@livekit/agents 0.7.9 → 1.0.0-next.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/_exceptions.cjs +109 -0
- package/dist/_exceptions.cjs.map +1 -0
- package/dist/_exceptions.d.cts +64 -0
- package/dist/_exceptions.d.ts +64 -0
- package/dist/_exceptions.d.ts.map +1 -0
- package/dist/_exceptions.js +80 -0
- package/dist/_exceptions.js.map +1 -0
- package/dist/audio.cjs +10 -3
- package/dist/audio.cjs.map +1 -1
- package/dist/audio.d.cts +2 -0
- package/dist/audio.d.ts +2 -0
- package/dist/audio.d.ts.map +1 -1
- package/dist/audio.js +8 -2
- package/dist/audio.js.map +1 -1
- package/dist/cli.cjs +25 -0
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.d.ts.map +1 -1
- package/dist/cli.js +25 -0
- package/dist/cli.js.map +1 -1
- package/dist/constants.cjs +6 -3
- package/dist/constants.cjs.map +1 -1
- package/dist/constants.d.cts +2 -1
- package/dist/constants.d.ts +2 -1
- package/dist/constants.d.ts.map +1 -1
- package/dist/constants.js +4 -2
- package/dist/constants.js.map +1 -1
- package/dist/http_server.cjs.map +1 -1
- package/dist/http_server.d.cts +1 -0
- package/dist/http_server.d.ts +1 -0
- package/dist/http_server.d.ts.map +1 -1
- package/dist/http_server.js.map +1 -1
- package/dist/index.cjs +27 -20
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +13 -10
- package/dist/index.d.ts +13 -10
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +15 -11
- package/dist/index.js.map +1 -1
- package/dist/inference_runner.cjs +0 -1
- package/dist/inference_runner.cjs.map +1 -1
- package/dist/inference_runner.d.cts +2 -3
- package/dist/inference_runner.d.ts +2 -3
- package/dist/inference_runner.d.ts.map +1 -1
- package/dist/inference_runner.js +0 -1
- package/dist/inference_runner.js.map +1 -1
- package/dist/ipc/inference_proc_executor.cjs +2 -2
- package/dist/ipc/inference_proc_executor.cjs.map +1 -1
- package/dist/ipc/inference_proc_executor.js +2 -2
- package/dist/ipc/inference_proc_executor.js.map +1 -1
- package/dist/ipc/job_executor.cjs.map +1 -1
- package/dist/ipc/job_executor.js.map +1 -1
- package/dist/ipc/job_proc_executor.cjs +1 -0
- package/dist/ipc/job_proc_executor.cjs.map +1 -1
- package/dist/ipc/job_proc_executor.js +1 -0
- package/dist/ipc/job_proc_executor.js.map +1 -1
- package/dist/ipc/job_proc_lazy_main.cjs +1 -1
- package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
- package/dist/ipc/job_proc_lazy_main.js +1 -1
- package/dist/ipc/job_proc_lazy_main.js.map +1 -1
- package/dist/ipc/supervised_proc.d.cts +1 -1
- package/dist/ipc/supervised_proc.d.ts +1 -1
- package/dist/ipc/supervised_proc.d.ts.map +1 -1
- package/dist/job.cjs +14 -2
- package/dist/job.cjs.map +1 -1
- package/dist/job.d.cts +8 -0
- package/dist/job.d.ts +8 -0
- package/dist/job.d.ts.map +1 -1
- package/dist/job.js +12 -1
- package/dist/job.js.map +1 -1
- package/dist/llm/chat_context.cjs +332 -82
- package/dist/llm/chat_context.cjs.map +1 -1
- package/dist/llm/chat_context.d.cts +152 -48
- package/dist/llm/chat_context.d.ts +152 -48
- package/dist/llm/chat_context.d.ts.map +1 -1
- package/dist/llm/chat_context.js +327 -81
- package/dist/llm/chat_context.js.map +1 -1
- package/dist/llm/chat_context.test.cjs +380 -0
- package/dist/llm/chat_context.test.cjs.map +1 -0
- package/dist/llm/chat_context.test.js +385 -0
- package/dist/llm/chat_context.test.js.map +1 -0
- package/dist/llm/index.cjs +37 -8
- package/dist/llm/index.cjs.map +1 -1
- package/dist/llm/index.d.cts +7 -3
- package/dist/llm/index.d.ts +7 -3
- package/dist/llm/index.d.ts.map +1 -1
- package/dist/llm/index.js +39 -9
- package/dist/llm/index.js.map +1 -1
- package/dist/llm/llm.cjs +98 -33
- package/dist/llm/llm.cjs.map +1 -1
- package/dist/llm/llm.d.cts +50 -24
