@livekit/agents 1.0.2 → 1.0.4

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 (118) hide show
  1. package/dist/index.cjs +2 -5
  2. package/dist/index.cjs.map +1 -1
  3. package/dist/index.d.cts +2 -3
  4. package/dist/index.d.ts +2 -3
  5. package/dist/index.d.ts.map +1 -1
  6. package/dist/index.js +1 -3
  7. package/dist/index.js.map +1 -1
  8. package/dist/ipc/job_proc_lazy_main.cjs +3 -2
  9. package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
  10. package/dist/ipc/job_proc_lazy_main.js +4 -3
  11. package/dist/ipc/job_proc_lazy_main.js.map +1 -1
  12. package/dist/job.cjs +20 -14
  13. package/dist/job.cjs.map +1 -1
  14. package/dist/job.d.cts +11 -5
  15. package/dist/job.d.ts +11 -5
  16. package/dist/job.d.ts.map +1 -1
  17. package/dist/job.js +17 -12
  18. package/dist/job.js.map +1 -1
  19. package/dist/tokenize/basic/hyphenator.cjs.map +1 -1
  20. package/dist/tokenize/basic/hyphenator.js.map +1 -1
  21. package/dist/utils.cjs +77 -0
  22. package/dist/utils.cjs.map +1 -1
  23. package/dist/utils.d.cts +21 -0
  24. package/dist/utils.d.ts +21 -0
  25. package/dist/utils.d.ts.map +1 -1
  26. package/dist/utils.js +76 -1
  27. package/dist/utils.js.map +1 -1
  28. package/dist/voice/agent_activity.cjs +112 -71
  29. package/dist/voice/agent_activity.cjs.map +1 -1
  30. package/dist/voice/agent_activity.d.ts.map +1 -1
  31. package/dist/voice/agent_activity.js +112 -71
  32. package/dist/voice/agent_activity.js.map +1 -1
  33. package/dist/voice/agent_session.cjs +9 -2
  34. package/dist/voice/agent_session.cjs.map +1 -1
  35. package/dist/voice/agent_session.d.ts.map +1 -1
  36. package/dist/voice/agent_session.js +9 -2
  37. package/dist/voice/agent_session.js.map +1 -1
  38. package/dist/voice/avatar/datastream_io.cjs +204 -0
  39. package/dist/voice/avatar/datastream_io.cjs.map +1 -0
  40. package/dist/voice/avatar/datastream_io.d.cts +37 -0
  41. package/dist/voice/avatar/datastream_io.d.ts +37 -0
  42. package/dist/voice/avatar/datastream_io.d.ts.map +1 -0
  43. package/dist/voice/avatar/datastream_io.js +188 -0
  44. package/dist/voice/avatar/datastream_io.js.map +1 -0
  45. package/dist/{multimodal → voice/avatar}/index.cjs +4 -4
  46. package/dist/voice/avatar/index.cjs.map +1 -0
  47. package/dist/voice/avatar/index.d.cts +2 -0
  48. package/dist/voice/avatar/index.d.ts +2 -0
  49. package/dist/voice/avatar/index.d.ts.map +1 -0
  50. package/dist/voice/avatar/index.js +2 -0
  51. package/dist/voice/avatar/index.js.map +1 -0
  52. package/dist/voice/index.cjs +2 -0
  53. package/dist/voice/index.cjs.map +1 -1
  54. package/dist/voice/index.d.cts +1 -0
  55. package/dist/voice/index.d.ts +1 -0
  56. package/dist/voice/index.d.ts.map +1 -1
  57. package/dist/voice/index.js +1 -0
  58. package/dist/voice/index.js.map +1 -1
  59. package/dist/voice/io.cjs.map +1 -1
  60. package/dist/voice/io.d.cts +1 -1
  61. package/dist/voice/io.d.ts +1 -1
  62. package/dist/voice/io.d.ts.map +1 -1
  63. package/dist/voice/io.js.map +1 -1
  64. package/dist/voice/room_io/_input.cjs +2 -1
  65. package/dist/voice/room_io/_input.cjs.map +1 -1
  66. package/dist/voice/room_io/_input.d.ts.map +1 -1
  67. package/dist/voice/room_io/_input.js +2 -1
  68. package/dist/voice/room_io/_input.js.map +1 -1
  69. package/dist/voice/run_context.cjs +13 -0
  70. package/dist/voice/run_context.cjs.map +1 -1
  71. package/dist/voice/run_context.d.cts +10 -0
  72. package/dist/voice/run_context.d.ts +10 -0
  73. package/dist/voice/run_context.d.ts.map +1 -1
  74. package/dist/voice/run_context.js +13 -0
  75. package/dist/voice/run_context.js.map +1 -1
  76. package/dist/voice/speech_handle.cjs +152 -30
  77. package/dist/voice/speech_handle.cjs.map +1 -1
  78. package/dist/voice/speech_handle.d.cts +67 -16
  79. package/dist/voice/speech_handle.d.ts +67 -16
  80. package/dist/voice/speech_handle.d.ts.map +1 -1
  81. package/dist/voice/speech_handle.js +153 -31
  82. package/dist/voice/speech_handle.js.map +1 -1
  83. package/dist/worker.cjs +4 -1
  84. package/dist/worker.cjs.map +1 -1
  85. package/dist/worker.d.ts.map +1 -1
  86. package/dist/worker.js +4 -1
  87. package/dist/worker.js.map +1 -1
  88. package/package.json +2 -2
  89. package/src/index.ts +2 -3
  90. package/src/ipc/job_proc_lazy_main.ts +6 -3
  91. package/src/job.ts +27 -12
  92. package/src/tokenize/basic/hyphenator.ts +1 -1
  93. package/src/utils.ts +121 -1
  94. package/src/voice/agent_activity.ts +128 -78
  95. package/src/voice/agent_session.ts +11 -2
  96. package/src/voice/avatar/datastream_io.ts +247 -0
  97. package/src/voice/avatar/index.ts +4 -0
  98. package/src/voice/index.ts +2 -0
  99. package/src/voice/io.ts +1 -1
  100. package/src/voice/room_io/_input.ts +8 -3
  101. package/src/voice/run_context.ts +16 -2
  102. package/src/voice/speech_handle.ts +183 -38
  103. package/src/worker.ts +5 -1
  104. package/dist/multimodal/agent_playout.cjs +0 -233
  105. package/dist/multimodal/agent_playout.cjs.map +0 -1
  106. package/dist/multimodal/agent_playout.d.cts +0 -34
  107. package/dist/multimodal/agent_playout.d.ts +0 -34
  108. package/dist/multimodal/agent_playout.d.ts.map +0 -1
  109. package/dist/multimodal/agent_playout.js +0 -207
  110. package/dist/multimodal/agent_playout.js.map +0 -1
  111. package/dist/multimodal/index.cjs.map +0 -1
  112. package/dist/multimodal/index.d.cts +0 -2
  113. package/dist/multimodal/index.d.ts +0 -2
  114. package/dist/multimodal/index.d.ts.map +0 -1
  115. package/dist/multimodal/index.js +0 -2
  116. package/dist/multimodal/index.js.map +0 -1
  117. package/src/multimodal/agent_playout.ts +0 -266
  118. package/src/multimodal/index.ts +0 -4
@@ -1 +1 @@
1
- {"version":3,"file":"agent_session.d.ts","sourceRoot":"","sources":["../../src/voice/agent_session.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAC1D,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,KAAK,EAAE,GAAG,EAAE,aAAa,EAAE,kBAAkB,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC1F,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAE9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACnD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AAExC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AAC5D,OAAO,EACL,sBAAsB,EACtB,KAAK,UAAU,EACf,KAAK,sBAAsB,EAC3B,KAAK,UAAU,EACf,WAAW,EACX,KAAK,0BAA0B,EAC/B,KAAK,UAAU,EACf,KAAK,0BAA0B,EAC/B,KAAK,qBAAqB,EAC1B,KAAK,kBAAkB,EACvB,KAAK,yBAAyB,EAC9B,KAAK,SAAS,EACd,KAAK,qBAAqB,EAK3B,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAClD,OAAO,EAAU,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC3F,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,WAAW,YAAY;IAC3B,kBAAkB,EAAE,OAAO,CAAC;IAC5B,6BAA6B,EAAE,OAAO,CAAC;IACvC,uBAAuB,EAAE,MAAM,CAAC;IAChC,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,YAAY,EAAE,MAAM,CAAC;CACtB;AAYD,MAAM,MAAM,iBAAiB,GAAG,KAAK,GAAG,KAAK,GAAG,cAAc,GAAG,QAAQ,GAAG,aAAa,CAAC;AAE1F,MAAM,MAAM,qBAAqB,GAAG;IAClC,CAAC,sBAAsB,CAAC,oBAAoB,CAAC,EAAE,CAAC,EAAE,EAAE,yBAAyB,KAAK,IAAI,CAAC;IACvF,CAAC,sBAAsB,CAAC,iBAAiB,CAAC,EAAE,CAAC,EAAE,EAAE,sBAAsB,KAAK,IAAI,CAAC;IACjF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,aAAa,CAAC,EAAE,CAAC,EAAE,EAAE,kBAAkB,KAAK,IAAI,CAAC;IACzE,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;IACzD,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;CAC1D,CAAC;AAEF,MAAM,MAAM,mBAAmB,CAAC,QAAQ,GAAG,eAAe,IAAI;IAC5D,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC,YAAY,CAAC,CAAC;CACtC,CAAC;2CAImC,aAAa,qBAAqB,CAAC;AAFxE,qBAAa,YAAY,CACvB,QAAQ,GAAG,eAAe,CAC1B,SAAQ,iBAA+D;IACvE,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAElC,QAAQ,CAAC,OAAO,EAAE,YAAY,CAAC;IAE/B,OAAO,CAAC,KAAK,CAAC,CAAQ;IACtB,OAAO,CAAC,QAAQ,CAAC,CAAgB;IACjC,OAAO,CAAC,YAAY,CAAC,CAAgB;IACrC,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,MAAM,CAAC,CAAS;IACxB,OAAO,CAAC,MAAM,CAAS;IAEvB,OAAO,CAAC,QAAQ,CAAc;IAC9B,OAAO,CAAC,SAAS,CAAuB;IACxC,OAAO,CAAC,WAAW,CAA8B;IAEjD,OAAO,CAAC,MAAM,CAAa;IAC3B,OAAO,CAAC,OAAO,CAAc;IAE7B,OAAO,CAAC,WAAW,CAA8B;gBAErC,IAAI,EAAE,mBAAmB,CAAC,QAAQ,CAAC;IA6B/C,IAAI,KAAK,IAAI,UAAU,CAEtB;IAED,IAAI,MAAM,IAAI,WAAW,CAExB;IAED,IAAI,QAAQ,IAAI,QAAQ,CAMvB;IAED,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,QAAQ,CAAC,KAAK,EAAE,QAAQ,EAE3B;IAEK,KAAK,CAAC,EACV,KAAK,EACL,IAAI,EACJ,YAAY,EACZ,aAAa,GACd,EAAE;QACD,KAAK,EAAE,KAAK,CAAC;QACb,IAAI,EAAE,IAAI,CAAC;QACX,YAAY,CAAC,EAAE,OAAO,CAAC,gBAAgB,CAAC,CAAC;QACzC,aAAa,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;KAC5C,GAAG,OAAO,CAAC,IAAI,CAAC;IAiDjB,WAAW,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI;IAQ/B,cAAc;IAQd,aAAa;IAOb,GAAG,CACD,IAAI,EAAE,MAAM,GAAG,cAAc,CAAC,MAAM,CAAC,EACrC,OAAO,CAAC,EAAE;QACR,KAAK,CAAC,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,YAAY,CAAC,EAAE,OAAO,CAAC;KACxB,GACA,YAAY;IAQf,SAAS;IAOT,aAAa,CAAC,OAAO,CAAC,EAAE;QACtB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,MAAM,CAAC;QACtB,UAAU,CAAC,EAAE,UAAU,CAAC;QACxB,kBAAkB,CAAC,EAAE,OAAO,CAAC;KAC9B,GAAG,YAAY;YAsBF,cAAc;IAmB5B,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,UAAU,IAAI,UAAU,CAE3B;IAED,IAAI,YAAY,IAAI,KAAK,CAMxB;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5B,gBAAgB;IAChB,UAAU,CAAC,EACT,MAAM,EACN,KAAa,EACb,KAAY,GACb,EAAE;QACD,MAAM,EAAE,WAAW,CAAC;QACpB,KAAK,CAAC,EAAE,OAAO,CAAC;QAChB,KAAK,CAAC,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI,CAAC;KACpE,GAAG,IAAI;IAOR,gBAAgB;IAChB,QAAQ,CAAC,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI;IAc1E,gBAAgB;IAChB,sBAAsB,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI;IAK/C,gBAAgB;IAChB,iBAAiB,CAAC,KAAK,EAAE,UAAU;IAanC,gBAAgB;IAChB,gBAAgB,CAAC,KAAK,EAAE,SAAS;IAcjC,OAAO,CAAC,mBAAmB;IAU3B,OAAO,CAAC,oBAAoB;IAE5B,OAAO,CAAC,mBAAmB;YAEb,SAAS;CA6CxB"}
