[],\n): STTFallbackModel[] {\n const makeFallback = (model: STTFallbackModelType): STTFallbackModel => {\n if (typeof model === 'string') {\n const [name] = parseSTTModelString(model);\n return { model: name };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n model?: TModel;\n language?: STTLanguages;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\n fallback?: STTFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n private opts: InferenceSTTOptions<TModel>;\n private streams: Set<SpeechStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts?: {\n model?: ModelWithLanguage;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\n fallback?: STTFallbackModelType | STTFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n super({ streaming: true, interimResults: true, alignedTranscript: 'word' });\n\n const {\n model,\n language,\n baseURL,\n encoding = DEFAULT_ENCODING,\n sampleRate = DEFAULT_SAMPLE_RATE,\n apiKey,\n apiSecret,\n modelOptions = {} as STTOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || getDefaultInferenceUrl();\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // Parse language from model string if provided: \"provider/model:language\"\n let nextModel = model;\n let nextLanguage = language;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const languageFromModel = nextModel.slice(idx + 1) as STTLanguages;\n if (nextLanguage && nextLanguage !== languageFromModel) {\n this.#logger.warn(\n '`language` is provided via both argument and model, using the one from the argument',\n { language: nextLanguage, model: nextModel },\n );\n } else {\n nextLanguage = languageFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel as TModel,\n language: nextLanguage,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n get model(): string {\n return this.opts.model ?? 'auto';\n }\n\n get provider(): string {\n return 'livekit';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n const [model, language] = parseSTTModelString(modelString);\n return new STT({ model, language });\n }\n\n protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n stream(options?: {\n language?: STTLanguages | string;\n connOptions?: APIConnectOptions;\n }): SpeechStream<TModel> {\n const { language, connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } =\n options || {};\n const streamOpts = {\n ...this.opts,\n language: language ?? this.opts.language,\n } as InferenceSTTOptions<TModel>;\n\n const stream = new SpeechStream(this, streamOpts, connOptions);\n this.streams.add(stream);\n\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n const params = {\n settings: {\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n },\n } as Record<string, unknown>;\n\n if (this.opts.model && this.opts.model !== 'auto') {\n params.model = this.opts.model;\n }\n\n if (this.opts.language) {\n (params.settings as Record<string, unknown>).language = this.opts.language;\n }\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/stt`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const socket = await connectWs(url, headers, timeout);\n const msg = { ...params, type: 'session.create' };\n socket.send(JSON.stringify(msg));\n\n return socket;\n }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n private opts: InferenceSTTOptions<TModel>;\n private requestId = shortuuid('stt_request_');\n private speaking = false;\n private speechDuration = 0;\n private reconnectEvent = new Event();\n private stt: STT<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(\n sttImpl: STT<TModel>,\n opts: InferenceSTTOptions<TModel>,\n connOptions: APIConnectOptions,\n ) {\n super(sttImpl, opts.sampleRate, connOptions);\n this.opts = opts;\n this.stt = sttImpl;\n this.connOptions = connOptions;\n }\n\n get label(): string {\n return 'inference.SpeechStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n this.reconnectEvent.set();\n }\n\n protected async run(): Promise<void> {\n while (true) {\n // Create fresh resources for each connection attempt\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n\n const eventChannel = createStreamChannel<SttServerEvent>();\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const createWsListener = async (ws: WebSocket, signal: AbortSignal) => {\n return new Promise<void>((resolve, reject) => {\n const onAbort = () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n };\n\n signal.addEventListener('abort', onAbort, { once: true });\n\n ws.on('message', (data) => {\n const json = JSON.parse(data.toString()) as SttServerEvent;\n eventChannel.write(json);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', (code: number) => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'LiveKit STT connection closed unexpectedly',\n options: { statusCode: code },\n }),\n );\n });\n });\n };\n\n const send = async (socket: WebSocket, signal: AbortSignal) => {\n const audioStream = new AudioByteStream(\n this.opts.sampleRate,\n 1,\n Math.floor(this.opts.sampleRate / 20), // 50ms\n );\n\n // Create abort promise once to avoid memory leak\n const abortPromise = new Promise<never>((_, reject) => {\n if (signal.aborted) {\n return reject(new Error('Send aborted'));\n }\n const onAbort = () => reject(new Error('Send aborted'));\n signal.addEventListener('abort', onAbort, { once: true });\n });\n\n // Manual iteration to support cancellation\n const iterator = this.input[Symbol.asyncIterator]();\n try {\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result.done) break;\n const ev = result.value;\n\n let frames: AudioFrame[];\n if (ev === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else {\n const frame = ev as AudioFrame;\n frames = audioStream.write(new Int16Array(frame.data).buffer);\n }\n\n for (const frame of frames) {\n this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n const base64 = Buffer.from(frame.data.buffer).toString('base64');\n const msg = { type: 'input_audio', audio: base64 };\n socket.send(JSON.stringify(msg));\n }\n }\n\n closing = true;\n socket.send(JSON.stringify({ type: 'session.finalize' }));\n } catch (e) {\n if ((e as Error).message === 'Send aborted') {\n // Expected abort, don't log\n return;\n }\n throw e;\n }\n };\n\n const recv = async (signal: AbortSignal) => {\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n // Parse and validate with Zod schema\n const parseResult = await sttServerEventSchema.safeParseAsync(result.value);\n if (!parseResult.success) {\n this.#logger.warn(\n { error: parseResult.error, rawData: result.value },\n 'Failed to parse STT server event',\n );\n continue;\n }\n\n const event: SttServerEvent = parseResult.data;\n\n switch (event.type) {\n case 'session.created':\n case 'session.finalized':\n break;\n case 'session.closed':\n finalReceived = true;\n resourceCleanup();\n break;\n case 'interim_transcript':\n this.processTranscript(event, false);\n break;\n case 'final_transcript':\n this.processTranscript(event, true);\n break;\n case 'error':\n this.#logger.error({ error: event }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.stt.connectWs(this.connOptions.timeoutMs);\n\n const controller = this.abortController; // Use base class abortController for proper cancellation\n const sendTask = Task.from(({ signal }) => send(ws!, signal), controller);\n const wsListenerTask = Task.from(({ signal }) => createWsListener(ws!, signal), controller);\n const recvTask = Task.from(({ signal }) => recv(signal), controller);\n const waitReconnectTask = Task.from(\n ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),\n controller,\n );\n\n try {\n await Promise.race([\n Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),\n waitReconnectTask.result,\n ]);\n\n // If reconnect didn't trigger, tasks finished - exit loop\n if (!waitReconnectTask.done) break;\n\n // Reconnect triggered - clear event and continue loop\n this.reconnectEvent.clear();\n } finally {\n // Cancel all tasks to ensure cleanup\n await cancelAndWait(\n [sendTask, wsListenerTask, recvTask, waitReconnectTask],\n DEFAULT_CANCEL_TIMEOUT,\n );\n resourceCleanup();\n }\n } finally {\n // Ensure cleanup even if connectWs throws\n resourceCleanup();\n }\n }\n }\n\n private processTranscript(data: SttTranscriptEvent, isFinal: boolean) {\n // Check if queue is closed to avoid race condition during disconnect\n if (this.queue.closed) return;\n\n const requestId = data.session_id || this.requestId;\n const text = data.transcript;\n const language = data.language || this.opts.language || 'en';\n\n if (!text && !isFinal) return;\n\n try {\n // We'll have a more accurate way of detecting when speech started when we have VAD\n if (!this.speaking) {\n this.speaking = true;\n this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n }\n\n const speechData: SpeechData = {\n language,\n startTime: this.startTimeOffset + data.start,\n endTime: this.startTimeOffset + data.start + data.duration,\n confidence: data.confidence,\n text,\n words: data.words.map(\n (word): TimedString =>\n createTimedString({\n text: word.word,\n startTime: word.start + this.startTimeOffset,\n endTime: word.end + this.startTimeOffset,\n startTimeOffset: this.startTimeOffset,\n confidence: word.confidence,\n }),\n ),\n };\n\n if (isFinal) {\n if (this.speechDuration > 0) {\n this.queue.put({\n type: SpeechEventType.RECOGNITION_USAGE,\n requestId,\n recognitionUsage: { audioDuration: this.speechDuration },\n });\n this.speechDuration = 0;\n }\n\n this.queue.put({\n type: SpeechEventType.FINAL_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n\n if (this.speaking) {\n this.speaking = false;\n this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n }\n } else {\n this.queue.put({\n type: SpeechEventType.INTERIM_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n }\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n // Expected behavior on disconnect, log as warning\n this.#logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err: e }, 'Error putting transcript to queue');\n }\n }\n }\n}\n"],"mappings":"AAGA,eAAgC;AAEhC,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC;AAAA,EACE,OAAO;AAAA,EACP,gBAAgB;AAAA,EAGhB;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAA2B,OAAO,MAAM,eAAe,WAAW,oBAAoB;AACtF,SAA2B,yBAAyB;AACpD;AAAA,EAGE;AAAA,OACK;AACP,SAAyB,WAAW,mBAAmB,8BAA8B;AAmG9E,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAGO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,IAAI,IAAI,oBAAoB,KAAK;AACxC,aAAO,EAAE,OAAO,KAAK;AAAA,IACvB;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,yBAAyB;AAkBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,YAAY,MAWT;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,mBAAmB,OAAO,CAAC;AAE1E,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,uBAAuB;AACpD,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,eAAe;AACnB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,oBAAoB,UAAU,MAAM,MAAM,CAAC;AACjD,YAAI,gBAAgB,iBAAiB,mBAAmB;AACtD,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,UAAU,cAAc,OAAO,UAAU;AAAA,UAC7C;AAAA,QACF,OAAO;AACL,yBAAe;AAAA,QACjB;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AACA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,UAAU;AAAA,MACV;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK,SAAS;AAAA,EAC5B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,WAAW;AACzD,WAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,EACpC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,KAAK,KAAK,eAAe,4BAA4B,IACnF,WAAW,CAAC;AACd,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AAxSvD;AAySI,UAAM,SAAS;AAAA,MACb,UAAU;AAAA,QACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,QACxC,UAAU,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,QAAI,KAAK,KAAK,UAAU;AACtB,MAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,IACpE;AAEA,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,UAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,WAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,iBAAiB;AAAA,EACnE;AAAA,EACA,YAAY,UAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,MAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,YAAM,eAAe,oBAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOA,KAAe,WAAwB;AACrE,eAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,gBAAM,UAAU,MAAM;AACpB,4BAAgB;AAChB,mBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,UAClD;AAEA,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAExD,UAAAA,IAAG,GAAG,WAAW,CAAC,SAAS;AACzB,kBAAM,OAAO,KAAK,MAAM,KAAK,SAAS,CAAC;AACvC,yBAAa,MAAM,IAAI;AAAA,UACzB,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,iBAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,4BAAgB;AAChB,mBAAO,CAAC;AAAA,UACV,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,SAAiB;AAC/B,4BAAgB;AAEhB,gBAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,gBAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,cACE,IAAI,eAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS,EAAE,YAAY,KAAK;AAAA,cAC9B,CAAC;AAAA,YACH;AAAA,UACF,CAAC;AAAA,QACH,CAAC;AAAA,MACH;AAEA,YAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,cAAM,cAAc,IAAI;AAAA,UACtB,KAAK,KAAK;AAAA,UACV;AAAA,UACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,QACtC;AAGA,cAAM,eAAe,IAAI,QAAe,CAAC,GAAG,WAAW;AACrD,cAAI,OAAO,SAAS;AAClB,mBAAO,OAAO,IAAI,MAAM,cAAc,CAAC;AAAA,UACzC;AACA,gBAAM,UAAU,MAAM,OAAO,IAAI,MAAM,cAAc,CAAC;AACtD,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,QAC1D,CAAC;AAGD,cAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,YAAI;AACF,iBAAO,MAAM;AACX,kBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,gBAAI,OAAO,KAAM;AACjB,kBAAM,KAAK,OAAO;AAElB,gBAAI;AACJ,gBAAI,OAAO,aAAa,gBAAgB;AACtC,uBAAS,YAAY,MAAM;AAAA,YAC7B,OAAO;AACL,oBAAM,QAAQ;AACd,uBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,YAC9D;AAEA,uBAAW,SAAS,QAAQ;AAC1B,mBAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,oBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,oBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,qBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,YACjC;AAAA,UACF;AAEA,oBAAU;AACV,iBAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,QAC1D,SAAS,GAAG;AACV,cAAK,EAAY,YAAY,gBAAgB;AAE3C;AAAA,UACF;AACA,gBAAM;AAAA,QACR;AAAA,MACF;AAEA,YAAM,OAAO,OAAO,WAAwB;AAC1C,cAAM,oBAAoB,aAAa,OAAO;AAC9C,cAAM,SAAS,kBAAkB,UAAU;AAE3C,YAAI;AACF,iBAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,kBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,gBAAI,OAAO,QAAS;AACpB,gBAAI,OAAO,KAAM;AAGjB,kBAAM,cAAc,MAAM,qBAAqB,eAAe,OAAO,KAAK;AAC1E,gBAAI,CAAC,YAAY,SAAS;AACxB,mBAAK,QAAQ;AAAA,gBACX,EAAE,OAAO,YAAY,OAAO,SAAS,OAAO,MAAM;AAAA,gBAClD;AAAA,cACF;AACA;AAAA,YACF;AAEA,kBAAM,QAAwB,YAAY;AAE1C,oBAAQ,MAAM,MAAM;AAAA,cAClB,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,KAAK;AACnC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,IAAI;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,MAAM,GAAG,iCAAiC;AACtE,gCAAgB;AAChB,sBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,KAAK,CAAC,EAAE;AAAA,YAC7E;AAAA,UACF;AAAA,QACF,UAAE;AACA,iBAAO,YAAY;AACnB,cAAI;AACF,kBAAM,kBAAkB,OAAO;AAAA,UACjC,SAAS,GAAG;AACV,iBAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,UACxF;AAAA,QACF;AAAA,MACF;AAEA,UAAI;AACF,aAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,cAAM,aAAa,KAAK;AACxB,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,KAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,UAC/E;AAAA,QACF;AAEA,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,CAAC,SAAS,QAAQ,eAAe,QAAQ,SAAS,MAAM,CAAC;AAAA,YACrE,kBAAkB;AAAA,UACpB,CAAC;AAGD,cAAI,CAAC,kBAAkB,KAAM;AAG7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AAEA,gBAAM;AAAA,YACJ,CAAC,UAAU,gBAAgB,UAAU,iBAAiB;AAAA,YACtD;AAAA,UACF;AACA,0BAAgB;AAAA,QAClB;AAAA,MACF,UAAE;AAEA,wBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA0B,SAAkB;AAEpE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK;AAClB,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AAAA,MAC1D;AAEA,YAAM,aAAyB;AAAA,QAC7B;AAAA,QACA,WAAW,KAAK,kBAAkB,KAAK;AAAA,QACvC,SAAS,KAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,QAClD,YAAY,KAAK;AAAA,QACjB;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,UAChB,CAAC,SACC,kBAAkB;AAAA,YAChB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB,CAAC;AAAA,QACL;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,gBAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAAA,MACH;AAAA,IACF,SAAS,GAAG;AACV,UAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAE/D,aAAK,QAAQ;AAAA,UACX,EAAE,KAAK,EAAE;AAAA,UACT;AAAA,QACF;AAAA,MACF,OAAO;AACL,aAAK,QAAQ,MAAM,EAAE,KAAK,EAAE,GAAG,mCAAmC;AAAA,MACpE;AAAA,IACF;AAAA,EACF;AACF;","names":["ws"]}
@@ -58,7 +58,6 @@ function normalizeTTSFallback(fallback) {
58
58
  }
59
59
  const DEFAULT_ENCODING = "pcm_s16le";
60
60
  const DEFAULT_SAMPLE_RATE = 16e3;
61
- const DEFAULT_BASE_URL = "https://agent-gateway.livekit.cloud/v1";
62
61
  const NUM_CHANNELS = 1;
63
62
  const DEFAULT_LANGUAGE = "en";
64
63
  class TTS extends import_tts.TTS {
@@ -81,7 +80,7 @@ class TTS extends import_tts.TTS {
81
80
  fallback,
82
81
  connOptions
83
82
  } = opts || {};
84
- const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;
83
+ const lkBaseURL = baseURL || (0, import_utils2.getDefaultInferenceUrl)();
85
84
  const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;
86
85
  if (!lkApiKey) {
87
86
  throw new Error("apiKey is required: pass apiKey or set LIVEKIT_API_KEY");
@@ -133,6 +132,12 @@ class TTS extends import_tts.TTS {
133
132
  get label() {
134
133
  return "inference.TTS";
135
134
  }
135
+ get model() {
136
+ return this.opts.model ?? "unknown";
137
+ }
138
+ get provider() {
139
+ return "livekit";
140
+ }
136
141
  static fromModelString(modelString) {
137
142
  const [model, voice] = parseTTSModelString(modelString);
138
143
  return new TTS({ model, voice: voice || void 0 });