- package/dist/llm/llm.d.ts +50 -24
- package/dist/llm/llm.d.ts.map +1 -1
- package/dist/llm/llm.js +99 -33
- package/dist/llm/llm.js.map +1 -1
- package/dist/llm/provider_format/google.cjs +128 -0
- package/dist/llm/provider_format/google.cjs.map +1 -0
- package/dist/llm/provider_format/google.d.cts +6 -0
- package/dist/llm/provider_format/google.d.ts +6 -0
- package/dist/llm/provider_format/google.d.ts.map +1 -0
- package/dist/llm/provider_format/google.js +104 -0
- package/dist/llm/provider_format/google.js.map +1 -0
- package/dist/llm/provider_format/google.test.cjs +676 -0
- package/dist/llm/provider_format/google.test.cjs.map +1 -0
- package/dist/llm/provider_format/google.test.js +675 -0
- package/dist/llm/provider_format/google.test.js.map +1 -0
- package/dist/llm/provider_format/index.cjs +40 -0
- package/dist/llm/provider_format/index.cjs.map +1 -0
- package/dist/llm/provider_format/index.d.cts +4 -0
- package/dist/llm/provider_format/index.d.ts +4 -0
- package/dist/llm/provider_format/index.d.ts.map +1 -0
- package/dist/llm/provider_format/index.js +16 -0
- package/dist/llm/provider_format/index.js.map +1 -0
- package/dist/llm/provider_format/openai.cjs +116 -0
- package/dist/llm/provider_format/openai.cjs.map +1 -0
- package/dist/llm/provider_format/openai.d.cts +3 -0
- package/dist/llm/provider_format/openai.d.ts +3 -0
- package/dist/llm/provider_format/openai.d.ts.map +1 -0
- package/dist/llm/provider_format/openai.js +92 -0
- package/dist/llm/provider_format/openai.js.map +1 -0
- package/dist/llm/provider_format/openai.test.cjs +490 -0
- package/dist/llm/provider_format/openai.test.cjs.map +1 -0
- package/dist/llm/provider_format/openai.test.js +489 -0
- package/dist/llm/provider_format/openai.test.js.map +1 -0
- package/dist/llm/provider_format/utils.cjs +146 -0
- package/dist/llm/provider_format/utils.cjs.map +1 -0
- package/dist/llm/provider_format/utils.d.cts +38 -0
- package/dist/llm/provider_format/utils.d.ts +38 -0
- package/dist/llm/provider_format/utils.d.ts.map +1 -0
- package/dist/llm/provider_format/utils.js +122 -0
- package/dist/llm/provider_format/utils.js.map +1 -0
- package/dist/llm/realtime.cjs +77 -0
- package/dist/llm/realtime.cjs.map +1 -0
- package/dist/llm/realtime.d.cts +98 -0
- package/dist/llm/realtime.d.ts +98 -0
- package/dist/llm/realtime.d.ts.map +1 -0
- package/dist/llm/realtime.js +52 -0
- package/dist/llm/realtime.js.map +1 -0
- package/dist/llm/remote_chat_context.cjs +112 -0
- package/dist/llm/remote_chat_context.cjs.map +1 -0
- package/dist/llm/remote_chat_context.d.cts +23 -0
- package/dist/llm/remote_chat_context.d.ts +23 -0
- package/dist/llm/remote_chat_context.d.ts.map +1 -0
- package/dist/llm/remote_chat_context.js +88 -0
- package/dist/llm/remote_chat_context.js.map +1 -0
- package/dist/llm/remote_chat_context.test.cjs +225 -0
- package/dist/llm/remote_chat_context.test.cjs.map +1 -0
- package/dist/llm/remote_chat_context.test.js +224 -0
- package/dist/llm/remote_chat_context.test.js.map +1 -0
- package/dist/llm/tool_context.cjs +111 -0
- package/dist/llm/tool_context.cjs.map +1 -0
- package/dist/llm/tool_context.d.cts +125 -0
- package/dist/llm/tool_context.d.ts +125 -0
- package/dist/llm/tool_context.d.ts.map +1 -0
- package/dist/llm/tool_context.js +80 -0
- package/dist/llm/tool_context.js.map +1 -0
- package/dist/llm/tool_context.test.cjs +162 -0
- package/dist/llm/tool_context.test.cjs.map +1 -0
- package/dist/llm/tool_context.test.js +161 -0
- package/dist/llm/tool_context.test.js.map +1 -0
- package/dist/llm/tool_context.type.test.cjs +92 -0
- package/dist/llm/tool_context.type.test.cjs.map +1 -0
- package/dist/llm/tool_context.type.test.js +91 -0