1
+ {"version":3,"file":"agent_session.d.ts","sourceRoot":"","sources":["../../src/voice/agent_session.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAC1D,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAEtD,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,KAAK,EAAE,GAAG,EAAE,aAAa,EAAE,kBAAkB,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC1F,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAE9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACnD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AAExC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AAC5D,OAAO,EACL,sBAAsB,EACtB,KAAK,UAAU,EACf,KAAK,sBAAsB,EAC3B,KAAK,UAAU,EACf,WAAW,EACX,KAAK,0BAA0B,EAC/B,KAAK,UAAU,EACf,KAAK,0BAA0B,EAC/B,KAAK,qBAAqB,EAC1B,KAAK,kBAAkB,EACvB,KAAK,yBAAyB,EAC9B,KAAK,SAAS,EACd,KAAK,qBAAqB,EAK3B,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAClD,OAAO,EAAU,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC3F,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,WAAW,YAAY;IAC3B,kBAAkB,EAAE,OAAO,CAAC;IAC5B,6BAA6B,EAAE,OAAO,CAAC;IACvC,uBAAuB,EAAE,MAAM,CAAC;IAChC,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,YAAY,EAAE,MAAM,CAAC;CACtB;AAYD,MAAM,MAAM,iBAAiB,GAAG,KAAK,GAAG,KAAK,GAAG,cAAc,GAAG,QAAQ,GAAG,aAAa,CAAC;AAE1F,MAAM,MAAM,qBAAqB,GAAG;IAClC,CAAC,sBAAsB,CAAC,oBAAoB,CAAC,EAAE,CAAC,EAAE,EAAE,yBAAyB,KAAK,IAAI,CAAC;IACvF,CAAC,sBAAsB,CAAC,iBAAiB,CAAC,EAAE,CAAC,EAAE,EAAE,sBAAsB,KAAK,IAAI,CAAC;IACjF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,aAAa,CAAC,EAAE,CAAC,EAAE,EAAE,kBAAkB,KAAK,IAAI,CAAC;IACzE,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;IACzD,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;CAC1D,CAAC;AAEF,MAAM,MAAM,mBAAmB,CAAC,QAAQ,GAAG,eAAe,IAAI;IAC5D,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC,YAAY,CAAC,CAAC;CACtC,CAAC;2CAImC,aAAa,qBAAqB,CAAC;AAFxE,qBAAa,YAAY,CACvB,QAAQ,GAAG,eAAe,CAC1B,SAAQ,iBAA+D;IACvE,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAElC,QAAQ,CAAC,OAAO,EAAE,YAAY,CAAC;IAE/B,OAAO,CAAC,KAAK,CAAC,CAAQ;IACtB,OAAO,CAAC,QAAQ,CAAC,CAAgB;IACjC,OAAO,CAAC,YAAY,CAAC,CAAgB;IACrC,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,MAAM,CAAC,CAAS;IACxB,OAAO,CAAC,MAAM,CAAS;IAEvB,OAAO,CAAC,QAAQ,CAAc;IAC9B,OAAO,CAAC,SAAS,CAAuB;IACxC,OAAO,CAAC,WAAW,CAA8B;IAEjD,OAAO,CAAC,MAAM,CAAa;IAC3B,OAAO,CAAC,OAAO,CAAc;IAE7B,OAAO,CAAC,WAAW,CAA8B;gBAErC,IAAI,EAAE,mBAAmB,CAAC,QAAQ,CAAC;IA6B/C,IAAI,KAAK,IAAI,UAAU,CAEtB;IAED,IAAI,MAAM,IAAI,WAAW,CAExB;IAED,IAAI,QAAQ,IAAI,QAAQ,CAMvB;IAED,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,QAAQ,CAAC,KAAK,EAAE,QAAQ,EAE3B;IAEK,KAAK,CAAC,EACV,KAAK,EACL,IAAI,EACJ,YAAY,EACZ,aAAa,GACd,EAAE;QACD,KAAK,EAAE,KAAK,CAAC;QACb,IAAI,EAAE,IAAI,CAAC;QACX,YAAY,CAAC,EAAE,OAAO,CAAC,gBAAgB,CAAC,CAAC;QACzC,aAAa,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;KAC5C,GAAG,OAAO,CAAC,IAAI,CAAC;IAyDjB,WAAW,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI;IAQ/B,cAAc;IAQd,aAAa;IAOb,GAAG,CACD,IAAI,EAAE,MAAM,GAAG,cAAc,CAAC,MAAM,CAAC,EACrC,OAAO,CAAC,EAAE;QACR,KAAK,CAAC,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,YAAY,CAAC,EAAE,OAAO,CAAC;KACxB,GACA,YAAY;IAQf,SAAS;IAOT,aAAa,CAAC,OAAO,CAAC,EAAE;QACtB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,MAAM,CAAC;QACtB,UAAU,CAAC,EAAE,UAAU,CAAC;QACxB,kBAAkB,CAAC,EAAE,OAAO,CAAC;KAC9B,GAAG,YAAY;YAsBF,cAAc;IAmB5B,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,UAAU,IAAI,UAAU,CAE3B;IAED,IAAI,YAAY,IAAI,KAAK,CAMxB;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5B,gBAAgB;IAChB,UAAU,CAAC,EACT,MAAM,EACN,KAAa,EACb,KAAY,GACb,EAAE;QACD,MAAM,EAAE,WAAW,CAAC;QACpB,KAAK,CAAC,EAAE,OAAO,CAAC;QAChB,KAAK,CAAC,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI,CAAC;KACpE,GAAG,IAAI;IAOR,gBAAgB;IAChB,QAAQ,CAAC,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI;IAc1E,gBAAgB;IAChB,sBAAsB,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI;IAK/C,gBAAgB;IAChB,iBAAiB,CAAC,KAAK,EAAE,UAAU;IAanC,gBAAgB;IAChB,gBAAgB,CAAC,KAAK,EAAE,SAAS;IAcjC,OAAO,CAAC,mBAAmB;IAU3B,OAAO,CAAC,oBAAoB;IAE5B,OAAO,CAAC,mBAAmB;YAEb,SAAS;CA6CxB"}
@@ -1,4 +1,5 @@
1
1
  import { EventEmitter } from "node:events";
2
+ import { getJobContext } from "../job.js";
2
3
  import { ChatContext, ChatMessage } from "../llm/chat_context.js";
3
4
  import { log } from "../log.js";
4
5
  import { AgentActivity } from "./agent_activity.js";
@@ -92,6 +93,7 @@ class AgentSession extends EventEmitter {
92
93
  }
93
94
  this.agent = agent;
94
95
  this._updateAgentState("initializing");
96
+ const tasks = [];
95
97
  if (this.input.audio && (inputOptions == null ? void 0 : inputOptions.audioEnabled) !== false) {
96
98
  this.logger.warn("RoomIO audio input is enabled but input.audio is already set, ignoring..");
97
99
  }
@@ -112,14 +114,19 @@ class AgentSession extends EventEmitter {
112
114
  outputOptions
113
115
  });
114
116
  this.roomIO.start();
115
- this.updateActivity(this.agent);
117
+ const ctx = getJobContext();
118
+ if (ctx && ctx.room === room && !room.isConnected) {
119
+ this.logger.debug("Auto-connecting to room via job context");
120
+ tasks.push(ctx.connect());
121
+ }
122
+ tasks.push(this.updateActivity(this.agent));
123
+ await Promise.allSettled(tasks);
116
124
  this.logger.debug(
117
125
  `using audio io: ${this.input.audio ? "`" + this.input.audio.constructor.name + "`" : "(none)"} -> \`AgentSession\` -> ${this.output.audio ? "`" + this.output.audio.constructor.name + "`" : "(none)"}`
118
126
  );
119
127
  this.logger.debug(
120
128
  `using transcript io: \`AgentSession\` -> ${this.output.transcription ? "`" + this.output.transcription.constructor.name + "`" : "(none)"}`
121
129
  );
122
- this.logger.debug("AgentSession started");
123
130
  this.started = true;
124
131
  this._updateAgentState("listening");
125
132
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/voice/agent_session.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, Room } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { ChatContext, ChatMessage } from '../llm/chat_context.js';\nimport type { LLM, RealtimeModel, RealtimeModelError, ToolChoice } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport { log } from '../log.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS, TTSError } from '../tts/tts.js';\nimport type { VAD } from '../vad.js';\nimport type { Agent } from './agent.js';\nimport { AgentActivity } from './agent_activity.js';\nimport type { _TurnDetector } from './audio_recognition.js';\nimport {\n AgentSessionEventTypes,\n type AgentState,\n type AgentStateChangedEvent,\n type CloseEvent,\n CloseReason,\n type ConversationItemAddedEvent,\n type ErrorEvent,\n type FunctionToolsExecutedEvent,\n type MetricsCollectedEvent,\n type SpeechCreatedEvent,\n type UserInputTranscribedEvent,\n type UserState,\n type UserStateChangedEvent,\n createAgentStateChangedEvent,\n createCloseEvent,\n createConversationItemAddedEvent,\n createUserStateChangedEvent,\n} from './events.js';\nimport { AgentInput, AgentOutput } from './io.js';\nimport { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport interface VoiceOptions {\n allowInterruptions: boolean;\n discardAudioIfUninterruptible: boolean;\n minInterruptionDuration: number;\n minInterruptionWords: number;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n maxToolSteps: number;\n}\n\nconst defaultVoiceOptions: VoiceOptions = {\n allowInterruptions: true,\n discardAudioIfUninterruptible: true,\n minInterruptionDuration: 500,\n minInterruptionWords: 0,\n minEndpointingDelay: 500,\n maxEndpointingDelay: 6000,\n maxToolSteps: 3,\n} as const;\n\nexport type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;\n\nexport type AgentSessionCallbacks = {\n [AgentSessionEventTypes.UserInputTranscribed]: (ev: UserInputTranscribedEvent) => void;\n [AgentSessionEventTypes.AgentStateChanged]: (ev: AgentStateChangedEvent) => void;\n [AgentSessionEventTypes.UserStateChanged]: (ev: UserStateChangedEvent) => void;\n [AgentSessionEventTypes.ConversationItemAdded]: (ev: ConversationItemAddedEvent) => void;\n [AgentSessionEventTypes.FunctionToolsExecuted]: (ev: FunctionToolsExecutedEvent) => void;\n [AgentSessionEventTypes.MetricsCollected]: (ev: MetricsCollectedEvent) => void;\n [AgentSessionEventTypes.SpeechCreated]: (ev: SpeechCreatedEvent) => void;\n [AgentSessionEventTypes.Error]: (ev: ErrorEvent) => void;\n [AgentSessionEventTypes.Close]: (ev: CloseEvent) => void;\n};\n\nexport type AgentSessionOptions<UserData = UnknownUserData> = {\n turnDetection?: TurnDetectionMode;\n stt?: STT;\n vad?: VAD;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n userData?: UserData;\n voiceOptions?: Partial<VoiceOptions>;\n};\n\nexport class AgentSession<\n UserData = UnknownUserData,\n> extends (EventEmitter as new () => TypedEmitter<AgentSessionCallbacks>) {\n vad?: VAD;\n stt?: STT;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n turnDetection?: TurnDetectionMode;\n\n readonly options: VoiceOptions;\n\n private agent?: Agent;\n private activity?: AgentActivity;\n private nextActivity?: AgentActivity;\n private started = false;\n private userState: UserState = 'listening';\n\n private roomIO?: RoomIO;\n private logger = log();\n\n private _chatCtx: ChatContext;\n private _userData: UserData | undefined;\n private _agentState: AgentState = 'initializing';\n\n private _input: AgentInput;\n private _output: AgentOutput;\n\n private closingTask: Promise<void> | null = null;\n\n constructor(opts: AgentSessionOptions<UserData>) {\n super();\n\n const {\n vad,\n stt,\n llm,\n tts,\n turnDetection,\n userData,\n voiceOptions = defaultVoiceOptions,\n } = opts;\n\n this.vad = vad;\n this.stt = stt;\n this.llm = llm;\n this.tts = tts;\n this.turnDetection = turnDetection;\n this._userData = userData;\n\n // configurable IO\n this._input = new AgentInput(this.onAudioInputChanged);\n this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);\n\n // This is the \"global\" chat context, it holds the entire conversation history\n this._chatCtx = ChatContext.empty();\n this.options = { ...defaultVoiceOptions, ...voiceOptions };\n }\n\n get input(): AgentInput {\n return this._input;\n }\n\n get output(): AgentOutput {\n return this._output;\n }\n\n get userData(): UserData {\n if (this._userData === undefined) {\n throw new Error('Voice agent userData is not set');\n }\n\n return this._userData;\n }\n\n get history(): ChatContext {\n return this._chatCtx;\n }\n\n set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\n }\n\n if (this.output.audio && outputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio output is enabled but output.audio is already set, ignoring..',\n );\n }\n\n if (this.output.transcription && outputOptions?.transcriptionEnabled !== false) {\n this.logger.warn(\n 'RoomIO transcription output is enabled but output.transcription is already set, ignoring..',\n );\n }\n\n this.roomIO = new RoomIO({\n agentSession: this,\n room,\n inputOptions,\n outputOptions,\n });\n this.roomIO.start();\n\n this.updateActivity(this.agent);\n\n // Log used IO configuration\n this.logger.debug(\n `using audio io: ${this.input.audio ? '`' + this.input.audio.constructor.name + '`' : '(none)'} -> \\`AgentSession\\` -> ${this.output.audio ? '`' + this.output.audio.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug(\n `using transcript io: \\`AgentSession\\` -> ${this.output.transcription ? '`' + this.output.transcription.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug('AgentSession started');\n this.started = true;\n this._updateAgentState('listening');\n }\n\n updateAgent(agent: Agent): void {\n this.agent = agent;\n\n if (this.started) {\n this.updateActivity(agent);\n }\n }\n\n commitUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n this.activity.commitUserTurn();\n }\n\n clearUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n this.activity.clearUserTurn();\n }\n\n say(\n text: string | ReadableStream<string>,\n options?: {\n audio?: ReadableStream<AudioFrame>;\n allowInterruptions?: boolean;\n addToChatCtx?: boolean;\n },\n ): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n return this.activity.say(text, options);\n }\n\n interrupt() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n return this.activity.interrupt();\n }\n\n generateReply(options?: {\n userInput?: string;\n instructions?: string;\n toolChoice?: ToolChoice;\n allowInterruptions?: boolean;\n }): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const userMessage = options?.userInput\n ? new ChatMessage({\n role: 'user',\n content: options.userInput,\n })\n : undefined;\n\n if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\n\n if (this.activity) {\n await this.activity.drain();\n await this.activity.close();\n }\n\n this.activity = this.nextActivity;\n this.nextActivity = undefined;\n\n await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n get chatCtx(): ChatContext {\n return this._chatCtx.copy();\n }\n\n get agentState(): AgentState {\n return this._agentState;\n }\n\n get currentAgent(): Agent {\n if (!this.agent) {\n throw new Error('AgentSession is not running');\n }\n\n return this.agent;\n }\n\n async close(): Promise<void> {\n await this.closeImpl(CloseReason.USER_INITIATED);\n }\n\n /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\n drain?: boolean;\n error?: RealtimeModelError | STTError | TTSError | LLMError | null;\n }): void {\n if (this.closingTask) {\n return;\n }\n this.closeImpl(reason, error, drain);\n }\n\n /** @internal */\n _onError(error: RealtimeModelError | STTError | TTSError | LLMError): void {\n if (this.closingTask || error.recoverable) {\n return;\n }\n\n this.logger.error(error, 'AgentSession is closing due to unrecoverable error');\n\n this.closingTask = (async () => {\n await this.closeImpl(CloseReason.ERROR, error);\n })().then(() => {\n this.closingTask = null;\n });\n }\n\n /** @internal */\n _conversationItemAdded(item: ChatMessage): void {\n this._chatCtx.insert(item);\n this.emit(AgentSessionEventTypes.ConversationItemAdded, createConversationItemAddedEvent(item));\n }\n\n /** @internal */\n _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\n this.emit(\n AgentSessionEventTypes.AgentStateChanged,\n createAgentStateChangedEvent(oldState, state),\n );\n }\n\n /** @internal */\n _updateUserState(state: UserState) {\n if (this.userState === state) {\n return;\n }\n\n const oldState = this.userState;\n this.userState = state;\n this.emit(\n AgentSessionEventTypes.UserStateChanged,\n createUserStateChangedEvent(oldState, state),\n );\n }\n\n // -- User changed input/output streams/sinks --\n private onAudioInputChanged(): void {\n if (!this.started) {\n return;\n }\n\n if (this.activity && this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n private onAudioOutputChanged(): void {}\n\n private onTextOutputChanged(): void {}\n\n private async closeImpl(\n reason: CloseReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\n // TODO(shubhra): force interrupt or wait for it to finish?\n // it might be an audio played from the error callback\n }\n }\n await this.activity.drain();\n // wait any uninterruptible speech to finish\n await this.activity.currentSpeech?.waitForPlayout();\n this.activity.detachAudioInput();\n }\n\n // detach the inputs and outputs\n this.input.audio = null;\n this.output.audio = null;\n this.output.transcription = null;\n\n await this.roomIO?.close();\n this.roomIO = undefined;\n\n await this.activity?.close();\n this.activity = undefined;\n\n