@getpaseo/server 0.1.16 → 0.1.17

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 (346) hide show
  1. package/dist/scripts/daemon-runner.js +53 -14
  2. package/dist/scripts/daemon-runner.js.map +1 -1
  3. package/dist/scripts/dev-runner.js +9 -16
  4. package/dist/scripts/dev-runner.js.map +1 -1
  5. package/dist/scripts/supervisor.js +40 -13
  6. package/dist/scripts/supervisor.js.map +1 -1
  7. package/dist/server/client/daemon-client.d.ts +23 -3
  8. package/dist/server/client/daemon-client.d.ts.map +1 -1
  9. package/dist/server/client/daemon-client.js +81 -8
  10. package/dist/server/client/daemon-client.js.map +1 -1
  11. package/dist/server/server/agent/agent-manager.d.ts +3 -1
  12. package/dist/server/server/agent/agent-manager.d.ts.map +1 -1
  13. package/dist/server/server/agent/agent-manager.js +146 -24
  14. package/dist/server/server/agent/agent-manager.js.map +1 -1
  15. package/dist/server/server/agent/agent-metadata-generator.d.ts.map +1 -1
  16. package/dist/server/server/agent/agent-metadata-generator.js +13 -4
  17. package/dist/server/server/agent/agent-metadata-generator.js.map +1 -1
  18. package/dist/server/server/agent/agent-response-loop.js +1 -1
  19. package/dist/server/server/agent/agent-response-loop.js.map +1 -1
  20. package/dist/server/server/agent/agent-sdk-types.d.ts +9 -0
  21. package/dist/server/server/agent/agent-sdk-types.d.ts.map +1 -1
  22. package/dist/server/server/agent/agent-sdk-types.js +11 -1
  23. package/dist/server/server/agent/agent-sdk-types.js.map +1 -1
  24. package/dist/server/server/agent/agent-storage.d.ts +5 -1
  25. package/dist/server/server/agent/agent-storage.d.ts.map +1 -1
  26. package/dist/server/server/agent/agent-storage.js +41 -72
  27. package/dist/server/server/agent/agent-storage.js.map +1 -1
  28. package/dist/server/server/agent/agent-title-limits.d.ts +3 -0
  29. package/dist/server/server/agent/agent-title-limits.d.ts.map +1 -0
  30. package/dist/server/server/agent/agent-title-limits.js +3 -0
  31. package/dist/server/server/agent/agent-title-limits.js.map +1 -0
  32. package/dist/server/server/agent/providers/claude/model-catalog.d.ts +29 -0
  33. package/dist/server/server/agent/providers/claude/model-catalog.d.ts.map +1 -0
  34. package/dist/server/server/agent/providers/claude/model-catalog.js +70 -0
  35. package/dist/server/server/agent/providers/claude/model-catalog.js.map +1 -0
  36. package/dist/server/server/agent/providers/claude/task-notification-tool-call.d.ts +44 -0
  37. package/dist/server/server/agent/providers/claude/task-notification-tool-call.d.ts.map +1 -0
  38. package/dist/server/server/agent/providers/claude/task-notification-tool-call.js +250 -0
  39. package/dist/server/server/agent/providers/claude/task-notification-tool-call.js.map +1 -0
  40. package/dist/server/server/agent/providers/claude/tool-call-detail-parser.d.ts.map +1 -1
  41. package/dist/server/server/agent/providers/claude/tool-call-detail-parser.js +15 -0
  42. package/dist/server/server/agent/providers/claude/tool-call-detail-parser.js.map +1 -1
  43. package/dist/server/server/agent/providers/claude-agent.d.ts +3 -2
  44. package/dist/server/server/agent/providers/claude-agent.d.ts.map +1 -1
  45. package/dist/server/server/agent/providers/claude-agent.js +240 -106
  46. package/dist/server/server/agent/providers/claude-agent.js.map +1 -1
  47. package/dist/server/server/agent/providers/codex/tool-call-mapper.d.ts.map +1 -1
  48. package/dist/server/server/agent/providers/codex/tool-call-mapper.js +81 -28
  49. package/dist/server/server/agent/providers/codex/tool-call-mapper.js.map +1 -1
  50. package/dist/server/server/agent/providers/codex-app-server-agent.d.ts.map +1 -1
  51. package/dist/server/server/agent/providers/codex-app-server-agent.js +31 -5
  52. package/dist/server/server/agent/providers/codex-app-server-agent.js.map +1 -1
  53. package/dist/server/server/agent/providers/opencode-agent.d.ts +10 -1
  54. package/dist/server/server/agent/providers/opencode-agent.d.ts.map +1 -1
  55. package/dist/server/server/agent/providers/opencode-agent.js +207 -176
  56. package/dist/server/server/agent/providers/opencode-agent.js.map +1 -1
  57. package/dist/server/server/agent/providers/tool-call-detail-primitives.d.ts +15 -0
  58. package/dist/server/server/agent/providers/tool-call-detail-primitives.d.ts.map +1 -1
  59. package/dist/server/server/agent/timeline-projection.d.ts +20 -0
  60. package/dist/server/server/agent/timeline-projection.d.ts.map +1 -1
  61. package/dist/server/server/agent/timeline-projection.js +73 -0
  62. package/dist/server/server/agent/timeline-projection.js.map +1 -1
  63. package/dist/server/server/bootstrap.d.ts +15 -0
  64. package/dist/server/server/bootstrap.d.ts.map +1 -1
  65. package/dist/server/server/bootstrap.js +27 -4
  66. package/dist/server/server/bootstrap.js.map +1 -1
  67. package/dist/server/server/file-download/token-store.d.ts +0 -1
  68. package/dist/server/server/file-download/token-store.d.ts.map +1 -1
  69. package/dist/server/server/file-download/token-store.js.map +1 -1
  70. package/dist/server/server/file-explorer/service.d.ts.map +1 -1
  71. package/dist/server/server/file-explorer/service.js +56 -36
  72. package/dist/server/server/file-explorer/service.js.map +1 -1
  73. package/dist/server/server/index.js +85 -29
  74. package/dist/server/server/index.js.map +1 -1
  75. package/dist/server/server/logger.d.ts +24 -3
  76. package/dist/server/server/logger.d.ts.map +1 -1
  77. package/dist/server/server/logger.js +157 -21
  78. package/dist/server/server/logger.js.map +1 -1
  79. package/dist/server/server/persisted-config.d.ts +86 -0
  80. package/dist/server/server/persisted-config.d.ts.map +1 -1
  81. package/dist/server/server/persisted-config.js +25 -3
  82. package/dist/server/server/persisted-config.js.map +1 -1
  83. package/dist/server/server/pid-lock.d.ts +6 -2
  84. package/dist/server/server/pid-lock.d.ts.map +1 -1
  85. package/dist/server/server/pid-lock.js +7 -10
  86. package/dist/server/server/pid-lock.js.map +1 -1
  87. package/dist/server/server/relay-transport.d.ts.map +1 -1
  88. package/dist/server/server/relay-transport.js +1 -0
  89. package/dist/server/server/relay-transport.js.map +1 -1
  90. package/dist/server/server/session.d.ts +57 -3
  91. package/dist/server/server/session.d.ts.map +1 -1
  92. package/dist/server/server/session.js +755 -182
  93. package/dist/server/server/session.js.map +1 -1
  94. package/dist/server/server/websocket-server.d.ts +16 -1
  95. package/dist/server/server/websocket-server.d.ts.map +1 -1
  96. package/dist/server/server/websocket-server.js +135 -9
  97. package/dist/server/server/websocket-server.js.map +1 -1
  98. package/dist/server/server/worktree-bootstrap.d.ts.map +1 -1
  99. package/dist/server/server/worktree-bootstrap.js +45 -2
  100. package/dist/server/server/worktree-bootstrap.js.map +1 -1
  101. package/dist/server/shared/messages.d.ts +2841 -541
  102. package/dist/server/shared/messages.d.ts.map +1 -1
  103. package/dist/server/shared/messages.js +99 -5
  104. package/dist/server/shared/messages.js.map +1 -1
  105. package/dist/server/shared/tool-call-display.d.ts.map +1 -1
  106. package/dist/server/shared/tool-call-display.js +3 -0