- package/dist/llm/tool_context.type.test.js.map +1 -0
- package/dist/llm/utils.cjs +260 -0
- package/dist/llm/utils.cjs.map +1 -0
- package/dist/llm/utils.d.cts +42 -0
- package/dist/llm/utils.d.ts +42 -0
- package/dist/llm/utils.d.ts.map +1 -0
- package/dist/llm/utils.js +223 -0
- package/dist/llm/utils.js.map +1 -0
- package/dist/llm/utils.test.cjs +513 -0
- package/dist/llm/utils.test.cjs.map +1 -0
- package/dist/llm/utils.test.js +490 -0
- package/dist/llm/utils.test.js.map +1 -0
- package/dist/metrics/base.cjs +0 -27
- package/dist/metrics/base.cjs.map +1 -1
- package/dist/metrics/base.d.cts +105 -63
- package/dist/metrics/base.d.ts +105 -63
- package/dist/metrics/base.d.ts.map +1 -1
- package/dist/metrics/base.js +0 -19
- package/dist/metrics/base.js.map +1 -1
- package/dist/metrics/index.cjs +0 -3
- package/dist/metrics/index.cjs.map +1 -1
- package/dist/metrics/index.d.cts +2 -3
- package/dist/metrics/index.d.ts +2 -3
- package/dist/metrics/index.d.ts.map +1 -1
- package/dist/metrics/index.js +0 -2
- package/dist/metrics/index.js.map +1 -1
- package/dist/metrics/usage_collector.cjs +17 -12
- package/dist/metrics/usage_collector.cjs.map +1 -1
- package/dist/metrics/usage_collector.d.cts +3 -2
- package/dist/metrics/usage_collector.d.ts +3 -2
- package/dist/metrics/usage_collector.d.ts.map +1 -1
- package/dist/metrics/usage_collector.js +17 -12
- package/dist/metrics/usage_collector.js.map +1 -1
- package/dist/metrics/utils.cjs +22 -59
- package/dist/metrics/utils.cjs.map +1 -1
- package/dist/metrics/utils.d.cts +1 -8
- package/dist/metrics/utils.d.ts +1 -8
- package/dist/metrics/utils.d.ts.map +1 -1
- package/dist/metrics/utils.js +22 -52
- package/dist/metrics/utils.js.map +1 -1
- package/dist/multimodal/index.cjs +0 -2
- package/dist/multimodal/index.cjs.map +1 -1
- package/dist/multimodal/index.d.cts +0 -1
- package/dist/multimodal/index.d.ts +0 -1
- package/dist/multimodal/index.d.ts.map +1 -1
- package/dist/multimodal/index.js +0 -1
- package/dist/multimodal/index.js.map +1 -1
- package/dist/plugin.cjs +24 -8
- package/dist/plugin.cjs.map +1 -1
- package/dist/plugin.d.cts +18 -4
- package/dist/plugin.d.ts +18 -4
- package/dist/plugin.d.ts.map +1 -1
- package/dist/plugin.js +22 -7
- package/dist/plugin.js.map +1 -1
- package/dist/stream/deferred_stream.cjs +98 -0
- package/dist/stream/deferred_stream.cjs.map +1 -0
- package/dist/stream/deferred_stream.d.cts +27 -0
- package/dist/stream/deferred_stream.d.ts +27 -0
- package/dist/stream/deferred_stream.d.ts.map +1 -0
- package/dist/stream/deferred_stream.js +73 -0
- package/dist/stream/deferred_stream.js.map +1 -0
- package/dist/stream/deferred_stream.test.cjs +527 -0
- package/dist/stream/deferred_stream.test.cjs.map +1 -0
- package/dist/stream/deferred_stream.test.js +526 -0
- package/dist/stream/deferred_stream.test.js.map +1 -0
- package/dist/stream/identity_transform.cjs +42 -0
- package/dist/stream/identity_transform.cjs.map +1 -0
- package/dist/stream/identity_transform.d.cts +6 -0
- package/dist/stream/identity_transform.d.ts +6 -0
- package/dist/stream/identity_transform.d.ts.map +1 -0
- package/dist/stream/identity_transform.js +18 -0
- package/dist/stream/identity_transform.js.map +1 -0
- package/dist/stream/identity_transform.test.cjs +125 -0
- package/dist/stream/identity_transform.test.cjs.map +1 -0
- package/dist/stream/identity_transform.test.js +124 -0
- package/dist/stream/identity_transform.test.js.map +1 -0
- package/dist/stream/index.cjs +38 -0
- package/dist/stream/index.cjs.map +1 -0
- package/dist/stream/index.d.cts +5 -0
- package/dist/stream/index.d.ts +5 -0
- package/dist/stream/index.d.ts.map +1 -0
- package/dist/stream/index.js +11 -0