this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B,SAAS,aAAa,mBAAmB;AAGzC,SAAS,WAAW;AAMpB,SAAS,qBAAqB;AAE9B;AAAA,EACE;AAAA,EAIA;AAAA,EASA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,cAA6D;AActE,MAAM,sBAAoC;AAAA,EACxC,oBAAoB;AAAA,EACpB,+BAA+B;AAAA,EAC/B,yBAAyB;AAAA,EACzB,sBAAsB;AAAA,EACtB,qBAAqB;AAAA,EACrB,qBAAqB;AAAA,EACrB,cAAc;AAChB;AA0BO,MAAM,qBAEF,aAA+D;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAES;AAAA,EAED;AAAA,EACA;AAAA,EACA;AAAA,EACA,UAAU;AAAA,EACV,YAAuB;AAAA,EAEvB;AAAA,EACA,SAAS,IAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EAE5C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,SAAK,MAAM;AACX,SAAK,MAAM;AACX,SAAK,MAAM;AACX,SAAK,MAAM;AACX,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,WAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,YAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,YAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAAA,EAC3D;AAAA,EAEA,IAAI,QAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,WAAqB;AACvB,QAAI,KAAK,cAAc,QAAW;AAChC,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAGrC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,SAAK,eAAe,KAAK,KAAK;AAG9B,SAAK,OAAO;AAAA,MACV,mBAAmB,KAAK,MAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,YAAY,OAAO,MAAM,QAAQ,2BAA2B,KAAK,OAAO,QAAQ,MAAM,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,QAAQ;AAAA,IACxM;AAEA,SAAK,OAAO;AAAA,MACV,4CAA4C,KAAK,OAAO,gBAAgB,MAAM,KAAK,OAAO,cAAc,YAAY,OAAO,MAAM,QAAQ;AAAA,IAC3I;AAEA,SAAK,OAAO,MAAM,sBAAsB;AACxC,SAAK,UAAU;AACf,SAAK,kBAAkB,WAAW;AAAA,EACpC;AAAA,EAEA,YAAY,OAAoB;AAC9B,SAAK,QAAQ;AAEb,QAAI,KAAK,SAAS;AAChB,WAAK,eAAe,KAAK;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,iBAAiB;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,SAAK,SAAS,eAAe;AAAA,EAC/B;AAAA,EAEA,gBAAgB;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,SAAK,SAAS,cAAc;AAAA,EAC9B;AAAA,EAEA,IACE,MACA,SAKc;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;AAAA,EAEA,YAAY;AACV,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EAEA,cAAc,SAKG;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,eAAc,mCAAS,aACzB,IAAI,YAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAsB;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAuB;AAC3B,UAAM,KAAK,UAAU,YAAY,cAAc;AAAA,EACjD;AAAA;AAAA,EAGA,WAAW;AAAA,IACT;AAAA,IACA,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV,GAIS;AACP,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AACA,SAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,OAAkE;AACzE,QAAI,KAAK,eAAe,MAAM,aAAa;AACzC;AAAA,IACF;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,YAAY,OAAO,KAAK;AAAA,IAC/C,GAAG,EAAE,KAAK,MAAM;AACd,WAAK,cAAc;AAAA,IACrB,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,uBAAuB,MAAyB;AAC9C,SAAK,SAAS,OAAO,IAAI;AACzB,SAAK,KAAK,uBAAuB,uBAAuB,iCAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB;AACjC,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AACjB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,4BAA4B,UAAU,KAAK;AAAA,IAC7C;AAAA,EACF;AAAA;AAAA,EAGQ,sBAA4B;AAClC,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,YAAY,KAAK,OAAO,OAAO;AACtC,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEQ,uBAA6B;AAAA,EAAC;AAAA,EAE9B,sBAA4B;AAAA,EAAC;AAAA,EAErC,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AAtanB;AAuaI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;AAGA,SAAK,MAAM,QAAQ;AACnB,SAAK,OAAO,QAAQ;AACpB,SAAK,OAAO,gBAAgB;AAE5B,YAAM,UAAK,WAAL,mBAAa;AACnB,SAAK,SAAS;AAEd,YAAM,UAAK,aAAL,mBAAe;AACrB,SAAK,WAAW;AAEhB,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
1
+ {"version":3,"sources":["../../src/voice/agent_session.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, Room } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { getJobContext } from '../job.js';\nimport { ChatContext, ChatMessage } from '../llm/chat_context.js';\nimport type { LLM, RealtimeModel, RealtimeModelError, ToolChoice } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport { log } from '../log.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS, TTSError } from '../tts/tts.js';\nimport type { VAD } from '../vad.js';\nimport type { Agent } from './agent.js';\nimport { AgentActivity } from './agent_activity.js';\nimport type { _TurnDetector } from './audio_recognition.js';\nimport {\n AgentSessionEventTypes,\n type AgentState,\n type AgentStateChangedEvent,\n type CloseEvent,\n CloseReason,\n type ConversationItemAddedEvent,\n type ErrorEvent,\n type FunctionToolsExecutedEvent,\n type MetricsCollectedEvent,\n type SpeechCreatedEvent,\n type UserInputTranscribedEvent,\n type UserState,\n type UserStateChangedEvent,\n createAgentStateChangedEvent,\n createCloseEvent,\n createConversationItemAddedEvent,\n createUserStateChangedEvent,\n} from './events.js';\nimport { AgentInput, AgentOutput } from './io.js';\nimport { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport interface VoiceOptions {\n allowInterruptions: boolean;\n discardAudioIfUninterruptible: boolean;\n minInterruptionDuration: number;\n minInterruptionWords: number;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n maxToolSteps: number;\n}\n\nconst defaultVoiceOptions: VoiceOptions = {\n allowInterruptions: true,\n discardAudioIfUninterruptible: true,\n minInterruptionDuration: 500,\n minInterruptionWords: 0,\n minEndpointingDelay: 500,\n maxEndpointingDelay: 6000,\n maxToolSteps: 3,\n} as const;\n\nexport type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;\n\nexport type AgentSessionCallbacks = {\n [AgentSessionEventTypes.UserInputTranscribed]: (ev: UserInputTranscribedEvent) => void;\n [AgentSessionEventTypes.AgentStateChanged]: (ev: AgentStateChangedEvent) => void;\n [AgentSessionEventTypes.UserStateChanged]: (ev: UserStateChangedEvent) => void;\n [AgentSessionEventTypes.ConversationItemAdded]: (ev: ConversationItemAddedEvent) => void;\n [AgentSessionEventTypes.FunctionToolsExecuted]: (ev: FunctionToolsExecutedEvent) => void;\n [AgentSessionEventTypes.MetricsCollected]: (ev: MetricsCollectedEvent) => void;\n [AgentSessionEventTypes.SpeechCreated]: (ev: SpeechCreatedEvent) => void;\n [AgentSessionEventTypes.Error]: (ev: ErrorEvent) => void;\n [AgentSessionEventTypes.Close]: (ev: CloseEvent) => void;\n};\n\nexport type AgentSessionOptions<UserData = UnknownUserData> = {\n turnDetection?: TurnDetectionMode;\n stt?: STT;\n vad?: VAD;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n userData?: UserData;\n voiceOptions?: Partial<VoiceOptions>;\n};\n\nexport class AgentSession<\n UserData = UnknownUserData,\n> extends (EventEmitter as new () => TypedEmitter<AgentSessionCallbacks>) {\n vad?: VAD;\n stt?: STT;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n turnDetection?: TurnDetectionMode;\n\n readonly options: VoiceOptions;\n\n private agent?: Agent;\n private activity?: AgentActivity;\n private nextActivity?: AgentActivity;\n private started = false;\n private userState: UserState = 'listening';\n\n private roomIO?: RoomIO;\n private logger = log();\n\n private _chatCtx: ChatContext;\n private _userData: UserData | undefined;\n private _agentState: AgentState = 'initializing';\n\n private _input: AgentInput;\n private _output: AgentOutput;\n\n private closingTask: Promise<void> | null = null;\n\n constructor(opts: AgentSessionOptions<UserData>) {\n super();\n\n const {\n vad,\n stt,\n llm,\n tts,\n turnDetection,\n userData,\n voiceOptions = defaultVoiceOptions,\n } = opts;\n\n this.vad = vad;\n this.stt = stt;\n this.llm = llm;\n this.tts = tts;\n this.turnDetection = turnDetection;\n this._userData = userData;\n\n // configurable IO\n this._input = new AgentInput(this.onAudioInputChanged);\n this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);\n\n // This is the \"global\" chat context, it holds the entire conversation history\n this._chatCtx = ChatContext.empty();\n this.options = { ...defaultVoiceOptions, ...voiceOptions };\n }\n\n get input(): AgentInput {\n return this._input;\n }\n\n get output(): AgentOutput {\n return this._output;\n }\n\n get userData(): UserData {\n if (this._userData === undefined) {\n throw new Error('Voice agent userData is not set');\n }\n\n return this._userData;\n }\n\n get history(): ChatContext {\n return this._chatCtx;\n }\n\n set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\n }\n\n if (this.output.audio && outputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio output is enabled but output.audio is already set, ignoring..',\n );\n }\n\n if (this.output.transcription && outputOptions?.transcriptionEnabled !== false) {\n this.logger.warn(\n 'RoomIO transcription output is enabled but output.transcription is already set, ignoring..',\n );\n }\n\n this.roomIO = new RoomIO({\n agentSession: this,\n room,\n inputOptions,\n outputOptions,\n });\n this.roomIO.start();\n\n const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\n }\n // TODO(AJS-265): add shutdown callback to job context\n tasks.push(this.updateActivity(this.agent));\n\n await Promise.allSettled(tasks);\n\n // Log used IO configuration\n this.logger.debug(\n `using audio io: ${this.input.audio ? '`' + this.input.audio.constructor.name + '`' : '(none)'} -> \\`AgentSession\\` -> ${this.output.audio ? '`' + this.output.audio.