  107. package/dist/server/shared/tool-call-display.js.map +1 -1
  108. package/dist/server/terminal/terminal-manager.d.ts.map +1 -1
  109. package/dist/server/terminal/terminal-manager.js +1 -13
  110. package/dist/server/terminal/terminal-manager.js.map +1 -1
  111. package/dist/server/terminal/terminal.d.ts.map +1 -1
  112. package/dist/server/terminal/terminal.js +29 -5
  113. package/dist/server/terminal/terminal.js.map +1 -1
  114. package/dist/server/utils/worktree.d.ts +1 -0
  115. package/dist/server/utils/worktree.d.ts.map +1 -1
  116. package/dist/server/utils/worktree.js +17 -2
  117. package/dist/server/utils/worktree.js.map +1 -1
  118. package/dist/src/server/agent/activity-curator.js +228 -0
  119. package/dist/src/server/agent/activity-curator.js.map +1 -0
  120. package/dist/src/server/agent/agent-manager.js +1712 -0
  121. package/dist/src/server/agent/agent-manager.js.map +1 -0
  122. package/dist/src/server/agent/agent-metadata-generator.js +163 -0
  123. package/dist/src/server/agent/agent-metadata-generator.js.map +1 -0
  124. package/dist/src/server/agent/agent-projections.js +262 -0
  125. package/dist/src/server/agent/agent-projections.js.map +1 -0
  126. package/dist/src/server/agent/agent-response-loop.js +304 -0
  127. package/dist/src/server/agent/agent-response-loop.js.map +1 -0
  128. package/dist/src/server/agent/agent-sdk-types.js +12 -0
  129. package/dist/src/server/agent/agent-sdk-types.js.map +1 -0
  130. package/dist/src/server/agent/agent-storage.js +299 -0
  131. package/dist/src/server/agent/agent-storage.js.map +1 -0
  132. package/dist/src/server/agent/agent-title-limits.js +3 -0
  133. package/dist/src/server/agent/agent-title-limits.js.map +1 -0
  134. package/dist/src/server/agent/audio-utils.js +19 -0
  135. package/dist/src/server/agent/audio-utils.js.map +1 -0
  136. package/dist/src/server/agent/dictation-debug.js +50 -0
  137. package/dist/src/server/agent/dictation-debug.js.map +1 -0
  138. package/dist/src/server/agent/mcp-server.js +787 -0
  139. package/dist/src/server/agent/mcp-server.js.map +1 -0
  140. package/dist/src/server/agent/orchestrator-instructions.js +51 -0
  141. package/dist/src/server/agent/orchestrator-instructions.js.map +1 -0
  142. package/dist/src/server/agent/pcm16-resampler.js +63 -0
  143. package/dist/src/server/agent/pcm16-resampler.js.map +1 -0
  144. package/dist/src/server/agent/provider-launch-config.js +83 -0
  145. package/dist/src/server/agent/provider-launch-config.js.map +1 -0
  146. package/dist/src/server/agent/provider-manifest.js +97 -0
  147. package/dist/src/server/agent/provider-manifest.js.map +1 -0
  148. package/dist/src/server/agent/provider-registry.js +45 -0
  149. package/dist/src/server/agent/provider-registry.js.map +1 -0
  150. package/dist/src/server/agent/providers/claude/model-catalog.js +70 -0
  151. package/dist/src/server/agent/providers/claude/model-catalog.js.map +1 -0
  152. package/dist/src/server/agent/providers/claude/task-notification-tool-call.js +250 -0
  153. package/dist/src/server/agent/providers/claude/task-notification-tool-call.js.map +1 -0
  154. package/dist/src/server/agent/providers/claude/tool-call-detail-parser.js +109 -0
  155. package/dist/src/server/agent/providers/claude/tool-call-detail-parser.js.map +1 -0
  156. package/dist/src/server/agent/providers/claude/tool-call-mapper.js +238 -0
  157. package/dist/src/server/agent/providers/claude/tool-call-mapper.js.map +1 -0
  158. package/dist/src/server/agent/providers/claude-agent.js +3747 -0
  159. package/dist/src/server/agent/providers/claude-agent.js.map +1 -0
  160. package/dist/src/server/agent/providers/codex/tool-call-detail-parser.js +104 -0
  161. package/dist/src/server/agent/providers/codex/tool-call-detail-parser.js.map +1 -0
  162. package/dist/src/server/agent/providers/codex/tool-call-mapper.js +720 -0
  163. package/dist/src/server/agent/providers/codex/tool-call-mapper.js.map +1 -0
  164. package/dist/src/server/agent/providers/codex-app-server-agent.js +2601 -0
  165. package/dist/src/server/agent/providers/codex-app-server-agent.js.map +1 -0
  166. package/dist/src/server/agent/providers/codex-rollout-timeline.js +487 -0
  167. package/dist/src/server/agent/providers/codex-rollout-timeline.js.map +1 -0
  168. package/dist/src/server/agent/providers/opencode/tool-call-detail-parser.js +39 -0
  169. package/dist/src/server/agent/providers/opencode/tool-call-detail-parser.js.map +1 -0
  170. package/dist/src/server/agent/providers/opencode/tool-call-mapper.js +151 -0
  171. package/dist/src/server/agent/providers/opencode/tool-call-mapper.js.map +1 -0
  172. package/dist/src/server/agent/providers/opencode-agent.js +905 -0
  173. package/dist/src/server/agent/providers/opencode-agent.js.map +1 -0
  174. package/dist/src/server/agent/providers/tool-call-detail-primitives.js +552 -0
  175. package/dist/src/server/agent/providers/tool-call-detail-primitives.js.map +1 -0
  176. package/dist/src/server/agent/providers/tool-call-mapper-utils.js +109 -0
  177. package/dist/src/server/agent/providers/tool-call-mapper-utils.js.map +1 -0
  178. package/dist/src/server/agent/recordings-debug.js +19 -0
  179. package/dist/src/server/agent/recordings-debug.js.map +1 -0
  180. package/dist/src/server/agent/stt-debug.js +33 -0
  181. package/dist/src/server/agent/stt-debug.js.map +1 -0
  182. package/dist/src/server/agent/stt-manager.js +233 -0
  183. package/dist/src/server/agent/stt-manager.js.map +1 -0
  184. package/dist/src/server/agent/timeline-append.js +27 -0
  185. package/dist/src/server/agent/timeline-append.js.map +1 -0
  186. package/dist/src/server/agent/timeline-projection.js +215 -0
  187. package/dist/src/server/agent/timeline-projection.js.map +1 -0
  188. package/dist/src/server/agent/tool-name-normalization.js +45 -0
  189. package/dist/src/server/agent/tool-name-normalization.js.map +1 -0
  190. package/dist/src/server/agent/tts-debug.js +24 -0
  191. package/dist/src/server/agent/tts-debug.js.map +1 -0
  192. package/dist/src/server/agent/tts-manager.js +249 -0
  193. package/dist/src/server/agent/tts-manager.js.map +1 -0
  194. package/dist/src/server/agent/wait-for-agent-tracker.js +53 -0
  195. package/dist/src/server/agent/wait-for-agent-tracker.js.map +1 -0
  196. package/dist/src/server/agent-attention-policy.js +40 -0
  197. package/dist/src/server/agent-attention-policy.js.map +1 -0
  198. package/dist/src/server/allowed-hosts.js +94 -0
  199. package/dist/src/server/allowed-hosts.js.map +1 -0
  200. package/dist/src/server/bootstrap.js +498 -0
  201. package/dist/src/server/bootstrap.js.map +1 -0
  202. package/dist/src/server/client-message-id.js +12 -0
  203. package/dist/src/server/client-message-id.js.map +1 -0
  204. package/dist/src/server/config.js +84 -0
  205. package/dist/src/server/config.js.map +1 -0
  206. package/dist/src/server/connection-offer.js +60 -0
  207. package/dist/src/server/connection-offer.js.map +1 -0
  208. package/dist/src/server/daemon-keypair.js +40 -0
  209. package/dist/src/server/daemon-keypair.js.map +1 -0
  210. package/dist/src/server/daemon-version.js +22 -0
  211. package/dist/src/server/daemon-version.js.map +1 -0
  212. package/dist/src/server/dictation/dictation-stream-manager.js +568 -0
  213. package/dist/src/server/dictation/dictation-stream-manager.js.map +1 -0