- package/dist/stream/index.js.map +1 -0
- package/dist/stream/merge_readable_streams.cjs +59 -0
- package/dist/stream/merge_readable_streams.cjs.map +1 -0
- package/dist/stream/merge_readable_streams.d.cts +4 -0
- package/dist/stream/merge_readable_streams.d.ts +4 -0
- package/dist/stream/merge_readable_streams.d.ts.map +1 -0
- package/dist/stream/merge_readable_streams.js +35 -0
- package/dist/stream/merge_readable_streams.js.map +1 -0
- package/dist/stream/stream_channel.cjs +47 -0
- package/dist/stream/stream_channel.cjs.map +1 -0
- package/dist/stream/stream_channel.d.cts +9 -0
- package/dist/stream/stream_channel.d.ts +9 -0
- package/dist/stream/stream_channel.d.ts.map +1 -0
- package/dist/stream/stream_channel.js +23 -0
- package/dist/stream/stream_channel.js.map +1 -0
- package/dist/stream/stream_channel.test.cjs +97 -0
- package/dist/stream/stream_channel.test.cjs.map +1 -0
- package/dist/stream/stream_channel.test.js +96 -0
- package/dist/stream/stream_channel.test.js.map +1 -0
- package/dist/stt/stream_adapter.cjs +3 -4
- package/dist/stt/stream_adapter.cjs.map +1 -1
- package/dist/stt/stream_adapter.d.cts +1 -0
- package/dist/stt/stream_adapter.d.ts +1 -0
- package/dist/stt/stream_adapter.d.ts.map +1 -1
- package/dist/stt/stream_adapter.js +3 -4
- package/dist/stt/stream_adapter.js.map +1 -1
- package/dist/stt/stt.cjs +101 -10
- package/dist/stt/stt.cjs.map +1 -1
- package/dist/stt/stt.d.cts +26 -5
- package/dist/stt/stt.d.ts +26 -5
- package/dist/stt/stt.d.ts.map +1 -1
- package/dist/stt/stt.js +102 -11
- package/dist/stt/stt.js.map +1 -1
- package/dist/tokenize/basic/basic.cjs +10 -5
- package/dist/tokenize/basic/basic.cjs.map +1 -1
- package/dist/tokenize/basic/basic.d.cts +7 -1
- package/dist/tokenize/basic/basic.d.ts +7 -1
- package/dist/tokenize/basic/basic.d.ts.map +1 -1
- package/dist/tokenize/basic/basic.js +10 -5
- package/dist/tokenize/basic/basic.js.map +1 -1
- package/dist/tokenize/basic/sentence.cjs +14 -6
- package/dist/tokenize/basic/sentence.cjs.map +1 -1
- package/dist/tokenize/basic/sentence.d.cts +1 -1
- package/dist/tokenize/basic/sentence.d.ts +1 -1
- package/dist/tokenize/basic/sentence.d.ts.map +1 -1
- package/dist/tokenize/basic/sentence.js +14 -6
- package/dist/tokenize/basic/sentence.js.map +1 -1
- package/dist/tokenize/token_stream.cjs +5 -3
- package/dist/tokenize/token_stream.cjs.map +1 -1
- package/dist/tokenize/token_stream.d.cts +1 -0
- package/dist/tokenize/token_stream.d.ts +1 -0
- package/dist/tokenize/token_stream.d.ts.map +1 -1
- package/dist/tokenize/token_stream.js +6 -4
- package/dist/tokenize/token_stream.js.map +1 -1
- package/dist/transcription.cjs +1 -2
- package/dist/transcription.cjs.map +1 -1
- package/dist/transcription.d.ts.map +1 -1
- package/dist/transcription.js +2 -3
- package/dist/transcription.js.map +1 -1
- package/dist/tts/index.cjs +2 -4
- package/dist/tts/index.cjs.map +1 -1
- package/dist/tts/index.d.cts +1 -1
- package/dist/tts/index.d.ts +1 -1
- package/dist/tts/index.d.ts.map +1 -1
- package/dist/tts/index.js +1 -3
- package/dist/tts/index.js.map +1 -1
- package/dist/tts/stream_adapter.cjs +26 -13
- package/dist/tts/stream_adapter.cjs.map +1 -1
- package/dist/tts/stream_adapter.d.cts +1 -1
- package/dist/tts/stream_adapter.d.ts +1 -1
- package/dist/tts/stream_adapter.d.ts.map +1 -1
- package/dist/tts/stream_adapter.js +27 -14
- package/dist/tts/stream_adapter.js.map +1 -1
- package/dist/tts/tts.cjs +157 -25
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.cts +29 -5
- package/dist/tts/tts.d.ts +29 -5
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +157 -24
- package/dist/tts/tts.js.map +1 -1
- package/dist/types.cjs +60 -0