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug(\n `using transcript io: \\`AgentSession\\` -> ${this.output.transcription ? '`' + this.output.transcription.constructor.name + '`' : '(none)'}`,\n );\n\n this.started = true;\n this._updateAgentState('listening');\n }\n\n updateAgent(agent: Agent): void {\n this.agent = agent;\n\n if (this.started) {\n this.updateActivity(agent);\n }\n }\n\n commitUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n this.activity.commitUserTurn();\n }\n\n clearUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n this.activity.clearUserTurn();\n }\n\n say(\n text: string | ReadableStream<string>,\n options?: {\n audio?: ReadableStream<AudioFrame>;\n allowInterruptions?: boolean;\n addToChatCtx?: boolean;\n },\n ): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n return this.activity.say(text, options);\n }\n\n interrupt() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n return this.activity.interrupt();\n }\n\n generateReply(options?: {\n userInput?: string;\n instructions?: string;\n toolChoice?: ToolChoice;\n allowInterruptions?: boolean;\n }): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const userMessage = options?.userInput\n ? new ChatMessage({\n role: 'user',\n content: options.userInput,\n })\n : undefined;\n\n if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\n\n if (this.activity) {\n await this.activity.drain();\n await this.activity.close();\n }\n\n this.activity = this.nextActivity;\n this.nextActivity = undefined;\n\n await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n get chatCtx(): ChatContext {\n return this._chatCtx.copy();\n }\n\n get agentState(): AgentState {\n return this._agentState;\n }\n\n get currentAgent(): Agent {\n if (!this.agent) {\n throw new Error('AgentSession is not running');\n }\n\n return this.agent;\n }\n\n async close(): Promise<void> {\n await this.closeImpl(CloseReason.USER_INITIATED);\n }\n\n /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\n drain?: boolean;\n error?: RealtimeModelError | STTError | TTSError | LLMError | null;\n }): void {\n if (this.closingTask) {\n return;\n }\n this.closeImpl(reason, error, drain);\n }\n\n /** @internal */\n _onError(error: RealtimeModelError | STTError | TTSError | LLMError): void {\n if (this.closingTask || error.recoverable) {\n return;\n }\n\n this.logger.error(error, 'AgentSession is closing due to unrecoverable error');\n\n this.closingTask = (async () => {\n await this.closeImpl(CloseReason.ERROR, error);\n })().then(() => {\n this.closingTask = null;\n });\n }\n\n /** @internal */\n _conversationItemAdded(item: ChatMessage): void {\n this._chatCtx.insert(item);\n this.emit(AgentSessionEventTypes.ConversationItemAdded, createConversationItemAddedEvent(item));\n }\n\n /** @internal */\n _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\n this.emit(\n AgentSessionEventTypes.AgentStateChanged,\n createAgentStateChangedEvent(oldState, state),\n );\n }\n\n /** @internal */\n _updateUserState(state: UserState) {\n if (this.userState === state) {\n return;\n }\n\n const oldState = this.userState;\n this.userState = state;\n this.emit(\n AgentSessionEventTypes.UserStateChanged,\n createUserStateChangedEvent(oldState, state),\n );\n }\n\n // -- User changed input/output streams/sinks --\n private onAudioInputChanged(): void {\n if (!this.started) {\n return;\n }\n\n if (this.activity && this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n private onAudioOutputChanged(): void {}\n\n private onTextOutputChanged(): void {}\n\n private async closeImpl(\n reason: CloseReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\n // TODO(shubhra): force interrupt or wait for it to finish?\n // it might be an audio played from the error callback\n }\n }\n await this.activity.drain();\n // wait any uninterruptible speech to finish\n await this.activity.currentSpeech?.waitForPlayout();\n this.activity.detachAudioInput();\n }\n\n // detach the inputs and outputs\n this.input.audio = null;\n this.output.audio = null;\n this.output.transcription = null;\n\n await this.roomIO?.close();\n this.roomIO = undefined;\n\n await this.activity?.close();\n this.activity = undefined;\n\n this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B,SAAS,qBAAqB;AAC9B,SAAS,aAAa,mBAAmB;AAGzC,SAAS,WAAW;AAMpB,SAAS,qBAAqB;AAE9B;AAAA,EACE;AAAA,EAIA;AAAA,EASA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,cAA6D;AActE,MAAM,sBAAoC;AAAA,EACxC,oBAAoB;AAAA,EACpB,+BAA+B;AAAA,EAC/B,yBAAyB;AAAA,EACzB,sBAAsB;AAAA,EACtB,qBAAqB;AAAA,EACrB,qBAAqB;AAAA,EACrB,cAAc;AAChB;AA0BO,MAAM,qBAEF,aAA+D;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAES;AAAA,EAED;AAAA,EACA;AAAA,EACA;AAAA,EACA,UAAU;AAAA,EACV,YAAuB;AAAA,EAEvB;AAAA,EACA,SAAS,IAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EAE5C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,SAAK,MAAM;AACX,SAAK,MAAM;AACX,SAAK,MAAM;AACX,SAAK,MAAM;AACX,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,WAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,YAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,YAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAAA,EAC3D;AAAA,EAEA,IAAI,QAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,WAAqB;AACvB,QAAI,KAAK,cAAc,QAAW;AAChC,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,MAAM,cAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,UAAM,KAAK,KAAK,eAAe,KAAK,KAAK,CAAC;AAE1C,UAAM,QAAQ,WAAW,KAAK;AAG9B,SAAK,OAAO;AAAA,MACV,mBAAmB,KAAK,MAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,YAAY,OAAO,MAAM,QAAQ,2BAA2B,KAAK,OAAO,QAAQ,MAAM,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,QAAQ;AAAA,IACxM;AAEA,SAAK,OAAO;AAAA,MACV,4CAA4C,KAAK,OAAO,gBAAgB,MAAM,KAAK,OAAO,cAAc,YAAY,OAAO,MAAM,QAAQ;AAAA,IAC3I;AAEA,SAAK,UAAU;AACf,SAAK,kBAAkB,WAAW;AAAA,EACpC;AAAA,EAEA,YAAY,OAAoB;AAC9B,SAAK,QAAQ;AAEb,QAAI,KAAK,SAAS;AAChB,WAAK,eAAe,KAAK;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,iBAAiB;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,SAAK,SAAS,eAAe;AAAA,EAC/B;AAAA,EAEA,gBAAgB;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,SAAK,SAAS,cAAc;AAAA,EAC9B;AAAA,EAEA,IACE,MACA,SAKc;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;AAAA,EAEA,YAAY;AACV,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EAEA,cAAc,SAKG;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,eAAc,mCAAS,aACzB,IAAI,YAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAsB;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAuB;AAC3B,UAAM,KAAK,UAAU,YAAY,cAAc;AAAA,EACjD;AAAA;AAAA,EAGA,WAAW;AAAA,IACT;AAAA,IACA,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV,GAIS;AACP,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AACA,SAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,OAAkE;AACzE,QAAI,KAAK,eAAe,MAAM,aAAa;AACzC;AAAA,IACF;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,YAAY,OAAO,KAAK;AAAA,IAC/C,GAAG,EAAE,KAAK,MAAM;AACd,WAAK,cAAc;AAAA,IACrB,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,uBAAuB,MAAyB;AAC9C,SAAK,SAAS,OAAO,IAAI;AACzB,SAAK,KAAK,uBAAuB,uBAAuB,iCAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB;AACjC,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AACjB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,4BAA4B,UAAU,KAAK;AAAA,IAC7C;AAAA,EACF;AAAA;AAAA,EAGQ,sBAA4B;AAClC,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,YAAY,KAAK,OAAO,OAAO;AACtC,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEQ,uBAA6B;AAAA,EAAC;AAAA,EAE9B,sBAA4B;AAAA,EAAC;AAAA,EAErC,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AA/anB;AAgbI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;AAGA,SAAK,MAAM,QAAQ;AACnB,SAAK,OAAO,QAAQ;AACpB,SAAK,OAAO,gBAAgB;AAE5B,YAAM,UAAK,WAAL,mBAAa;AACnB,SAAK,SAAS;AAEd,YAAM,UAAK,aAAL,mBAAe;AACrB,SAAK,WAAW;AAEhB,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
@@ -0,0 +1,204 @@
1
+ "use strict";
2
+ var __defProp = Object.defineProperty;
3
+ var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
4
+ var __getOwnPropNames = Object.getOwnPropertyNames;
5
+ var __hasOwnProp = Object.prototype.hasOwnProperty;
6
+ var __export = (target, all) => {
7
+ for (var name in all)
8
+ __defProp(target, name, { get: all[name], enumerable: true });
9
+ };
10
+ var __copyProps = (to, from, except, desc) => {
11
+ if (from && typeof from === "object" || typeof from === "function") {
12
+ for (let key of __getOwnPropNames(from))
13
+ if (!__hasOwnProp.call(to, key) && key !== except)
14
+ __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
15
+ }
16
+ return to;
17
+ };
18
+ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
19
+ var datastream_io_exports = {};
20
+ __export(datastream_io_exports, {
21
+ DataStreamAudioOutput: () => DataStreamAudioOutput
22
+ });
23
+ module.exports = __toCommonJS(datastream_io_exports);
24
+ var import_mutex = require("@livekit/mutex");
25
+ var import_rtc_node = require("@livekit/rtc-node");
26