  214. package/dist/src/server/file-download/token-store.js +40 -0
  215. package/dist/src/server/file-download/token-store.js.map +1 -0
  216. package/dist/src/server/file-explorer/service.js +183 -0
  217. package/dist/src/server/file-explorer/service.js.map +1 -0
  218. package/dist/src/server/json-utils.js +45 -0
  219. package/dist/src/server/json-utils.js.map +1 -0
  220. package/dist/src/server/messages.js +29 -0
  221. package/dist/src/server/messages.js.map +1 -0
  222. package/dist/src/server/package-version.js +47 -0
  223. package/dist/src/server/package-version.js.map +1 -0
  224. package/dist/src/server/paseo-home.js +19 -0
  225. package/dist/src/server/paseo-home.js.map +1 -0
  226. package/dist/src/server/path-utils.js +20 -0
  227. package/dist/src/server/path-utils.js.map +1 -0
  228. package/dist/src/server/persisted-config.js +259 -0
  229. package/dist/src/server/persisted-config.js.map +1 -0
  230. package/dist/src/server/persistence-hooks.js +60 -0
  231. package/dist/src/server/persistence-hooks.js.map +1 -0
  232. package/dist/src/server/pid-lock.js +126 -0
  233. package/dist/src/server/pid-lock.js.map +1 -0
  234. package/dist/src/server/push/push-service.js +68 -0
  235. package/dist/src/server/push/push-service.js.map +1 -0
  236. package/dist/src/server/push/token-store.js +70 -0
  237. package/dist/src/server/push/token-store.js.map +1 -0
  238. package/dist/src/server/relay-transport.js +457 -0
  239. package/dist/src/server/relay-transport.js.map +1 -0
  240. package/dist/src/server/server-id.js +63 -0
  241. package/dist/src/server/server-id.js.map +1 -0
  242. package/dist/src/server/session.js +5947 -0
  243. package/dist/src/server/session.js.map +1 -0
  244. package/dist/src/server/speech/audio.js +101 -0
  245. package/dist/src/server/speech/audio.js.map +1 -0
  246. package/dist/src/server/speech/provider-resolver.js +7 -0
  247. package/dist/src/server/speech/provider-resolver.js.map +1 -0
  248. package/dist/src/server/speech/providers/local/config.js +83 -0
  249. package/dist/src/server/speech/providers/local/config.js.map +1 -0
  250. package/dist/src/server/speech/providers/local/models.js +17 -0
  251. package/dist/src/server/speech/providers/local/models.js.map +1 -0
  252. package/dist/src/server/speech/providers/local/pocket/pocket-tts-onnx.js +422 -0
  253. package/dist/src/server/speech/providers/local/pocket/pocket-tts-onnx.js.map +1 -0
  254. package/dist/src/server/speech/providers/local/runtime.js +253 -0
  255. package/dist/src/server/speech/providers/local/runtime.js.map +1 -0
  256. package/dist/src/server/speech/providers/local/sherpa/model-catalog.js +166 -0
  257. package/dist/src/server/speech/providers/local/sherpa/model-catalog.js.map +1 -0
  258. package/dist/src/server/speech/providers/local/sherpa/model-downloader.js +165 -0
  259. package/dist/src/server/speech/providers/local/sherpa/model-downloader.js.map +1 -0
  260. package/dist/src/server/speech/providers/local/sherpa/sherpa-offline-recognizer.js +68 -0
  261. package/dist/src/server/speech/providers/local/sherpa/sherpa-offline-recognizer.js.map +1 -0
  262. package/dist/src/server/speech/providers/local/sherpa/sherpa-online-recognizer.js +79 -0
  263. package/dist/src/server/speech/providers/local/sherpa/sherpa-online-recognizer.js.map +1 -0
  264. package/dist/src/server/speech/providers/local/sherpa/sherpa-onnx-loader.js +11 -0
  265. package/dist/src/server/speech/providers/local/sherpa/sherpa-onnx-loader.js.map +1 -0
  266. package/dist/src/server/speech/providers/local/sherpa/sherpa-onnx-node-loader.js +102 -0
  267. package/dist/src/server/speech/providers/local/sherpa/sherpa-onnx-node-loader.js.map +1 -0
  268. package/dist/src/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.js +131 -0
  269. package/dist/src/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.js.map +1 -0
  270. package/dist/src/server/speech/providers/local/sherpa/sherpa-parakeet-stt.js +132 -0
  271. package/dist/src/server/speech/providers/local/sherpa/sherpa-parakeet-stt.js.map +1 -0
  272. package/dist/src/server/speech/providers/local/sherpa/sherpa-realtime-session.js +112 -0
  273. package/dist/src/server/speech/providers/local/sherpa/sherpa-realtime-session.js.map +1 -0
  274. package/dist/src/server/speech/providers/local/sherpa/sherpa-stt.js +140 -0
  275. package/dist/src/server/speech/providers/local/sherpa/sherpa-stt.js.map +1 -0
  276. package/dist/src/server/speech/providers/local/sherpa/sherpa-tts.js +95 -0
  277. package/dist/src/server/speech/providers/local/sherpa/sherpa-tts.js.map +1 -0
  278. package/dist/src/server/speech/providers/openai/config.js +99 -0
  279. package/dist/src/server/speech/providers/openai/config.js.map +1 -0
  280. package/dist/src/server/speech/providers/openai/realtime-transcription-session.js +165 -0
  281. package/dist/src/server/speech/providers/openai/realtime-transcription-session.js.map +1 -0
  282. package/dist/src/server/speech/providers/openai/runtime.js +114 -0
  283. package/dist/src/server/speech/providers/openai/runtime.js.map +1 -0
  284. package/dist/src/server/speech/providers/openai/stt.js +208 -0
  285. package/dist/src/server/speech/providers/openai/stt.js.map +1 -0
  286. package/dist/src/server/speech/providers/openai/tts.js +46 -0
  287. package/dist/src/server/speech/providers/openai/tts.js.map +1 -0
  288. package/dist/src/server/speech/speech-config-resolver.js +85 -0
  289. package/dist/src/server/speech/speech-config-resolver.js.map +1 -0
  290. package/dist/src/server/speech/speech-provider.js +2 -0
  291. package/dist/src/server/speech/speech-provider.js.map +1 -0
  292. package/dist/src/server/speech/speech-runtime.js +497 -0
  293. package/dist/src/server/speech/speech-runtime.js.map +1 -0
  294. package/dist/src/server/speech/speech-types.js +8 -0
  295. package/dist/src/server/speech/speech-types.js.map +1 -0
  296. package/dist/src/server/utils/diff-highlighter.js +244 -0
  297. package/dist/src/server/utils/diff-highlighter.js.map +1 -0
  298. package/dist/src/server/utils/syntax-highlighter.js +145 -0
  299. package/dist/src/server/utils/syntax-highlighter.js.map +1 -0
  300. package/dist/src/server/voice-config.js +51 -0
  301. package/dist/src/server/voice-config.js.map +1 -0
  302. package/dist/src/server/voice-mcp-bridge-command.js +31 -0
  303. package/dist/src/server/voice-mcp-bridge-command.js.map +1 -0
  304. package/dist/src/server/voice-mcp-bridge.js +109 -0
  305. package/dist/src/server/voice-mcp-bridge.js.map +1 -0
  306. package/dist/src/server/voice-permission-policy.js +13 -0
  307. package/dist/src/server/voice-permission-policy.js.map +1 -0
  308. package/dist/src/server/voice-types.js +2 -0
  309. package/dist/src/server/voice-types.js.map +1 -0
  310. package/dist/src/server/websocket-server.js +967 -0
  311. package/dist/src/server/websocket-server.js.map +1 -0
  312. package/dist/src/server/worktree-bootstrap.js +497 -0
  313. package/dist/src/server/worktree-bootstrap.js.map +1 -0
  314. package/dist/src/shared/agent-attention-notification.js +130 -0
  315. package/dist/src/shared/agent-attention-notification.js.map +1 -0
  316. package/dist/src/shared/agent-lifecycle.js +8 -0
  317. package/dist/src/shared/agent-lifecycle.js.map +1 -0