- package/dist/types.cjs.map +1 -0
- package/dist/types.d.cts +13 -0
- package/dist/types.d.ts +13 -0
- package/dist/types.d.ts.map +1 -0
- package/dist/types.js +35 -0
- package/dist/types.js.map +1 -0
- package/dist/utils.cjs +281 -27
- package/dist/utils.cjs.map +1 -1
- package/dist/utils.d.cts +134 -9
- package/dist/utils.d.ts +134 -9
- package/dist/utils.d.ts.map +1 -1
- package/dist/utils.js +265 -26
- package/dist/utils.js.map +1 -1
- package/dist/utils.test.cjs +492 -0
- package/dist/utils.test.cjs.map +1 -0
- package/dist/utils.test.js +498 -0
- package/dist/utils.test.js.map +1 -0
- package/dist/vad.cjs +76 -20
- package/dist/vad.cjs.map +1 -1
- package/dist/vad.d.cts +25 -5
- package/dist/vad.d.ts +25 -5
- package/dist/vad.d.ts.map +1 -1
- package/dist/vad.js +76 -20
- package/dist/vad.js.map +1 -1
- package/dist/voice/agent.cjs +245 -0
- package/dist/voice/agent.cjs.map +1 -0
- package/dist/voice/agent.d.cts +78 -0
- package/dist/voice/agent.d.ts +78 -0
- package/dist/voice/agent.d.ts.map +1 -0
- package/dist/voice/agent.js +220 -0
- package/dist/voice/agent.js.map +1 -0
- package/dist/voice/agent.test.cjs +61 -0
- package/dist/voice/agent.test.cjs.map +1 -0
- package/dist/voice/agent.test.js +60 -0
- package/dist/voice/agent.test.js.map +1 -0
- package/dist/voice/agent_activity.cjs +1453 -0
- package/dist/voice/agent_activity.cjs.map +1 -0
- package/dist/voice/agent_activity.d.cts +94 -0
- package/dist/voice/agent_activity.d.ts +94 -0
- package/dist/voice/agent_activity.d.ts.map +1 -0
- package/dist/voice/agent_activity.js +1449 -0
- package/dist/voice/agent_activity.js.map +1 -0
- package/dist/voice/agent_session.cjs +312 -0
- package/dist/voice/agent_session.cjs.map +1 -0
- package/dist/voice/agent_session.d.cts +121 -0
- package/dist/voice/agent_session.d.ts +121 -0
- package/dist/voice/agent_session.d.ts.map +1 -0
- package/dist/voice/agent_session.js +295 -0
- package/dist/voice/agent_session.js.map +1 -0
- package/dist/voice/audio_recognition.cjs +375 -0
- package/dist/voice/audio_recognition.cjs.map +1 -0
- package/dist/voice/audio_recognition.d.cts +80 -0
- package/dist/voice/audio_recognition.d.ts +80 -0
- package/dist/voice/audio_recognition.d.ts.map +1 -0
- package/dist/voice/audio_recognition.js +351 -0
- package/dist/voice/audio_recognition.js.map +1 -0
- package/dist/voice/events.cjs +145 -0
- package/dist/voice/events.cjs.map +1 -0
- package/dist/voice/events.d.cts +124 -0
- package/dist/voice/events.d.ts +124 -0
- package/dist/voice/events.d.ts.map +1 -0
- package/dist/voice/events.js +110 -0
- package/dist/voice/events.js.map +1 -0
- package/dist/voice/generation.cjs +700 -0
- package/dist/voice/generation.cjs.map +1 -0
- package/dist/voice/generation.d.cts +115 -0
- package/dist/voice/generation.d.ts +115 -0
- package/dist/voice/generation.d.ts.map +1 -0
- package/dist/voice/generation.js +672 -0
- package/dist/voice/generation.js.map +1 -0
- package/dist/voice/index.cjs +40 -0
- package/dist/voice/index.cjs.map +1 -0
- package/dist/voice/index.d.cts +5 -0
- package/dist/voice/index.d.ts +5 -0
- package/dist/voice/index.d.ts.map +1 -0
- package/dist/voice/index.js +11 -0
- package/dist/voice/index.js.map +1 -0
- package/dist/voice/io.cjs +245 -0
- package/dist/voice/io.cjs.map +1 -0
- package/dist/voice/io.d.cts +101 -0
- package/dist/voice/io.d.ts +101 -0
- package/dist/voice/io.d.ts.map +1 -0
- package/dist/voice/io.js +217 -0
- package/dist/voice/io.js.map +1 -0
- package/dist/voice/room_io/_input.cjs +121 -0
- package/dist/voice/room_io/_input.cjs.map +1 -0
- package/dist/voice/room_io/_input.d.cts +24 -0
- package/dist/voice/room_io/_input.d.ts +24 -0
- package/dist/voice/room_io/_input.d.ts.map +1 -0