+ var import_log = require("../../log.cjs");
27
+ var import_utils = require("../../utils.cjs");
28
+ var import_io = require("../io.cjs");
29
+ const RPC_CLEAR_BUFFER = "lk.clear_buffer";
30
+ const RPC_PLAYBACK_FINISHED = "lk.playback_finished";
31
+ const AUDIO_STREAM_TOPIC = "lk.audio_stream";
32
+ class DataStreamAudioOutput extends import_io.AudioOutput {
33
+ static _playbackFinishedRpcRegistered = false;
34
+ static _playbackFinishedHandlers = {};
35
+ room;
36
+ destinationIdentity;
37
+ roomConnectedFuture;
38
+ waitRemoteTrack;
39
+ streamWriter;
40
+ pushedDuration = 0;
41
+ started = false;
42
+ lock = new import_mutex.Mutex();
43
+ startTask;
44
+ #logger = (0, import_log.log)();
45
+ constructor(opts) {
46
+ super(opts.sampleRate, void 0);
47
+ const { room, destinationIdentity, sampleRate, waitRemoteTrack } = opts;
48
+ this.room = room;
49
+ this.destinationIdentity = destinationIdentity;
50
+ this.sampleRate = sampleRate;
51
+ this.waitRemoteTrack = waitRemoteTrack;
52
+ const onRoomConnected = async () => {
53
+ if (this.startTask) return;
54
+ await this.roomConnectedFuture.await;
55
+ DataStreamAudioOutput.registerPlaybackFinishedRpc({
56
+ room,
57
+ callerIdentity: this.destinationIdentity,
58
+ handler: (data) => this.handlePlaybackFinished(data)
59
+ });
60
+ this.startTask = import_utils.Task.from(({ signal }) => this._start(signal));
61
+ };
62
+ this.roomConnectedFuture = new import_utils.Future();
63
+ this.room.on(import_rtc_node.RoomEvent.ConnectionStateChanged, (_) => {
64
+ if (room.isConnected && !this.roomConnectedFuture.done) {
65
+ this.roomConnectedFuture.resolve(void 0);
66
+ }
67
+ });
68
+ if (this.room.isConnected) {
69
+ this.roomConnectedFuture.resolve(void 0);
70
+ }
71
+ onRoomConnected();
72
+ }
73
+ async _start(_abortSignal) {
74
+ const unlock = await this.lock.lock();
75
+ try {
76
+ if (this.started) return;
77
+ await this.roomConnectedFuture.await;
78
+ this.#logger.debug(
79
+ {
80
+ identity: this.destinationIdentity
81
+ },
82
+ "waiting for the remote participant"
83
+ );
84
+ await (0, import_utils.waitForParticipant)({
85
+ room: this.room,
86
+ identity: this.destinationIdentity
87
+ });
88
+ if (this.waitRemoteTrack) {
89
+ this.#logger.debug(
90
+ {
91
+ identity: this.destinationIdentity,
92
+ kind: this.waitRemoteTrack
93
+ },
94
+ "waiting for the remote track"
95
+ );
96
+ await (0, import_utils.waitForTrackPublication)({
97
+ room: this.room,
98
+ identity: this.destinationIdentity,
99
+ kind: this.waitRemoteTrack
100
+ });
101
+ }
102
+ this.#logger.debug(
103
+ {
104
+ identity: this.destinationIdentity
105
+ },
106
+ "remote participant ready"
107
+ );
108
+ this.started = true;
109
+ } finally {
110
+ unlock();
111
+ }
112
+ }
113
+ async captureFrame(frame) {
114
+ if (!this.startTask) {
115
+ this.startTask = import_utils.Task.from(({ signal }) => this._start(signal));
116
+ }
117
+ await this.startTask.result;
118
+ await super.captureFrame(frame);
119
+ if (!this.streamWriter) {
120
+ this.streamWriter = await this.room.localParticipant.streamBytes({
121
+ name: (0, import_utils.shortuuid)("AUDIO_"),
122
+ topic: AUDIO_STREAM_TOPIC,
123
+ destinationIdentities: [this.destinationIdentity],
124
+ attributes: {
125
+ sample_rate: frame.sampleRate.toString(),
126
+ num_channels: frame.channels.toString()
127
+ }
128
+ });
129
+ this.pushedDuration = 0;
130
+ }
131
+ await this.streamWriter.write(new Uint8Array(frame.data.buffer));
132
+ this.pushedDuration += frame.samplesPerChannel / frame.sampleRate;
133
+ }
134
+ flush() {
135
+ super.flush();
136
+ if (this.streamWriter === void 0 || !this.started) {
137
+ return;
138
+ }
139
+ this.streamWriter.close().finally(() => {
140
+ this.streamWriter = void 0;
141
+ });
142
+ }
143
+ clearBuffer() {
144
+ if (!this.started) return;
145
+ this.room.localParticipant.performRpc({
146
+ destinationIdentity: this.destinationIdentity,
147
+ method: RPC_CLEAR_BUFFER,
148
+ payload: ""
149
+ });
150
+ }
151
+ handlePlaybackFinished(data) {
152
+ if (data.callerIdentity !== this.destinationIdentity) {
153
+ this.#logger.warn(
154
+ {
155
+ callerIdentity: data.callerIdentity,
156
+ destinationIdentity: this.destinationIdentity
157
+ },
158
+ "playback finished event received from unexpected participant"
159
+ );
160
+ return "reject";
161
+ }
162
+ this.#logger.info(
163
+ {
164
+ callerIdentity: data.callerIdentity
165
+ },
166
+ "playback finished event received"
167
+ );
168
+ const playbackFinishedEvent = JSON.parse(data.payload);
169
+ this.onPlaybackFinished(playbackFinishedEvent);
170
+ return "ok";
171
+ }
172
+ static registerPlaybackFinishedRpc({
173
+ room,
174
+ callerIdentity,
175
+ handler
176
+ }) {
177
+ var _a;
178
+ DataStreamAudioOutput._playbackFinishedHandlers[callerIdentity] = handler;
179
+ if (DataStreamAudioOutput._playbackFinishedRpcRegistered) {
180
+ return;
181
+ }
182
+ const rpcHandler = async (data) => {
183
+ const handler2 = DataStreamAudioOutput._playbackFinishedHandlers[data.callerIdentity];
184
+ if (!handler2) {
185
+ (0, import_log.log)().warn(
186
+ {
187
+ callerIdentity: data.callerIdentity,
188
+ expectedIdentities: Object.keys(DataStreamAudioOutput._playbackFinishedHandlers)
189
+ },
190
+ "playback finished event received from unexpected participant"
191
+ );
192
+ return "reject";
193
+ }
194
+ return handler2(data);
195
+ };
196
+ (_a = room.localParticipant) == null ? void 0 : _a.registerRpcMethod(RPC_PLAYBACK_FINISHED, rpcHandler);
197
+ DataStreamAudioOutput._playbackFinishedRpcRegistered = true;
198
+ }
199
+ }
200
+ // Annotate the CommonJS export names for ESM import in node:
201
+ 0 && (module.exports = {
202
+ DataStreamAudioOutput
203
+ });
204
+ //# sourceMappingURL=datastream_io.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../../../src/voice/avatar/datastream_io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { Mutex } from '@livekit/mutex';\nimport {\n type AudioFrame,\n type ByteStreamWriter,\n type Room,\n RoomEvent,\n type RpcInvocationData,\n type TrackKind,\n} from '@livekit/rtc-node';\nimport { log } from '../../log.js';\nimport {\n Future,\n Task,\n shortuuid,\n waitForParticipant,\n waitForTrackPublication,\n} from '../../utils.js';\nimport { AudioOutput, type PlaybackFinishedEvent } from '../io.js';\n\nconst RPC_CLEAR_BUFFER = 'lk.clear_buffer';\nconst RPC_PLAYBACK_FINISHED = 'lk.playback_finished';\nconst AUDIO_STREAM_TOPIC = 'lk.audio_stream';\n\nexport interface DataStreamAudioOutputOptions {\n room: Room;\n destinationIdentity: string;\n sampleRate?: number;\n waitRemoteTrack?: TrackKind;\n}\n\n/**\n * AudioOutput implementation that streams audio to a remote avatar worker using LiveKit DataStream.\n */\nexport class DataStreamAudioOutput extends AudioOutput {\n static _playbackFinishedRpcRegistered: boolean = false;\n static _playbackFinishedHandlers: Record<string, (data: RpcInvocationData) => string> = {};\n\n private room: Room;\n private destinationIdentity: string;\n private roomConnectedFuture: Future<void>;\n private waitRemoteTrack?: TrackKind;\n private streamWriter?: ByteStreamWriter;\n private pushedDuration: number = 0;\n private started: boolean = false;\n private lock = new Mutex();\n private startTask?: Task<void>;\n\n #logger = log();\n\n constructor(opts: DataStreamAudioOutputOptions) {\n super(opts.sampleRate, undefined);\n\n const { room, destinationIdentity, sampleRate, waitRemoteTrack } = opts;\n this.room = room;\n this.destinationIdentity = destinationIdentity;\n this.sampleRate = sampleRate;\n this.waitRemoteTrack = waitRemoteTrack;\n\n const onRoomConnected = async () => {\n if (this.startTask) return;\n\n await this.roomConnectedFuture.await;\n\n // register the rpc method right after the room is connected\n DataStreamAudioOutput.registerPlaybackFinishedRpc({\n room,\n callerIdentity: this.destinationIdentity,\n handler: (data) => this.handlePlaybackFinished(data),\n });\n\n this.startTask = Task.from(({ signal }) => this._start(signal));\n };\n\n this.roomConnectedFuture = new Future<void>();\n\n this.room.on(RoomEvent.ConnectionStateChanged, (_) => {\n if (room.isConnected && !this.roomConnectedFuture.done) {\n this.roomConnectedFuture.resolve(undefined);\n }\n });\n\n if (this.room.isConnected) {\n this.roomConnectedFuture.resolve(undefined);\n }\n\n onRoomConnected();\n }\n\n private async _start(_abortSignal: AbortSignal) {\n const unlock = await this.lock.lock();\n\n try {\n if (this.started) return;\n\n await this.roomConnectedFuture.await;\n\n this.