  318. package/dist/src/shared/binary-mux.js +114 -0
  319. package/dist/src/shared/binary-mux.js.map +1 -0
  320. package/dist/src/shared/connection-offer.js +17 -0
  321. package/dist/src/shared/connection-offer.js.map +1 -0
  322. package/dist/src/shared/daemon-endpoints.js +113 -0
  323. package/dist/src/shared/daemon-endpoints.js.map +1 -0
  324. package/dist/src/shared/messages.js +2001 -0
  325. package/dist/src/shared/messages.js.map +1 -0
  326. package/dist/src/shared/path-utils.js +16 -0
  327. package/dist/src/shared/path-utils.js.map +1 -0
  328. package/dist/src/shared/tool-call-display.js +93 -0
  329. package/dist/src/shared/tool-call-display.js.map +1 -0
  330. package/dist/src/terminal/terminal-manager.js +136 -0
  331. package/dist/src/terminal/terminal-manager.js.map +1 -0
  332. package/dist/src/terminal/terminal.js +410 -0
  333. package/dist/src/terminal/terminal.js.map +1 -0
  334. package/dist/src/utils/checkout-git.js +1397 -0
  335. package/dist/src/utils/checkout-git.js.map +1 -0
  336. package/dist/src/utils/directory-suggestions.js +655 -0
  337. package/dist/src/utils/directory-suggestions.js.map +1 -0
  338. package/dist/src/utils/path.js +15 -0
  339. package/dist/src/utils/path.js.map +1 -0
  340. package/dist/src/utils/project-icon.js +391 -0
  341. package/dist/src/utils/project-icon.js.map +1 -0
  342. package/dist/src/utils/worktree-metadata.js +116 -0
  343. package/dist/src/utils/worktree-metadata.js.map +1 -0
  344. package/dist/src/utils/worktree.js +741 -0
  345. package/dist/src/utils/worktree.js.map +1 -0
  346. package/package.json +14 -6
@@ -0,0 +1,112 @@
1
+ import { EventEmitter } from "node:events";
2
+ import { v4 as uuidv4 } from "uuid";
3
+ import { pcm16lePeakAbs, pcm16leToFloat32 } from "../../../audio.js";
4
+ export class SherpaRealtimeTranscriptionSession extends EventEmitter {
5
+ constructor(params) {
6
+ super();
7
+ this.stream = null;
8
+ this.connected = false;
9
+ this.currentSegmentId = null;
10
+ this.previousSegmentId = null;
11
+ this.lastPartialText = "";
12
+ this.engine = params.engine;
13
+ this.requiredSampleRate = this.engine.sampleRate;
14
+ this.tailPaddingMs = params.tailPaddingMs ?? 500;
15
+ }
16
+ async connect() {
17
+ if (this.connected) {
18
+ return;
19
+ }
20
+ this.stream = this.engine.createStream();
21
+ this.currentSegmentId = uuidv4();
22
+ this.connected = true;
23
+ }
24
+ appendPcm16(pcm16le) {
25
+ if (!this.connected || !this.stream || !this.currentSegmentId) {
26
+ this.emit("error", new Error("Sherpa realtime session not connected"));
27
+ return;
28
+ }
29
+ try {
30
+ const peak = pcm16lePeakAbs(pcm16le);
31
+ const peakFloat = peak / 32768.0;
32
+ const targetPeak = 0.6;
33
+ const maxGain = 50;
34
+ const gain = peakFloat > 0 && peakFloat < targetPeak
35
+ ? Math.min(maxGain, targetPeak / peakFloat)
36
+ : 1;
37
+ const floatSamples = pcm16leToFloat32(pcm16le, gain);
38
+ this.stream.acceptWaveform(this.engine.sampleRate, floatSamples);
39
+ while (this.engine.recognizer.isReady(this.stream)) {
40
+ this.engine.recognizer.decode(this.stream);
41
+ }
42
+ const text = String(this.engine.recognizer.getResult(this.stream)?.text ?? "").trim();
43
+ if (text !== this.lastPartialText) {
44
+ this.lastPartialText = text;
45
+ this.emit("transcript", {
46
+ segmentId: this.currentSegmentId,
47
+ transcript: text,
48
+ isFinal: false,
49
+ });
50
+ }
51
+ }
52
+ catch (err) {
53
+ this.emit("error", err instanceof Error ? err : new Error(String(err)));
54
+ }
55
+ }
56
+ commit() {
57
+ if (!this.connected || !this.stream || !this.currentSegmentId) {
58
+ this.emit("error", new Error("Sherpa realtime session not connected"));
59
+ return;
60
+ }
61
+ try {
62
+ const padSamples = Math.max(0, Math.round((this.engine.sampleRate * this.tailPaddingMs) / 1000));
63
+ if (padSamples > 0) {
64
+ this.stream.acceptWaveform(this.engine.sampleRate, new Float32Array(padSamples));
65
+ }
66
+ while (this.engine.recognizer.isReady(this.stream)) {
67
+ this.engine.recognizer.decode(this.stream);
68
+ }
69
+ const finalText = String(this.engine.recognizer.getResult(this.stream)?.text ?? "").trim();
70
+ const segmentId = this.currentSegmentId;
71
+ const previousSegmentId = this.previousSegmentId;
72
+ this.emit("committed", { segmentId, previousSegmentId });
73
+ this.emit("transcript", { segmentId, transcript: finalText, isFinal: true });
74
+ this.previousSegmentId = segmentId;
75
+ this.currentSegmentId = uuidv4();
76
+ this.lastPartialText = "";
77
+ this.engine.recognizer.reset(this.stream);
78
+ }
79
+ catch (err) {
80
+ this.emit("error", err instanceof Error ? err : new Error(String(err)));
81
+ }
82
+ }
83
+ clear() {
84
+ if (!this.connected || !this.stream) {
85
+ return;
86
+ }
87
+ try {
88
+ this.engine.recognizer.reset(this.stream);
89
+ this.currentSegmentId = uuidv4();
90
+ this.lastPartialText = "";
91
+ }
92
+ catch (err) {
93
+ this.emit("error", err instanceof Error ? err : new Error(String(err)));
94
+ }
95
+ }
96
+ close() {
97
+ if (!this.stream) {
98
+ return;
99
+ }
100
+ try {
101
+ this.stream.free?.();
102
+ }
103
+ catch {
104
+ // ignore
105
+ }
106
+ finally {
107
+ this.stream = null;
108
+ this.connected = false;
109
+ }
110
+ }
111
+ }
112
+ //# sourceMappingURL=sherpa-realtime-session.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"sherpa-realtime-session.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-realtime-session.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,EAAE,IAAI,MAAM,EAAE,MAAM,MAAM,CAAC;AAGpC,OAAO,EAAE,cAAc,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAGrE,MAAM,OAAO,kCACX,SAAQ,YAAY;IAapB,YAAY,MAAwE;QAClF,KAAK,EAAE,CAAC;QAVF,WAAM,GAAe,IAAI,CAAC;QAC1B,cAAS,GAAG,KAAK,CAAC;QAGlB,qBAAgB,GAAkB,IAAI,CAAC;QACvC,sBAAiB,GAAkB,IAAI,CAAC;QACxC,oBAAe,GAAG,EAAE,CAAC;QAK3B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QACjD,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,GAAG,CAAC;IACnD,CAAC;IAED,KAAK,CAAC,OAAO;QACX,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,OAAO;QACT,CAAC;QACD,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;QACzC,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;QACjC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;IACxB,CAAC;IAED,WAAW,CAAC,OAAe;QACzB,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,uCAAuC,CAAC,CAAC,CAAC;YACvE,OAAO;QACT,CAAC;QAED,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,cAAc,CAAC,OAAO,CAAC,CAAC;YACrC,MAAM,SAAS,GAAG,IAAI,GAAG,OAAO,CAAC;YACjC,MAAM,UAAU,GAAG,GAAG,CAAC;YACvB,MAAM,OAAO,GAAG,EAAE,CAAC;YACnB,MAAM,IAAI,GACR,SAAS,GAAG,CAAC,IAAI,SAAS,GAAG,UAAU;gBACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,GAAG,SAAS,CAAC;gBAC3C,CAAC,CAAC,CAAC,CAAC;YACR,MAAM,YAAY,GAAG,gBAAgB,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;YACrD,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;YAEjE,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;gBACnD,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAC7C,CAAC;YAED,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YACtF,IAAI,IAAI,KAAK,IAAI,CAAC,eAAe,EAAE,CAAC;gBAClC,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC;gBAC5B,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE;oBACtB,SAAS,EAAE,IAAI,CAAC,gBAAgB;oBAChC,UAAU,EAAE,IAAI;oBAChB,OAAO,EAAE,KAAK;iBACf,CAAC,CAAC;YACL,CAAC;QACH,