- package/dist/voice/room_io/_input.js +102 -0
- package/dist/voice/room_io/_input.js.map +1 -0
- package/dist/voice/room_io/_output.cjs +358 -0
- package/dist/voice/room_io/_output.cjs.map +1 -0
- package/dist/voice/room_io/_output.d.cts +75 -0
- package/dist/voice/room_io/_output.d.ts +75 -0
- package/dist/voice/room_io/_output.d.ts.map +1 -0
- package/dist/voice/room_io/_output.js +342 -0
- package/dist/voice/room_io/_output.js.map +1 -0
- package/dist/voice/room_io/index.cjs +25 -0
- package/dist/voice/room_io/index.cjs.map +1 -0
- package/dist/voice/room_io/index.d.cts +3 -0
- package/dist/voice/room_io/index.d.ts +3 -0
- package/dist/voice/room_io/index.d.ts.map +1 -0
- package/dist/voice/room_io/index.js +3 -0
- package/dist/voice/room_io/index.js.map +1 -0
- package/dist/voice/room_io/room_io.cjs +370 -0
- package/dist/voice/room_io/room_io.cjs.map +1 -0
- package/dist/voice/room_io/room_io.d.cts +73 -0
- package/dist/voice/room_io/room_io.d.ts +73 -0
- package/dist/voice/room_io/room_io.d.ts.map +1 -0
- package/dist/voice/room_io/room_io.js +361 -0
- package/dist/voice/room_io/room_io.js.map +1 -0
- package/dist/{pipeline/index.cjs → voice/run_context.cjs} +16 -11
- package/dist/voice/run_context.cjs.map +1 -0
- package/dist/voice/run_context.d.cts +12 -0
- package/dist/voice/run_context.d.ts +12 -0
- package/dist/voice/run_context.d.ts.map +1 -0
- package/dist/voice/run_context.js +14 -0
- package/dist/voice/run_context.js.map +1 -0
- package/dist/voice/speech_handle.cjs +105 -0
- package/dist/voice/speech_handle.cjs.map +1 -0
- package/dist/voice/speech_handle.d.cts +46 -0
- package/dist/voice/speech_handle.d.ts +46 -0
- package/dist/voice/speech_handle.d.ts.map +1 -0
- package/dist/voice/speech_handle.js +81 -0
- package/dist/voice/speech_handle.js.map +1 -0
- package/dist/voice/transcription/_utils.cjs +45 -0
- package/dist/voice/transcription/_utils.cjs.map +1 -0
- package/dist/voice/transcription/_utils.d.cts +3 -0
- package/dist/voice/transcription/_utils.d.ts +3 -0
- package/dist/voice/transcription/_utils.d.ts.map +1 -0
- package/dist/voice/transcription/_utils.js +21 -0
- package/dist/voice/transcription/_utils.js.map +1 -0
- package/dist/voice/transcription/index.cjs +23 -0
- package/dist/voice/transcription/index.cjs.map +1 -0
- package/dist/voice/transcription/index.d.cts +2 -0
- package/dist/voice/transcription/index.d.ts +2 -0
- package/dist/voice/transcription/index.d.ts.map +1 -0
- package/dist/voice/transcription/index.js +2 -0
- package/dist/voice/transcription/index.js.map +1 -0
- package/dist/voice/transcription/synchronizer.cjs +380 -0
- package/dist/voice/transcription/synchronizer.cjs.map +1 -0
- package/dist/voice/transcription/synchronizer.d.cts +86 -0
- package/dist/voice/transcription/synchronizer.d.ts +86 -0
- package/dist/voice/transcription/synchronizer.d.ts.map +1 -0
- package/dist/voice/transcription/synchronizer.js +355 -0
- package/dist/voice/transcription/synchronizer.js.map +1 -0
- package/dist/worker.cjs +22 -4
- package/dist/worker.cjs.map +1 -1
- package/dist/worker.d.cts +1 -1
- package/dist/worker.d.ts +1 -1
- package/dist/worker.d.ts.map +1 -1
- package/dist/worker.js +22 -4
- package/dist/worker.js.map +1 -1
- package/package.json +9 -2
- package/src/_exceptions.ts +137 -0
- package/src/audio.ts +12 -1
- package/src/cli.ts +37 -0
- package/src/constants.ts +2 -1
- package/src/http_server.ts +1 -0
- package/src/index.ts +13 -10
- package/src/inference_runner.ts +2 -3
- package/src/ipc/inference_proc_executor.ts +2 -2
- package/src/ipc/job_executor.ts +1 -1
- package/src/ipc/job_proc_executor.ts +1 -1
- package/src/ipc/job_proc_lazy_main.ts +1 -1