#logger.debug(\n {\n identity: this.destinationIdentity,\n },\n 'waiting for the remote participant',\n );\n\n await waitForParticipant({\n room: this.room,\n identity: this.destinationIdentity,\n });\n\n if (this.waitRemoteTrack) {\n this.#logger.debug(\n {\n identity: this.destinationIdentity,\n kind: this.waitRemoteTrack,\n },\n 'waiting for the remote track',\n );\n\n await waitForTrackPublication({\n room: this.room,\n identity: this.destinationIdentity,\n kind: this.waitRemoteTrack,\n });\n }\n\n this.#logger.debug(\n {\n identity: this.destinationIdentity,\n },\n 'remote participant ready',\n );\n\n this.started = true;\n } finally {\n unlock();\n }\n }\n\n async captureFrame(frame: AudioFrame): Promise<void> {\n if (!this.startTask) {\n this.startTask = Task.from(({ signal }) => this._start(signal));\n }\n\n await this.startTask.result;\n await super.captureFrame(frame);\n\n if (!this.streamWriter) {\n this.streamWriter = await this.room.localParticipant!.streamBytes({\n name: shortuuid('AUDIO_'),\n topic: AUDIO_STREAM_TOPIC,\n destinationIdentities: [this.destinationIdentity],\n attributes: {\n sample_rate: frame.sampleRate.toString(),\n num_channels: frame.channels.toString(),\n },\n });\n this.pushedDuration = 0;\n }\n\n // frame.data is a Int16Array, write accepts a Uint8Array\n await this.streamWriter.write(new Uint8Array(frame.data.buffer));\n this.pushedDuration += frame.samplesPerChannel / frame.sampleRate;\n }\n\n flush(): void {\n super.flush();\n\n if (this.streamWriter === undefined || !this.started) {\n return;\n }\n\n this.streamWriter.close().finally(() => {\n this.streamWriter = undefined;\n });\n }\n\n clearBuffer(): void {\n if (!this.started) return;\n\n this.room.localParticipant!.performRpc({\n destinationIdentity: this.destinationIdentity,\n method: RPC_CLEAR_BUFFER,\n payload: '',\n });\n }\n\n private handlePlaybackFinished(data: RpcInvocationData): string {\n if (data.callerIdentity !== this.destinationIdentity) {\n this.#logger.warn(\n {\n callerIdentity: data.callerIdentity,\n destinationIdentity: this.destinationIdentity,\n },\n 'playback finished event received from unexpected participant',\n );\n return 'reject';\n }\n\n this.#logger.info(\n {\n callerIdentity: data.callerIdentity,\n },\n 'playback finished event received',\n );\n\n const playbackFinishedEvent = JSON.parse(data.payload) as PlaybackFinishedEvent;\n this.onPlaybackFinished(playbackFinishedEvent);\n return 'ok';\n }\n\n static registerPlaybackFinishedRpc({\n room,\n callerIdentity,\n handler,\n }: {\n room: Room;\n callerIdentity: string;\n handler: (data: RpcInvocationData) => string;\n }) {\n DataStreamAudioOutput._playbackFinishedHandlers[callerIdentity] = handler;\n\n if (DataStreamAudioOutput._playbackFinishedRpcRegistered) {\n return;\n }\n\n const rpcHandler = async (data: RpcInvocationData): Promise<string> => {\n const handler = DataStreamAudioOutput._playbackFinishedHandlers[data.callerIdentity];\n if (!handler) {\n log().warn(\n {\n callerIdentity: data.callerIdentity,\n expectedIdentities: Object.keys(DataStreamAudioOutput._playbackFinishedHandlers),\n },\n 'playback finished event received from unexpected participant',\n );\n\n return 'reject';\n }\n return handler(data);\n };\n\n room.localParticipant?.registerRpcMethod(RPC_PLAYBACK_FINISHED, rpcHandler);\n DataStreamAudioOutput._playbackFinishedRpcRegistered = true;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,mBAAsB;AACtB,sBAOO;AACP,iBAAoB;AACpB,mBAMO;AACP,gBAAwD;AAExD,MAAM,mBAAmB;AACzB,MAAM,wBAAwB;AAC9B,MAAM,qBAAqB;AAYpB,MAAM,8BAA8B,sBAAY;AAAA,EACrD,OAAO,iCAA0C;AAAA,EACjD,OAAO,4BAAiF,CAAC;AAAA,EAEjF;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,iBAAyB;AAAA,EACzB,UAAmB;AAAA,EACnB,OAAO,IAAI,mBAAM;AAAA,EACjB;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YAAY,MAAoC;AAC9C,UAAM,KAAK,YAAY,MAAS;AAEhC,UAAM,EAAE,MAAM,qBAAqB,YAAY,gBAAgB,IAAI;AACnE,SAAK,OAAO;AACZ,SAAK,sBAAsB;AAC3B,SAAK,aAAa;AAClB,SAAK,kBAAkB;AAEvB,UAAM,kBAAkB,YAAY;AAClC,UAAI,KAAK,UAAW;AAEpB,YAAM,KAAK,oBAAoB;AAG/B,4BAAsB,4BAA4B;AAAA,QAChD;AAAA,QACA,gBAAgB,KAAK;AAAA,QACrB,SAAS,CAAC,SAAS,KAAK,uBAAuB,IAAI;AAAA,MACrD,CAAC;AAED,WAAK,YAAY,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAChE;AAEA,SAAK,sBAAsB,IAAI,oBAAa;AAE5C,SAAK,KAAK,GAAG,0BAAU,wBAAwB,CAAC,MAAM;AACpD,UAAI,KAAK,eAAe,CAAC,KAAK,oBAAoB,MAAM;AACtD,aAAK,oBAAoB,QAAQ,MAAS;AAAA,MAC5C;AAAA,IACF,CAAC;AAED,QAAI,KAAK,KAAK,aAAa;AACzB,WAAK,oBAAoB,QAAQ,MAAS;AAAA,IAC5C;AAEA,oBAAgB;AAAA,EAClB;AAAA,EAEA,MAAc,OAAO,cAA2B;AAC9C,UAAM,SAAS,MAAM,KAAK,KAAK,KAAK;AAEpC,QAAI;AACF,UAAI,KAAK,QAAS;AAElB,YAAM,KAAK,oBAAoB;AAE/B,WAAK,QAAQ;AAAA,QACX;AAAA,UACE,UAAU,KAAK;AAAA,QACjB;AAAA,QACA;AAAA,MACF;AAEA,gBAAM,iCAAmB;AAAA,QACvB,MAAM,KAAK;AAAA,QACX,UAAU,KAAK;AAAA,MACjB,CAAC;AAED,UAAI,KAAK,iBAAiB;AACxB,aAAK,QAAQ;AAAA,UACX;AAAA,YACE,UAAU,KAAK;AAAA,YACf,MAAM,KAAK;AAAA,UACb;AAAA,UACA;AAAA,QACF;AAEA,kBAAM,sCAAwB;AAAA,UAC5B,MAAM,KAAK;AAAA,UACX,UAAU,KAAK;AAAA,UACf,MAAM,KAAK;AAAA,QACb,CAAC;AAAA,MACH;AAEA,WAAK,QAAQ;AAAA,QACX;AAAA,UACE,UAAU,KAAK;AAAA,QACjB;AAAA,QACA;AAAA,MACF;AAEA,WAAK,UAAU;AAAA,IACjB,UAAE;AACA,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EAEA,MAAM,aAAa,OAAkC;AACnD,QAAI,CAAC,KAAK,WAAW;AACnB,WAAK,YAAY,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAChE;AAEA,UAAM,KAAK,UAAU;AACrB,UAAM,MAAM,aAAa,KAAK;AAE9B,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,eAAe,MAAM,KAAK,KAAK,iBAAkB,YAAY;AAAA,QAChE,UAAM,wBAAU,QAAQ;AAAA,QACxB,OAAO;AAAA,QACP,uBAAuB,CAAC,KAAK,mBAAmB;AAAA,QAChD,YAAY;AAAA,UACV,aAAa,MAAM,WAAW,SAAS;AAAA,UACvC,cAAc,MAAM,SAAS,SAAS;AAAA,QACxC;AAAA,MACF,CAAC;AACD,WAAK,iBAAiB;AAAA,IACxB;AAGA,UAAM,KAAK,aAAa,MAAM,IAAI,WAAW,MAAM,KAAK,MAAM,CAAC;AAC/D,SAAK,kBAAkB,MAAM,oBAAoB,MAAM;AAAA,EACzD;AAAA,EAEA,QAAc;AACZ,UAAM,MAAM;AAEZ,QAAI,KAAK,iBAAiB,UAAa,CAAC,KAAK,SAAS;AACpD;AAAA,IACF;AAEA,SAAK,aAAa,MAAM,EAAE,QAAQ,MAAM;AACtC,WAAK,eAAe;AAAA,IACtB,CAAC;AAAA,EACH;AAAA,EAEA,cAAoB;AAClB,QAAI,CAAC,KAAK,QAAS;AAEnB,SAAK,KAAK,iBAAkB,WAAW;AAAA,MACrC,qBAAqB,KAAK;AAAA,MAC1B,QAAQ;AAAA,MACR,SAAS;AAAA,IACX,CAAC;AAAA,EACH;AAAA,EAEQ,uBAAuB,MAAiC;AAC9D,QAAI,KAAK,mBAAmB,KAAK,qBAAqB;AACpD,WAAK,QAAQ;AAAA,QACX;AAAA,UACE,gBAAgB,KAAK;AAAA,UACrB,qBAAqB,KAAK;AAAA,QAC5B;AAAA,QACA;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAEA,SAAK,QAAQ;AAAA,MACX;AAAA,QACE,gBAAgB,KAAK;AAAA,MACvB;AAAA,MACA;AAAA,IACF;AAEA,UAAM,wBAAwB,KAAK,MAAM,KAAK,OAAO;AACrD,SAAK,mBAAmB,qBAAqB;AAC7C,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,4BAA4B;AAAA,IACjC;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAIG;AA5NL;AA6NI,0BAAsB,0BAA0B,cAAc,IAAI;AAElE,QAAI,sBAAsB,gCAAgC;AACxD;AAAA,IACF;AAEA,UAAM,aAAa,OAAO,SAA6C;AACrE,YAAMA,WAAU,sBAAsB,0BAA0B,KAAK,cAAc;AACnF,UAAI,CAACA,UAAS;AACZ,4BAAI,EAAE;AAAA,UACJ;AAAA,YACE,gBAAgB,KAAK;AAAA,YACrB,oBAAoB,OAAO,KAAK,sBAAsB,yBAAyB;AAAA,UACjF;AAAA,UACA;AAAA,QACF;AAEA,eAAO;AAAA,MACT;AACA,aAAOA,SAAQ,IAAI;AAAA,IACrB;AAEA,eAAK,qBAAL,mBAAuB,kBAAkB,uBAAuB;AAChE,0BAAsB,iCAAiC;AAAA,EACzD;AACF;","names":["handler"]}
@@ -0,0 +1,37 @@
1
+ import { type AudioFrame, type Room, type RpcInvocationData, type TrackKind } from '@livekit/rtc-node';
2
+ import { AudioOutput } from '../io.js';
3
+ export interface DataStreamAudioOutputOptions {
4
+ room: Room;
5
+ destinationIdentity: string;
6
+ sampleRate?: number;
7
+ waitRemoteTrack?: TrackKind;
8
+ }
9
+ /**
10
+ * AudioOutput implementation that streams audio to a remote avatar worker using LiveKit DataStream.
11
+ */
12
+ export declare class DataStreamAudioOutput extends AudioOutput {
13
+ #private;
14
+ static _playbackFinishedRpcRegistered: boolean;
15
+ static _playbackFinishedHandlers: Record<string, (data: RpcInvocationData) => string>;
16
+ private room;
17
+ private destinationIdentity;
18
+ private roomConnectedFuture;
19
+ private waitRemoteTrack?;
20
+ private streamWriter?;
21
+ private pushedDuration;
22
+ private started;
23
+ private lock;
24
+ private startTask?;
25
+ constructor(opts: DataStreamAudioOutputOptions);
26
+ private _start;
27
+ captureFrame(frame: AudioFrame): Promise<void>;
28
+ flush(): void;
29
+ clearBuffer(): void;
30
+ private handlePlaybackFinished;
31
+ static registerPlaybackFinishedRpc({ room, callerIdentity, handler, }: {
32
+ room: Room;
33
+ callerIdentity: string;
34
+ handler: (data: RpcInvocationData) => string;
35
+ }): void;
36
+ }
37
+ //# sourceMappingURL=datastream_io.d.ts.map
@@ -0,0 +1,37 @@
1
+ import { type AudioFrame, type Room, type RpcInvocationData, type TrackKind } from '@livekit/rtc-node';
2
+ import { AudioOutput } from '../io.js';
3
+ export interface DataStreamAudioOutputOptions {
4
+ room: Room;
5
+ destinationIdentity: string;
6
+ sampleRate?: number;
7
+ waitRemoteTrack?: TrackKind;
8
+ }
9
+ /**
10
+ * AudioOutput implementation that streams audio to a remote avatar worker using LiveKit DataStream.