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,MAAM;QACJ,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,uCAAuC,CAAC,CAAC,CAAC;YACvE,OAAO;QACT,CAAC;QAED,IAAI,CAAC;YACH,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC;YACjG,IAAI,UAAU,GAAG,CAAC,EAAE,CAAC;gBACnB,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,IAAI,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC;YACnF,CAAC;YAED,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;gBACnD,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAC7C,CAAC;YAED,MAAM,SAAS,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YAC3F,MAAM,SAAS,GAAG,IAAI,CAAC,gBAAgB,CAAC;YACxC,MAAM,iBAAiB,GAAG,IAAI,CAAC,iBAAiB,CAAC;YAEjD,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,iBAAiB,EAAE,CAAC,CAAC;YACzD,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,UAAU,EAAE,SAAS,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;YAE7E,IAAI,CAAC,iBAAiB,GAAG,SAAS,CAAC;YACnC,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;YACjC,IAAI,CAAC,eAAe,GAAG,EAAE,CAAC;YAC1B,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC5C,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;YACpC,OAAO;QACT,CAAC;QACD,IAAI,CAAC;YACH,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAC1C,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;YACjC,IAAI,CAAC,eAAe,GAAG,EAAE,CAAC;QAC5B,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;YACjB,OAAO;QACT,CAAC;QACD,IAAI,CAAC;YACH,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;QACvB,CAAC;QAAC,MAAM,CAAC;YACP,SAAS;QACX,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;YACnB,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;QACzB,CAAC;IACH,CAAC;CACF"}
@@ -0,0 +1,140 @@
1
+ import { EventEmitter } from "node:events";
2
+ import { v4 as uuidv4 } from "uuid";
3
+ import { Pcm16MonoResampler } from "../../../../agent/pcm16-resampler.js";
4
+ import { parsePcm16MonoWav, parsePcmRateFromFormat, pcm16lePeakAbs, pcm16leToFloat32 } from "../../../audio.js";
5
+ export class SherpaOnnxSTT {
6
+ constructor(config, logger) {
7
+ this.id = "local";
8
+ this.engine = config.engine;
9
+ this.silencePeakThreshold = config.silencePeakThreshold ?? 300;
10
+ this.tailPaddingMs = config.tailPaddingMs ?? 500;
11
+ this.logger = logger.child({ module: "speech", provider: "local", component: "stt" });
12
+ }
13
+ createSession(params) {
14
+ const emitter = new EventEmitter();
15
+ void params;
16
+ const requiredSampleRate = this.engine.sampleRate;
17
+ let connected = false;
18
+ let segmentId = uuidv4();
19
+ let previousSegmentId = null;
20
+ let pcm16 = Buffer.alloc(0);
21
+ return {
22
+ requiredSampleRate,
23
+ async connect() {
24
+ connected = true;
25
+ },
26
+ appendPcm16(chunk) {
27
+ if (!connected) {
28
+ emitter.emit("error", new Error("STT session not connected"));
29
+ return;
30
+ }
31
+ pcm16 = pcm16.length === 0 ? chunk : Buffer.concat([pcm16, chunk]);
32
+ },
33
+ commit: () => {
34
+ if (!connected) {
35
+ emitter.emit("error", new Error("STT session not connected"));
36
+ return;
37
+ }
38
+ const committedId = segmentId;
39
+ const prev = previousSegmentId;
40
+ emitter.emit("committed", { segmentId: committedId, previousSegmentId: prev });
41
+ void (async () => {
42
+ try {
43
+ const rt = await this.transcribeAudio(pcm16, `audio/pcm;rate=${requiredSampleRate}`);
44
+ emitter.emit("transcript", {
45
+ segmentId: committedId,
46
+ transcript: rt.text,
47
+ isFinal: true,
48
+ language: rt.language,
49
+ logprobs: rt.logprobs,
50
+ avgLogprob: rt.avgLogprob,
51
+ isLowConfidence: rt.isLowConfidence,
52
+ });
53
+ }
54
+ catch (err) {
55
+ emitter.emit("error", err);
56
+ }
57
+ finally {
58
+ previousSegmentId = committedId;
59
+ segmentId = uuidv4();
60
+ pcm16 = Buffer.alloc(0);
61
+ }
62
+ })();
63
+ },
64
+ clear() {
65
+ pcm16 = Buffer.alloc(0);
66
+ segmentId = uuidv4();
67
+ },
68
+ close() {
69
+ connected = false;
70
+ pcm16 = Buffer.alloc(0);
71
+ },
72
+ on(event, handler) {
73
+ emitter.on(event, handler);
74
+ return undefined;
75
+ },
76
+ };
77
+ }
78
+ async transcribeAudio(audioBuffer, format) {
79
+ const start = Date.now();
80
+ let inputRate;
81
+ let pcm16;
82
+ if (format.toLowerCase().includes("audio/wav")) {
83
+ const parsed = parsePcm16MonoWav(audioBuffer);
84
+ inputRate = parsed.sampleRate;
85
+ pcm16 = parsed.pcm16;
86
+ }
87
+ else if (format.toLowerCase().includes("audio/pcm")) {
88
+ inputRate = parsePcmRateFromFormat(format, this.engine.sampleRate) ?? this.engine.sampleRate;
89
+ pcm16 = audioBuffer;
90
+ }
91
+ else {
92
+ throw new Error(`Unsupported audio format for sherpa STT: ${format}`);
93
+ }
94
+ const peak = pcm16lePeakAbs(pcm16);
95
+ if (peak < this.silencePeakThreshold) {
96
+ return { text: "", duration: Date.now() - start, isLowConfidence: true };
97
+ }
98
+ let pcmForModel = pcm16;
99
+ if (inputRate !== this.engine.sampleRate) {
100
+ const resampler = new Pcm16MonoResampler({ inputRate, outputRate: this.engine.sampleRate });
101
+ pcmForModel = resampler.processChunk(pcm16);
102
+ inputRate = this.engine.sampleRate;
103
+ }
104
+ const peakForModel = pcm16lePeakAbs(pcmForModel);
105
+ const peakFloat = peakForModel / 32768.0;
106
+ const targetPeak = 0.6;
107
+ const maxGain = 50;
108
+ const gain = peakFloat > 0 && peakFloat < targetPeak
109
+ ? Math.min(maxGain, targetPeak / peakFloat)
110
+ : 1;
111
+ const stream = this.engine.createStream();
112
+ try {
113
+ const floatSamples = pcm16leToFloat32(pcmForModel, gain);
114
+ stream.acceptWaveform(inputRate, floatSamples);
115
+ while (this.engine.recognizer.isReady(stream)) {
116
+ this.engine.recognizer.decode(stream);
117
+ }
118
+ const padSamples = Math.max(0, Math.round((this.engine.sampleRate * this.tailPaddingMs) / 1000));
119
+ if (padSamples > 0) {
120
+ stream.acceptWaveform(this.engine.sampleRate, new Float32Array(padSamples));
121
+ }
122
+ while (this.engine.recognizer.isReady(stream)) {
123
+ this.engine.recognizer.decode(stream);
124
+ }
125
+ const text = String(this.engine.recognizer.getResult(stream)?.text ?? "").trim();
126
+ const duration = Date.now() - start;
127
+ this.logger.debug({ duration, textLength: text.length }, "Sherpa transcription complete");
128
+ return { text, duration, ...(text.length === 0 ? { isLowConfidence: true } : {}) };
129
+ }
130
+ finally {
131
+ try {
132
+ stream.free?.();
133
+ }
134
+ catch {
135
+ // ignore
136
+ }
137
+ }
138
+ }
139
+ }
140