- package/src/job.ts +18 -0
- package/src/llm/__snapshots__/chat_context.test.ts.snap +527 -0
- package/src/llm/__snapshots__/tool_context.test.ts.snap +177 -0
- package/src/llm/__snapshots__/utils.test.ts.snap +65 -0
- package/src/llm/chat_context.test.ts +450 -0
- package/src/llm/chat_context.ts +501 -103
- package/src/llm/index.ts +53 -18
- package/src/llm/llm.ts +149 -50
- package/src/llm/provider_format/google.test.ts +772 -0
- package/src/llm/provider_format/google.ts +130 -0
- package/src/llm/provider_format/index.ts +23 -0
- package/src/llm/provider_format/openai.test.ts +581 -0
- package/src/llm/provider_format/openai.ts +118 -0
- package/src/llm/provider_format/utils.ts +183 -0
- package/src/llm/realtime.ts +151 -0
- package/src/llm/remote_chat_context.test.ts +290 -0
- package/src/llm/remote_chat_context.ts +114 -0
- package/src/llm/tool_context.test.ts +198 -0
- package/src/llm/tool_context.ts +259 -0
- package/src/llm/tool_context.type.test.ts +115 -0
- package/src/llm/utils.test.ts +670 -0
- package/src/llm/utils.ts +324 -0
- package/src/metrics/base.ts +110 -78
- package/src/metrics/index.ts +3 -9
- package/src/metrics/usage_collector.ts +19 -13
- package/src/metrics/utils.ts +24 -69
- package/src/multimodal/index.ts +0 -1
- package/src/plugin.ts +26 -8
- package/src/stream/deferred_stream.test.ts +755 -0
- package/src/stream/deferred_stream.ts +110 -0
- package/src/stream/identity_transform.test.ts +179 -0
- package/src/stream/identity_transform.ts +18 -0
- package/src/stream/index.ts +7 -0
- package/src/stream/merge_readable_streams.ts +40 -0
- package/src/stream/stream_channel.test.ts +129 -0
- package/src/stream/stream_channel.ts +32 -0
- package/src/stt/stream_adapter.ts +3 -5
- package/src/stt/stt.ts +135 -17
- package/src/tokenize/basic/basic.ts +13 -5
- package/src/tokenize/basic/sentence.ts +20 -6
- package/src/tokenize/token_stream.ts +7 -4
- package/src/transcription.ts +2 -3
- package/src/tts/index.ts +0 -1
- package/src/tts/stream_adapter.ts +42 -16
- package/src/tts/tts.ts +203 -21
- package/src/types.ts +42 -0
- package/src/utils.test.ts +658 -0
- package/src/utils.ts +375 -44
- package/src/vad.ts +90 -22
- package/src/voice/agent.test.ts +80 -0
- package/src/voice/agent.ts +332 -0
- package/src/voice/agent_activity.ts +1913 -0
- package/src/voice/agent_session.ts +460 -0
- package/src/voice/audio_recognition.ts +474 -0
- package/src/voice/events.ts +252 -0
- package/src/voice/generation.ts +881 -0
- package/src/voice/index.ts +7 -0
- package/src/voice/io.ts +304 -0
- package/src/voice/room_io/_input.ts +144 -0
- package/src/voice/room_io/_output.ts +436 -0
- package/src/voice/room_io/index.ts +5 -0
- package/src/voice/room_io/room_io.ts +495 -0
- package/src/voice/run_context.ts +20 -0
- package/src/voice/speech_handle.ts +104 -0
- package/src/voice/transcription/_utils.ts +25 -0
- package/src/voice/transcription/index.ts +4 -0
- package/src/voice/transcription/synchronizer.ts +478 -0
- package/src/worker.ts +22 -2
- package/dist/llm/function_context.cjs +0 -103
- package/dist/llm/function_context.cjs.map +0 -1
- package/dist/llm/function_context.d.cts +0 -47
- package/dist/llm/function_context.d.ts +0 -47
- package/dist/llm/function_context.d.ts.map +0 -1
- package/dist/llm/function_context.js +0 -78
- package/dist/llm/function_context.js.map +0 -1
- package/dist/llm/function_context.test.cjs +0 -218
- package/dist/llm/function_context.test.cjs.map +0 -1
- package/dist/llm/function_context.test.js +0 -217
- package/dist/llm/function_context.test.js.map +0 -1
- package/dist/multimodal/multimodal_agent.cjs +0 -486
- package/dist/multimodal/multimodal_agent.cjs.map +0 -1