11
+ */
12
+ export declare class DataStreamAudioOutput extends AudioOutput {
13
+ #private;
14
+ static _playbackFinishedRpcRegistered: boolean;
15
+ static _playbackFinishedHandlers: Record<string, (data: RpcInvocationData) => string>;
16
+ private room;
17
+ private destinationIdentity;
18
+ private roomConnectedFuture;
19
+ private waitRemoteTrack?;
20
+ private streamWriter?;
21
+ private pushedDuration;
22
+ private started;
23
+ private lock;
24
+ private startTask?;
25
+ constructor(opts: DataStreamAudioOutputOptions);
26
+ private _start;
27
+ captureFrame(frame: AudioFrame): Promise<void>;
28
+ flush(): void;
29
+ clearBuffer(): void;
30
+ private handlePlaybackFinished;
31
+ static registerPlaybackFinishedRpc({ room, callerIdentity, handler, }: {
32
+ room: Room;
33
+ callerIdentity: string;
34
+ handler: (data: RpcInvocationData) => string;
35
+ }): void;
36
+ }
37
+ //# sourceMappingURL=datastream_io.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"datastream_io.d.ts","sourceRoot":"","sources":["../../../src/voice/avatar/datastream_io.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,UAAU,EAEf,KAAK,IAAI,EAET,KAAK,iBAAiB,EACtB,KAAK,SAAS,EACf,MAAM,mBAAmB,CAAC;AAS3B,OAAO,EAAE,WAAW,EAA8B,MAAM,UAAU,CAAC;AAMnE,MAAM,WAAW,4BAA4B;IAC3C,IAAI,EAAE,IAAI,CAAC;IACX,mBAAmB,EAAE,MAAM,CAAC;IAC5B,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,eAAe,CAAC,EAAE,SAAS,CAAC;CAC7B;AAED;;GAEG;AACH,qBAAa,qBAAsB,SAAQ,WAAW;;IACpD,MAAM,CAAC,8BAA8B,EAAE,OAAO,CAAS;IACvD,MAAM,CAAC,yBAAyB,EAAE,MAAM,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,iBAAiB,KAAK,MAAM,CAAC,CAAM;IAE3F,OAAO,CAAC,IAAI,CAAO;IACnB,OAAO,CAAC,mBAAmB,CAAS;IACpC,OAAO,CAAC,mBAAmB,CAAe;IAC1C,OAAO,CAAC,eAAe,CAAC,CAAY;IACpC,OAAO,CAAC,YAAY,CAAC,CAAmB;IACxC,OAAO,CAAC,cAAc,CAAa;IACnC,OAAO,CAAC,OAAO,CAAkB;IACjC,OAAO,CAAC,IAAI,CAAe;IAC3B,OAAO,CAAC,SAAS,CAAC,CAAa;gBAInB,IAAI,EAAE,4BAA4B;YAuChC,MAAM;IAiDd,YAAY,CAAC,KAAK,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC;IA0BpD,KAAK,IAAI,IAAI;IAYb,WAAW,IAAI,IAAI;IAUnB,OAAO,CAAC,sBAAsB;IAwB9B,MAAM,CAAC,2BAA2B,CAAC,EACjC,IAAI,EACJ,cAAc,EACd,OAAO,GACR,EAAE;QACD,IAAI,EAAE,IAAI,CAAC;QACX,cAAc,EAAE,MAAM,CAAC;QACvB,OAAO,EAAE,CAAC,IAAI,EAAE,iBAAiB,KAAK,MAAM,CAAC;KAC9C;CA0BF"}
@@ -0,0 +1,188 @@
1
+ import { Mutex } from "@livekit/mutex";
2
+ import {
3
+ RoomEvent
4
+ } from "@livekit/rtc-node";
5
+ import { log } from "../../log.js";
6
+ import {
7
+ Future,
8
+ Task,
9
+ shortuuid,
10
+ waitForParticipant,
11
+ waitForTrackPublication
12
+ } from "../../utils.js";
13
+ import { AudioOutput } from "../io.js";
14
+ const RPC_CLEAR_BUFFER = "lk.clear_buffer";
15
+ const RPC_PLAYBACK_FINISHED = "lk.playback_finished";
16
+ const AUDIO_STREAM_TOPIC = "lk.audio_stream";
17
+ class DataStreamAudioOutput extends AudioOutput {
18
+ static _playbackFinishedRpcRegistered = false;
19
+ static _playbackFinishedHandlers = {};
20
+ room;
21
+ destinationIdentity;
22
+ roomConnectedFuture;
23
+ waitRemoteTrack;
24
+ streamWriter;
25
+ pushedDuration = 0;
26
+ started = false;
27
+ lock = new Mutex();
28
+ startTask;
29
+ #logger = log();
30
+ constructor(opts) {
31
+ super(opts.sampleRate, void 0);
32
+ const { room, destinationIdentity, sampleRate, waitRemoteTrack } = opts;
33
+ this.room = room;
34
+ this.destinationIdentity = destinationIdentity;
35
+ this.sampleRate = sampleRate;
36
+ this.waitRemoteTrack = waitRemoteTrack;
37
+ const onRoomConnected = async () => {
38
+ if (this.startTask) return;
39
+ await this.roomConnectedFuture.await;
40
+ DataStreamAudioOutput.registerPlaybackFinishedRpc({
41
+ room,
42
+ callerIdentity: this.destinationIdentity,
43
+ handler: (data) => this.handlePlaybackFinished(data)
44
+ });
45
+ this.startTask = Task.from(({ signal }) => this._start(signal));
46
+ };
47
+ this.roomConnectedFuture = new Future();
48
+ this.room.on(RoomEvent.ConnectionStateChanged, (_) => {
49
+ if (room.isConnected && !this.roomConnectedFuture.done) {
50
+ this.roomConnectedFuture.resolve(void 0);
51
+ }
52
+ });
53
+ if (this.room.isConnected) {
54
+ this.roomConnectedFuture.resolve(void 0);
55
+ }
56
+ onRoomConnected();
57
+ }
58
+ async _start(_abortSignal) {
59
+ const unlock = await this.lock.lock();
60
+ try {
61
+ if (this.started) return;
62
+ await this.roomConnectedFuture.await;
63
+ this.#logger.debug(
64
+ {
65
+ identity: this.destinationIdentity
66
+ },
67
+ "waiting for the remote participant"
68
+ );
69
+ await waitForParticipant({
70
+ room: this.room,
71
+ identity: this.destinationIdentity
72
+ });
73
+ if (this.waitRemoteTrack) {
74
+ this.#logger.debug(
75
+ {
76
+ identity: this.destinationIdentity,
77
+ kind: this.waitRemoteTrack
78
+ },
79
+ "waiting for the remote track"
80
+ );
81
+ await waitForTrackPublication({
82
+ room: this.room,
83
+ identity: this.destinationIdentity,
84
+ kind: this.waitRemoteTrack
85
+ });
86
+ }
87
+ this.#logger.debug(
88
+ {
89
+ identity: this.destinationIdentity
90
+ },
91
+ "remote participant ready"
92
+ );
93
+ this.started = true;
94
+ } finally {
95
+ unlock();
96
+ }
97
+ }
98
+ async captureFrame(frame) {
99
+ if (!this.startTask) {
100
+ this.startTask = Task.from(({ signal }) => this._start(signal));
101
+ }
102
+ await this.startTask.result;
103
+ await super.captureFrame(frame);
104
+ if (!this.streamWriter) {
105
+ this.streamWriter = await this.room.localParticipant.streamBytes({
106
+ name: shortuuid("AUDIO_"),
107
+ topic: AUDIO_STREAM_TOPIC,
108
+ destinationIdentities: [this.destinationIdentity],
109
+ attributes: {
110
+ sample_rate: frame.sampleRate.toString(),
111
+ num_channels: frame.channels.toString()
112
+ }
113
+ });
114
+ this.pushedDuration = 0;
115
+ }
116
+ await this.streamWriter.write(new Uint8Array(frame.data.buffer));
117
+ this.pushedDuration += frame.samplesPerChannel / frame.sampleRate;
118
+ }
119
+ flush() {
120
+ super.flush();
121
+ if (this.streamWriter === void 0 || !this.started) {
122
+ return;
123
+ }
124
+ this.streamWriter.close().finally(() => {
125
+ this.streamWriter = void 0;
126
+ });
127
+ }
128
+ clearBuffer() {
129
+ if (!this.started) return;
130
+ this.room.localParticipant.performRpc({
131
+ destinationIdentity: this.destinationIdentity,
132
+ method: RPC_CLEAR_BUFFER,
133
+ payload: ""
134
+ });
135
+ }
136
+ handlePlaybackFinished(data) {
137
+ if (data.callerIdentity !== this.destinationIdentity) {
138
+ this.#logger.warn(
139
+ {
140
+ callerIdentity: data.callerIdentity,
141
+ destinationIdentity: this.destinationIdentity
142
+ },
143
+ "playback finished event received from unexpected participant"
144
+ );
145
+ return "reject";
146
+ }
147
+ this.#logger.info(
148
+ {
149
+ callerIdentity: data.callerIdentity
150
+ },
151
+ "playback finished event received"
152
+ );
153
+ const playbackFinishedEvent = JSON.parse(data.payload);
154
+ this.onPlaybackFinished(playbackFinishedEvent);
155
+ return "ok";
156
+ }
157
+ static registerPlaybackFinishedRpc({
158
+ room,
159
+ callerIdentity,
160
+ handler
161
+ }) {
162
+ var _a;
163
+ DataStreamAudioOutput._playbackFinishedHandlers[callerIdentity] = handler;
164
+ if (DataStreamAudioOutput._playbackFinishedRpcRegistered) {
165
+ return;
166
+ }
167
+ const rpcHandler = async (data) => {
168
+ const handler2 = DataStreamAudioOutput._playbackFinishedHandlers[data.callerIdentity];
169
+ if (!handler2) {
170
+ log().warn(
171
+ {
172
+ callerIdentity: data.callerIdentity,
173
+ expectedIdentities: Object.keys(DataStreamAudioOutput._playbackFinishedHandlers)
174
+ },
175
+ "playback finished event received from unexpected participant"
176
+ );
177
+ return "reject";
178
+ }
179
+ return handler2(data);
180
+ };
181
+ (_a = room.localParticipant) == null ? void 0 : _a.registerRpcMethod(RPC_PLAYBACK_FINISHED, rpcHandler);
182
+ DataStreamAudioOutput._playbackFinishedRpcRegistered = true;
183
+ }
184
+ }
185
+ export {
186
+ DataStreamAudioOutput
187
+ };
188
+ //# sourceMappingURL=datastream_io.js.map