+ //# sourceMappingURL=sherpa-stt.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"sherpa-stt.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,EAAE,IAAI,MAAM,EAAE,MAAM,MAAM,CAAC;AAQpC,OAAO,EAAE,kBAAkB,EAAE,MAAM,sCAAsC,CAAC;AAC1E,OAAO,EAAE,iBAAiB,EAAE,sBAAsB,EAAE,cAAc,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAShH,MAAM,OAAO,aAAa;IAOxB,YAAY,MAAuB,EAAE,MAAmB;QAFxC,OAAE,GAAG,OAAgB,CAAC;QAGpC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,oBAAoB,GAAG,MAAM,CAAC,oBAAoB,IAAI,GAAG,CAAC;QAC/D,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,GAAG,CAAC;QACjD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,EAAE,OAAO,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;IACxF,CAAC;IAEM,aAAa,CAAC,MAIpB;QACC,MAAM,OAAO,GAAG,IAAI,YAAY,EAAE,CAAC;QACnC,KAAK,MAAM,CAAC;QACZ,MAAM,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QAClD,IAAI,SAAS,GAAG,KAAK,CAAC;QACtB,IAAI,SAAS,GAAG,MAAM,EAAE,CAAC;QACzB,IAAI,iBAAiB,GAAkB,IAAI,CAAC;QAC5C,IAAI,KAAK,GAAW,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAEpC,OAAO;YACL,kBAAkB;YAClB,KAAK,CAAC,OAAO;gBACX,SAAS,GAAG,IAAI,CAAC;YACnB,CAAC;YACD,WAAW,CAAC,KAAa;gBACvB,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBACD,KAAK,GAAG,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC;YACrE,CAAC;YACD,MAAM,EAAE,GAAG,EAAE;gBACX,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBAED,MAAM,WAAW,GAAG,SAAS,CAAC;gBAC9B,MAAM,IAAI,GAAG,iBAAiB,CAAC;gBAC9B,OAAe,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,WAAW,EAAE,iBAAiB,EAAE,IAAI,EAAE,CAAC,CAAC;gBAExF,KAAK,CAAC,KAAK,IAAI,EAAE;oBACf,IAAI,CAAC;wBACH,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,eAAe,CAAC,KAAK,EAAE,kBAAkB,kBAAkB,EAAE,CAAC,CAAC;wBACpF,OAAe,CAAC,IAAI,CAAC,YAAY,EAAE;4BAClC,SAAS,EAAE,WAAW;4BACtB,UAAU,EAAE,EAAE,CAAC,IAAI;4BACnB,OAAO,EAAE,IAAI;4BACb,QAAQ,EAAE,EAAE,CAAC,QAAQ;4BACrB,QAAQ,EAAE,EAAE,CAAC,QAAQ;4BACrB,UAAU,EAAE,EAAE,CAAC,UAAU;4BACzB,eAAe,EAAE,EAAE,CAAC,eAAe;yBACpC,CAAC,CAAC;oBACL,CAAC;oBAAC,OAAO,GAAG,EAAE,CAAC;wBACZ,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;oBACtC,CAAC;4BAAS,CAAC;wBACT,iBAAiB,GAAG,WAAW,CAAC;wBAChC,SAAS,GAAG,MAAM,EAAE,CAAC;wBACrB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;oBAC1B,CAAC;gBACH,CAAC,CAAC,EAAE,CAAC;YACP,CAAC;YACD,KAAK;gBACH,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACxB,SAAS,GAAG,MAAM,EAAE,CAAC;YACvB,CAAC;YACD,KAAK;gBACH,SAAS,GAAG,KAAK,CAAC;gBAClB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC1B,CAAC;YACD,EAAE,CAAC,KAAU,EAAE,OAAY;gBACzB,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;gBAC3B,OAAO,SAAS,CAAC;YACnB,CAAC;SACF,CAAC;IACJ,CAAC;IAEM,KAAK,CAAC,eAAe,CAAC,WAAmB,EAAE,MAAc;QAC9D,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAEzB,IAAI,SAAiB,CAAC;QACtB,IAAI,KAAa,CAAC;QAElB,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;YAC/C,MAAM,MAAM,GAAG,iBAAiB,CAAC,WAAW,CAAC,CAAC;YAC9C,SAAS,GAAG,MAAM,CAAC,UAAU,CAAC;YAC9B,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC;QACvB,CAAC;aAAM,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;YACtD,SAAS,GAAG,sBAAsB,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,IAAI,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;YAC7F,KAAK,GAAG,WAAW,CAAC;QACtB,CAAC;aAAM,CAAC;YACN,MAAM,IAAI,KAAK,CAAC,4CAA4C,MAAM,EAAE,CAAC,CAAC;QACxE,CAAC;QAED,MAAM,IAAI,GAAG,cAAc,CAAC,KAAK,CAAC,CAAC;QACnC,IAAI,IAAI,GAAG,IAAI,CAAC,oBAAoB,EAAE,CAAC;YACrC,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,QAAQ,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC;QAC3E,CAAC;QAED,IAAI,WAAW,GAAG,KAAK,CAAC;QACxB,IAAI,SAAS,KAAK,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC;YACzC,MAAM,SAAS,GAAG,IAAI,kBAAkB,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,CAAC;YAC5F,WAAW,GAAG,SAAS,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;YAC5C,SAAS,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QACrC,CAAC;QAED,MAAM,YAAY,GAAG,cAAc,CAAC,WAAW,CAAC,CAAC;QACjD,MAAM,SAAS,GAAG,YAAY,GAAG,OAAO,CAAC;QACzC,MAAM,UAAU,GAAG,GAAG,CAAC;QACvB,MAAM,OAAO,GAAG,EAAE,CAAC;QACnB,MAAM,IAAI,GACR,SAAS,GAAG,CAAC,IAAI,SAAS,GAAG,UAAU;YACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,GAAG,SAAS,CAAC;YAC3C,CAAC,CAAC,CAAC,CAAC;QAER,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;QAC1C,IAAI,CAAC;YACH,MAAM,YAAY,GAAG,gBAAgB,CAAC,WAAW,EAAE,IAAI,CAAC,CAAC;YACzD,MAAM,CAAC,cAAc,CAAC,SAAS,EAAE,YAAY,CAAC,CAAC;YAE/C,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC9C,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACxC,CAAC;YAED,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC;YACjG,IAAI,UAAU,GAAG,CAAC,EAAE,CAAC;gBACnB,MAAM,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,IAAI,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC;YAC9E,CAAC;YAED,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC9C,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACxC,CAAC;YAED,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,MAAM,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YACjF,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC;YACpC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,EAAE,EAAE,+BAA+B,CAAC,CAAC;YAC1F,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,GAAG,CAAC,IAAI,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC;QACrF,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC;gBACH,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;YAClB,CAAC;YAAC,MAAM,CAAC;gBACP,SAAS;YACX,CAAC;QACH,CAAC;IACH,CAAC;CACF"}
@@ -0,0 +1,95 @@
1
+ import { Readable } from "node:stream";
2
+ import { existsSync } from "node:fs";
3
+ import { chunkBuffer, float32ToPcm16le } from "../../../audio.js";
4
+ import { loadSherpaOnnxNode } from "./sherpa-onnx-node-loader.js";
5
+ function assertFileExists(filePath, label) {
6
+ if (!existsSync(filePath)) {
7
+ throw new Error(`Missing ${label}: ${filePath}`);
8
+ }
9
+ }
10
+ export class SherpaOnnxTTS {
11
+ constructor(config, logger) {
12
+ if (config.preset !== "kokoro-en-v0_19" && config.preset !== "kitten-nano-en-v0_1-fp16") {
13
+ throw new Error(`Unsupported Sherpa TTS preset: ${config.preset}`);
14
+ }
15
+ this.logger = logger.child({ module: "speech", provider: "local", component: "tts" });
16
+ this.speakerId = config.speakerId ?? 0;
17
+ this.speed = config.speed ?? 1.0;
18
+ const sherpa = loadSherpaOnnxNode();
19
+ if (typeof sherpa.OfflineTts !== "function") {
20
+ throw new Error("sherpa-onnx-node OfflineTts is unavailable");
21
+ }
22
+ const modelFile = config.preset === "kokoro-en-v0_19" ? "model.onnx" : "model.fp16.onnx";
23
+ const modelPath = `${config.modelDir}/${modelFile}`;
24
+ const voicesPath = `${config.modelDir}/voices.bin`;
25
+ const tokensPath = `${config.modelDir}/tokens.txt`;
26
+ const dataDir = `${config.modelDir}/espeak-ng-data`;
27
+ assertFileExists(modelPath, "TTS model");
28
+ assertFileExists(voicesPath, "TTS voices");
29
+ assertFileExists(tokensPath, "TTS tokens");
30
+ assertFileExists(dataDir, "TTS espeak-ng dataDir");
31
+ const modelConfig = config.preset === "kokoro-en-v0_19"
32
+ ? {
33
+ kokoro: {
34
+ model: modelPath,
35
+ voices: voicesPath,
36
+ tokens: tokensPath,
37
+ dataDir,
38
+ lengthScale: config.lengthScale ?? 1.0,
39
+ },
40
+ }
41
+ : {
42
+ kitten: {
43
+ model: modelPath,
44
+ voices: voicesPath,
45
+ tokens: tokensPath,
46
+ dataDir,
47
+ lengthScale: config.lengthScale ?? 1.0,
48
+ },
49
+ };
50
+ const offlineTtsConfig = {
51
+ model: modelConfig,
52
+ numThreads: config.numThreads ?? 2,
53
+ provider: "cpu",
54
+ maxNumSentences: 1,
55
+ };
56
+ this.tts = new sherpa.OfflineTts(offlineTtsConfig);
57
+ this.logger.info({ preset: config.preset, modelDir: config.modelDir }, "Sherpa offline TTS initialized");
58
+ }
59
+ async synthesizeSpeech(text) {
60
+ const trimmed = text.trim();
61
+ if (!trimmed) {
62
+ throw new Error("Cannot synthesize empty text");
63
+ }
64
+ const audio = this.tts.generate({ text: trimmed, sid: this.speakerId, speed: this.speed });
65
+ const samples = audio && audio.samples instanceof Float32Array
66
+ ? audio.samples
67
+ : audio && Array.isArray(audio.samples)
68
+ ? Float32Array.from(audio.samples)
69
+ : null;
70
+ const sampleRate = audio && typeof audio.sampleRate === "number" && Number.isFinite(audio.sampleRate) && audio.sampleRate > 0
71
+ ? audio.sampleRate
72
+ : typeof this.tts.sampleRate === "number"
73
+ ? this.tts.sampleRate
74
+ : 24000;
75
+ if (!samples) {
76
+ throw new Error("Unexpected sherpa TTS output: missing Float32 samples");
77
+ }
78
+ const pcm16 = float32ToPcm16le(samples);
79
+ const chunkBytes = Math.max(2, Math.round(sampleRate * 0.05) * 2); // ~50ms
80
+ const chunks = chunkBuffer(pcm16, chunkBytes);
81
+ return {
82
+ stream: Readable.from(chunks),
83
+ format: `pcm;rate=${sampleRate}`,
84
+ };
85
+ }
86
+ free() {
87
+ try {
88
+ this.tts?.free?.();
89
+ }
90
+ catch {
91
+ // ignore
92
+ }
93
+ }
94
+ }
95
+ //# sourceMappingURL=sherpa-tts.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"sherpa-tts.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-tts.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvC,OAAO,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AAGrC,OAAO,EAAE,WAAW,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAClE,OAAO,EAAE,kBAAkB,EAAE,MAAM,8BAA8B,CAAC;AAalE,SAAS,gBAAgB,CAAC,QAAgB,EAAE,KAAa;IACvD,IAAI,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,WAAW,KAAK,KAAK,QAAQ,EAAE,CAAC,CAAC;IACnD,CAAC;AACH,CAAC;AAED,MAAM,OAAO,aAAa;IAMxB,YAAY,MAAuB,EAAE,MAAmB;QACtD,IAAI,MAAM,CAAC,MAAM,KAAK,iBAAiB,IAAI,MAAM,CAAC,MAAM,KAAK,0BAA0B,EAAE,CAAC;YACxF,MAAM,IAAI,KAAK,CAAC,kCAAkC,MAAM,CAAC,MAAM,EAAE,CAAC,CAAC;QACrE,CAAC;QACD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,EAAE,OAAO,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;QACtF,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,CAAC,CAAC;QACvC,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,GAAG,CAAC;QAEjC,MAAM,MAAM,GAAG,kBAAkB,EAAE,CAAC;QACpC,IAAI,OAAO,MAAM,CAAC,UAAU,KAAK,UAAU,EAAE,CAAC;YAC5C,MAAM,IAAI,KAAK,CAAC,4CAA4C,CAAC,CAAC;QAChE,CAAC;QAED,MAAM,SAAS,GAAG,MAAM,CAAC,MAAM,KAAK,iBAAiB,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,iBAAiB,CAAC;QACzF,MAAM,SAAS,GAAG,GAAG,MAAM,CAAC,QAAQ,IAAI,SAAS,EAAE,CAAC;QACpD,MAAM,UAAU,GAAG,GAAG,MAAM,CAAC,QAAQ,aAAa,CAAC;QACnD,MAAM,UAAU,GAAG,GAAG,MAAM,CAAC,QAAQ,aAAa,CAAC;QACnD,MAAM,OAAO,GAAG,GAAG,MAAM,CAAC,QAAQ,iBAAiB,CAAC;QAEpD,gBAAgB,CAAC,SAAS,EAAE,WAAW,CAAC,CAAC;QACzC,gBAAgB,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;QAC3C,gBAAgB,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;QAC3C,gBAAgB,CAAC,OAAO,EAAE,uBAAuB,CAAC,CAAC;QAEnD,MAAM,WAAW,GACf,MAAM,CAAC,MAAM,KAAK,iBAAiB;YACjC,CAAC,CAAC;gBACE,MAAM,EAAE;oBACN,KAAK,EAAE,SAAS;oBAChB,MAAM,EAAE,UAAU;oBAClB,MAAM,EAAE,UAAU;oBAClB,OAAO;oBACP,WAAW,EAAE,MAAM,CAAC,WAAW,IAAI,GAAG;iBACvC;aACF;YACH,CAAC,CAAC;gBACE,MAAM,EAAE;oBACN,KAAK,EAAE,SAAS;oBAChB,MAAM,EAAE,UAAU;oBAClB,MAAM,EAAE,UAAU;oBAClB,OAAO;oBACP,WAAW,EAAE,MAAM,CAAC,WAAW,IAAI,GAAG;iBACvC;aACF,CAAC;QAER,MAAM,gBAAgB,GAAG;YACvB,KAAK,EAAE,WAAW;YAClB,UAAU,EAAE,MAAM,CAAC,UAAU,IAAI,CAAC;YAClC,QAAQ,EAAE,KAAK;YACf,eAAe,EAAE,CAAC;SACnB,CAAC;QAEF,IAAI,CAAC,GAAG,GAAG,IAAI,MAAM,CAAC,UAAU,CAAC,gBAAgB,CAAC,CAAC;QACnD,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,EAAE,MAAM,CAAC,QAAQ,EAAE,EAAE,gCAAgC,CAAC,CAAC;IAC3G,CAAC;IAED,KAAK,CAAC,gBAAgB,CAAC,IAAY;QACjC,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC;QAC5B,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAC;QAClD,CAAC;QAED,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,GAAG,EAAE,IAAI,CAAC,SAAS,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,CAAC,CAAC;QAC3F,MAAM,OAAO,GACX,KAAK,IAAI,KAAK,CAAC,OAAO,YAAY,YAAY;YAC5C,CAAC,CAAC,KAAK,CAAC,OAAO;YACf,CAAC,CAAC,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,OAAO,CAAC;gBACrC,CAAC,CAAC,YAAY,CAAC,IAAI,CAAC,KAAK,CAAC,OAAmB,CAAC;gBAC9C,CAAC,CAAC,IAAI,CAAC;QACb,MAAM,UAAU,GACd,KAAK,IAAI,OAAO,KAAK,CAAC,UAAU,KAAK,QAAQ,IAAI,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,UAAU,CAAC,IAAI,KAAK,CAAC,UAAU,GAAG,CAAC;YACxG,CAAC,CAAC,KAAK,CAAC,UAAU;YAClB,CAAC,CAAC,OAAO,IAAI,CAAC,GAAG,CAAC,UAAU,KAAK,QAAQ;gBACvC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,UAAU;gBACrB,CAAC,CAAC,KAAK,CAAC;QAEd,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,IAAI,KAAK,CAAC,uDAAuD,CAAC,CAAC;QAC3E,CAAC;QAED,MAAM,KAAK,GAAG,gBAAgB,CAAC,OAAO,CAAC,CAAC;QACxC,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ;QAC3E,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC;QAE9C,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,IAAI,CAAC,MAAM,CAAC;YAC7B,MAAM,EAAE,YAAY,UAAU,EAAE;SACjC,CAAC;IACJ,CAAC;IAED,IAAI;QACF,IAAI,CAAC;YACH,IAAI,CAAC,GAAG,EAAE,IAAI,EAAE,EAAE,CAAC;QACrB,CAAC;QAAC,MAAM,CAAC;YACP,SAAS;QACX,CAAC;IACH,CAAC;CACF"}
@@ -0,0 +1,99 @@
1
+ import { z } from "zod";
2
+ export const DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL = "gpt-4o-transcribe";
3
+ export const DEFAULT_OPENAI_TTS_MODEL = "tts-1";
4
+ const OpenAiTtsVoiceSchema = z.enum([
5
+ "alloy",
6
+ "echo",
7
+ "fable",
8
+ "onyx",
9
+ "nova",
10
+ "shimmer",
11
+ ]);
12
+ const OpenAiTtsModelSchema = z.enum(["tts-1", "tts-1-hd"]);
13
+ const NumberLikeSchema = z.union([
14
+ z.number(),
15
+ z.string().trim().min(1),
16
+ ]);
17
+ const OptionalFiniteNumberSchema = NumberLikeSchema
18
+ .pipe(z.coerce.number().finite())
19
+ .optional();
20
+ const OptionalTrimmedStringSchema = z
21
+ .string()
22
+ .trim()
23
+ .optional()
24
+ .transform((value) => (value && value.length > 0 ? value : undefined));
25
+ const OpenAiSpeechResolutionSchema = z.object({
26
+ apiKey: OptionalTrimmedStringSchema,
27
+ sttConfidenceThreshold: OptionalFiniteNumberSchema,
28
+ sttModel: OptionalTrimmedStringSchema,
29
+ ttsVoice: z
30
+ .string()
31
+ .trim()
32
+ .toLowerCase()
33
+ .pipe(OpenAiTtsVoiceSchema)
34
+ .default("alloy"),
35
+ ttsModel: z
36
+ .string()
37
+ .trim()
38
+ .toLowerCase()
39
+ .pipe(OpenAiTtsModelSchema)
40
+ .default(DEFAULT_OPENAI_TTS_MODEL),
41
+ realtimeTranscriptionModel: OptionalTrimmedStringSchema.default(DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL),
42
+ });
43
+ export function resolveOpenAiSpeechConfig(params) {
44
+ const parsed = OpenAiSpeechResolutionSchema.parse({
45
+ apiKey: params.env.OPENAI_API_KEY ?? params.persisted.providers?.openai?.apiKey,
46
+ sttConfidenceThreshold: params.env.STT_CONFIDENCE_THRESHOLD ??
47
+ params.persisted.features?.dictation?.stt?.confidenceThreshold,
48
+ sttModel: params.env.STT_MODEL ??
49
+ (params.providers.voiceStt.enabled !== false &&
50
+ params.providers.voiceStt.provider === "openai"
51
+ ? params.persisted.features?.voiceMode?.stt?.model
52
+ : undefined) ??
53
+ (params.providers.dictationStt.enabled !== false &&
54
+ params.providers.dictationStt.provider === "openai"
55
+ ? params.persisted.features?.dictation?.stt?.model
56
+ : undefined),
57
+ ttsVoice: params.env.TTS_VOICE ??
58
+ (params.providers.voiceTts.enabled !== false &&
59
+ params.providers.voiceTts.provider === "openai"
60
+ ? params.persisted.features?.voiceMode?.tts?.voice
61
+ : undefined) ??
62
+ "alloy",
63
+ ttsModel: params.env.TTS_MODEL ??
64
+ (params.providers.voiceTts.enabled !== false &&
65
+ params.providers.voiceTts.provider === "openai"
66
+ ? params.persisted.features?.voiceMode?.tts?.model
67
+ : undefined) ??
68
+ DEFAULT_OPENAI_TTS_MODEL,
69
+ realtimeTranscriptionModel: params.env.OPENAI_REALTIME_TRANSCRIPTION_MODEL ??
70
+ (params.providers.dictationStt.enabled !== false &&
71
+ params.providers.dictationStt.provider === "openai"
72
+ ? params.persisted.features?.dictation?.stt?.model
73
+ : undefined) ??
74
+ DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL,
75
+ });
76
+ if (!parsed.apiKey) {
77
+ return undefined;
78
+ }
79
+ return {
80
+ apiKey: parsed.apiKey,
81
+ stt: {
82
+ apiKey: parsed.apiKey,
83
+ ...(parsed.sttConfidenceThreshold !== undefined
84
+ ? { confidenceThreshold: parsed.sttConfidenceThreshold }
85
+ : {}),
86
+ ...(parsed.sttModel
87
+ ? { model: parsed.sttModel }
88
+ : {}),
89
+ },
90
+ tts: {
91
+ apiKey: parsed.apiKey,
92
+ voice: parsed.ttsVoice,
93
+ model: parsed.ttsModel,
94
+ responseFormat: "pcm",
95
+ },
96
+ realtimeTranscriptionModel: parsed.realtimeTranscriptionModel,
97
+ };
98
+ }
99
+ //# sourceMappingURL=config.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"config.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/config.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAOxB,MAAM,CAAC,MAAM,2CAA2C,GAAG,mBAAmB,CAAC;AAC/E,MAAM,CAAC,MAAM,wBAAwB,GAAG,OAAO,CAAC;AAShD,MAAM,oBAAoB,GAAG,CAAC,CAAC,IAAI,CAAC;IAClC,OAAO;IACP,MAAM;IACN,OAAO;IACP,MAAM;IACN,MAAM;IACN,SAAS;CACV,CAAC,CAAC;AAEH,MAAM,oBAAoB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,OAAO,EAAE,UAAU,CAAC,CAAC,CAAC;AAE3D,MAAM,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC;IAC/B,CAAC,CAAC,MAAM,EAAE;IACV,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;CACzB,CAAC,CAAC;AAEH,MAAM,0BAA0B,GAAG,gBAAgB;KAChD,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,EAAE,CAAC,MAAM,EAAE,CAAC;KAChC,QAAQ,EAAE,CAAC;AAEd,MAAM,2BAA2B,GAAG,CAAC;KAClC,MAAM,EAAE;KACR,IAAI,EAAE;KACN,QAAQ,EAAE;KACV,SAAS,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC;AAEzE,MAAM,4BAA4B,GAAG,CAAC,CAAC,MAAM,CAAC;IAC5C,MAAM,EAAE,2BAA2B;IACnC,sBAAsB,EAAE,0BAA0B;IAClD,QAAQ,EAAE,2BAA2B;IACrC,QAAQ,EAAE,CAAC;SACR,MAAM,EAAE;SACR,IAAI,EAAE;SACN,WAAW,EAAE;SACb,IAAI,CAAC,oBAAoB,CAAC;SAC1B,OAAO,CAAC,OAAO,CAAC;IACnB,QAAQ,EAAE,CAAC;SACR,MAAM,EAAE;SACR,IAAI,EAAE;SACN,WAAW,EAAE;SACb,IAAI,CAAC,oBAAoB,CAAC;SAC1B,OAAO,CAAC,wBAAwB,CAAC;IACpC,0BAA0B,EAAE,2BAA2B,CAAC,OAAO,CAC7D,2CAA2C,CAC5C;CACF,CAAC,CAAC;AAEH,MAAM,UAAU,yBAAyB,CAAC,MAIzC;IACC,MAAM,MAAM,GAAG,4BAA4B,CAAC,KAAK,CAAC;QAChD,MAAM,EAAE,MAAM,CAAC,GAAG,CAAC,cAAc,IAAI,MAAM,CAAC,SAAS,CAAC,SAAS,EAAE,MAAM,EAAE,MAAM;QAC/E,sBAAsB,EACpB,MAAM,CAAC,GAAG,CAAC,wBAAwB;YACnC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,mBAAmB;QAChE,QAAQ,EACN,MAAM,CAAC,GAAG,CAAC,SAAS;YACpB,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;gBAC5C,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;gBAC7C,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,CAAC,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,OAAO,KAAK,KAAK;gBAChD,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ;gBACjD,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;QAChB,QAAQ,EACN,MAAM,CAAC,GAAG,CAAC,SAAS;YACpB,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;gBAC5C,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;gBAC7C,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,OAAO;QACT,QAAQ,EACN,MAAM,CAAC,GAAG,CAAC,SAAS;YACpB,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;gBAC5C,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;gBAC7C,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,wBAAwB;QAC1B,0BAA0B,EACxB,MAAM,CAAC,GAAG,CAAC,mCAAmC;YAC9C,CAAC,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,OAAO,KAAK,KAAK;gBAChD,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ;gBACjD,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,2CAA2C;KAC9C,CAAC,CAAC;IAEH,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE,CAAC;QACnB,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,OAAO;QACL,MAAM,EAAE,MAAM,CAAC,MAAM;QACrB,GAAG,EAAE;YACH,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,GAAG,CAAC,MAAM,CAAC,sBAAsB,KAAK,SAAS;gBAC7C,CAAC,CAAC,EAAE,mBAAmB,EAAE,MAAM,CAAC,sBAAsB,EAAE;gBACxD,CAAC,CAAC,EAAE,CAAC;YACP,GAAG,CAAC,MAAM,CAAC,QAAQ;gBACjB,CAAC,CAAC,EAAE,KAAK,EAAE,MAAM,CAAC,QAAQ,EAAE;gBAC5B,CAAC,CAAC,EAAE,CAAC;SACR;QACD,GAAG,EAAE;YACH,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,KAAK,EAAE,MAAM,CAAC,QAAQ;YACtB,KAAK,EAAE,MAAM,CAAC,QAAQ;YACtB,cAAc,EAAE,KAAK;SACtB;QACD,0BAA0B,EAAE,MAAM,CAAC,0BAA0B;KAC9D,CAAC;AACJ,CAAC"}
@@ -0,0 +1,165 @@
1
+ import WebSocket from "ws";
2
+ import { EventEmitter } from "node:events";
3
+ export class OpenAIRealtimeTranscriptionSession extends EventEmitter {
4
+ constructor(params) {
5
+ super();
6
+ this.requiredSampleRate = 24000;
7
+ this.ws = null;
8
+ this.ready = null;
9
+ this.closing = false;
10
+ this.partialByItemId = new Map();
11
+ this.apiKey = params.apiKey;
12
+ this.logger = params.logger.child({ provider: "openai", component: "realtime-transcription" });
13
+ this.transcriptionModel = params.transcriptionModel;
14
+ this.language = params.language;
15
+ this.prompt = params.prompt;
16
+ this.turnDetection = params.turnDetection ?? null;
17
+ }
18
+ async connect() {
19
+ if (this.ready) {
20
+ return this.ready;
21
+ }
22
+ this.closing = false;
23
+ this.ready = new Promise((resolve, reject) => {
24
+ const url = "wss://api.openai.com/v1/realtime?intent=transcription";
25
+ const ws = new WebSocket(url, {
26
+ headers: {
27
+ Authorization: `Bearer ${this.apiKey}`,
28
+ },
29
+ });
30
+ this.ws = ws;
31
+ let resolved = false;
32
+ const fail = (error) => {
33
+ if (resolved) {
34
+ this.emit("error", error);
35
+ return;
36
+ }
37
+ resolved = true;
38
+ reject(error);
39
+ };
40
+ ws.on("open", () => {
41
+ this.logger.debug("OpenAI realtime transcription websocket connected");
42
+ const update = {
43
+ type: "session.update",
44
+ session: {
45
+ type: "transcription",
46
+ audio: {
47
+ input: {
48
+ format: { type: "audio/pcm", rate: 24000 },
49
+ transcription: {
50
+ model: this.transcriptionModel,
51
+ ...(this.language ? { language: this.language } : {}),
52
+ ...(this.prompt ? { prompt: this.prompt } : {}),
53
+ },
54
+ turn_detection: this.turnDetection,
55
+ },
56
+ },
57
+ },
58
+ };
59
+ ws.send(JSON.stringify(update));
60
+ });
61
+ ws.on("message", (data) => {
62
+ const text = typeof data === "string" ? data : data.toString("utf-8");
63
+ let parsed;
64
+ try {
65
+ parsed = JSON.parse(text);
66
+ }
67
+ catch {
68
+ return;
69
+ }
70
+ const event = parsed;
71
+ if (event.type === "session.created" ||
72
+ event.type === "session.updated") {
73
+ if (!resolved) {
74
+ resolved = true;
75
+ resolve();
76
+ }
77
+ return;
78
+ }
79
+ if (event.type === "input_audio_buffer.committed") {
80
+ this.emit("committed", {
81
+ segmentId: event.item_id,
82
+ previousSegmentId: event.previous_item_id,
83
+ });
84
+ return;
85
+ }
86
+ if (event.type === "input_audio_buffer.speech_started") {
87
+ this.emit("speech_started");
88
+ return;
89
+ }
90
+ if (event.type === "input_audio_buffer.speech_stopped") {
91
+ this.emit("speech_stopped");
92
+ return;
93
+ }
94
+ if (event.type === "conversation.item.input_audio_transcription.delta") {
95
+ const replaceDelta = this.transcriptionModel === "whisper-1";
96
+ const prev = this.partialByItemId.get(event.item_id) ?? "";
97
+ const next = replaceDelta ? event.delta : prev + event.delta;
98
+ this.partialByItemId.set(event.item_id, next);
99
+ this.emit("transcript", { segmentId: event.item_id, transcript: next, isFinal: false });
100
+ return;
101
+ }
102
+ if (event.type === "conversation.item.input_audio_transcription.completed") {
103
+ this.partialByItemId.set(event.item_id, event.transcript);
104
+ this.emit("transcript", { segmentId: event.item_id, transcript: event.transcript, isFinal: true });
105
+ return;
106
+ }
107
+ if (event.type === "error") {
108
+ const message = event.error?.message ?? "OpenAI realtime error";
109
+ fail(new Error(message));
110
+ }
111
+ });
112
+ ws.on("error", (err) => {
113
+ fail(err instanceof Error ? err : new Error(String(err)));
114
+ });
115
+ ws.on("close", () => {
116
+ this.logger.debug("OpenAI realtime websocket closed");
117
+ if (this.closing) {
118
+ return;
119
+ }
120
+ if (!resolved) {
121
+ fail(new Error("OpenAI realtime websocket closed before ready"));
122
+ return;
123
+ }
124
+ fail(new Error("OpenAI realtime websocket closed"));
125
+ });
126
+ });
127
+ return this.ready;
128
+ }
129
+ appendPcm16(pcm16le) {
130
+ if (!this.ws || this.ws.readyState !== WebSocket.OPEN) {
131
+ throw new Error("OpenAI realtime websocket not connected");
132
+ }
133
+ const base64Audio = pcm16le.toString("base64");
134
+ const event = { type: "input_audio_buffer.append", audio: base64Audio };
135
+ this.ws.send(JSON.stringify(event));
136
+ }
137
+ commit() {
138
+ if (!this.ws || this.ws.readyState !== WebSocket.OPEN) {
139
+ throw new Error("OpenAI realtime websocket not connected");
140
+ }
141
+ const event = { type: "input_audio_buffer.commit" };
142
+ this.ws.send(JSON.stringify(event));
143
+ }
144
+ clear() {
145
+ if (!this.ws || this.ws.readyState !== WebSocket.OPEN) {
146
+ return;
147
+ }
148
+ const event = { type: "input_audio_buffer.clear" };
149
+ this.ws.send(JSON.stringify(event));
150
+ }
151
+ close() {
152
+ try {
153
+ this.closing = true;
154
+ this.ws?.close();
155
+ }
156
+ catch {
157
+ // no-op
158
+ }
159
+ finally {
160
+ this.ws = null;
161
+ this.ready = null;
162
+ }
163
+ }
164
+ }
165
+ //# sourceMappingURL=realtime-transcription-session.js.map