- package/dist/multimodal/multimodal_agent.d.cts +0 -48
- package/dist/multimodal/multimodal_agent.d.ts +0 -48
- package/dist/multimodal/multimodal_agent.d.ts.map +0 -1
- package/dist/multimodal/multimodal_agent.js +0 -461
- package/dist/multimodal/multimodal_agent.js.map +0 -1
- package/dist/pipeline/agent_output.cjs +0 -197
- package/dist/pipeline/agent_output.cjs.map +0 -1
- package/dist/pipeline/agent_output.d.cts +0 -33
- package/dist/pipeline/agent_output.d.ts +0 -33
- package/dist/pipeline/agent_output.d.ts.map +0 -1
- package/dist/pipeline/agent_output.js +0 -172
- package/dist/pipeline/agent_output.js.map +0 -1
- package/dist/pipeline/agent_playout.cjs +0 -175
- package/dist/pipeline/agent_playout.cjs.map +0 -1
- package/dist/pipeline/agent_playout.d.cts +0 -40
- package/dist/pipeline/agent_playout.d.ts +0 -40
- package/dist/pipeline/agent_playout.d.ts.map +0 -1
- package/dist/pipeline/agent_playout.js +0 -139
- package/dist/pipeline/agent_playout.js.map +0 -1
- package/dist/pipeline/human_input.cjs +0 -171
- package/dist/pipeline/human_input.cjs.map +0 -1
- package/dist/pipeline/human_input.d.cts +0 -30
- package/dist/pipeline/human_input.d.ts +0 -30
- package/dist/pipeline/human_input.d.ts.map +0 -1
- package/dist/pipeline/human_input.js +0 -146
- package/dist/pipeline/human_input.js.map +0 -1
- package/dist/pipeline/index.cjs.map +0 -1
- package/dist/pipeline/index.d.cts +0 -2
- package/dist/pipeline/index.d.ts +0 -2
- package/dist/pipeline/index.d.ts.map +0 -1
- package/dist/pipeline/index.js +0 -11
- package/dist/pipeline/index.js.map +0 -1
- package/dist/pipeline/pipeline_agent.cjs +0 -859
- package/dist/pipeline/pipeline_agent.cjs.map +0 -1
- package/dist/pipeline/pipeline_agent.d.cts +0 -150
- package/dist/pipeline/pipeline_agent.d.ts +0 -150
- package/dist/pipeline/pipeline_agent.d.ts.map +0 -1
- package/dist/pipeline/pipeline_agent.js +0 -837
- package/dist/pipeline/pipeline_agent.js.map +0 -1
- package/dist/pipeline/speech_handle.cjs +0 -176
- package/dist/pipeline/speech_handle.cjs.map +0 -1
- package/dist/pipeline/speech_handle.d.cts +0 -37
- package/dist/pipeline/speech_handle.d.ts +0 -37
- package/dist/pipeline/speech_handle.d.ts.map +0 -1
- package/dist/pipeline/speech_handle.js +0 -152
- package/dist/pipeline/speech_handle.js.map +0 -1
- package/src/llm/function_context.test.ts +0 -248
- package/src/llm/function_context.ts +0 -142
- package/src/multimodal/multimodal_agent.ts +0 -592
- package/src/pipeline/agent_output.ts +0 -219
- package/src/pipeline/agent_playout.ts +0 -192
- package/src/pipeline/human_input.ts +0 -188
- package/src/pipeline/index.ts +0 -15
- package/src/pipeline/pipeline_agent.ts +0 -1197
- package/src/pipeline/speech_handle.ts +0 -201
|
@@ -0,0 +1,351 @@
|
|
|
1
|
+
import { AudioFrame } from "@livekit/rtc-node";
|
|
2
|
+
import { delay } from "@std/async";
|
|
3
|
+
import { ReadableStream } from "node:stream/web";
|
|
4
|
+
import {} from "../llm/chat_context.js";
|
|
5
|
+
import { log } from "../log.js";
|
|
6
|
+
import { DeferredReadableStream, isStreamReaderReleaseError } from "../stream/deferred_stream.js";
|
|
7
|
+
import { IdentityTransform } from "../stream/identity_transform.js";
|
|
8
|
+
import { mergeReadableStreams } from "../stream/merge_readable_streams.js";
|
|
9
|
+
import { SpeechEventType } from "../stt/stt.js";
|
|
10
|
+
import { Task } from "../utils.js";
|
|
11
|
+
import { VADEventType } from "../vad.js";
|
|
12
|
+
class AudioRecognition {
|
|
13
|
+
hooks;
|
|
14
|
+
stt;
|
|
15
|
+
vad;
|
|
16
|
+
turnDetector;
|
|
17
|
+
turnDetectionMode;
|
|
18
|
+
minEndpointingDelay;
|
|
19
|
+
maxEndpointingDelay;
|
|
20
|
+
lastLanguage;
|
|
21
|
+
deferredInputStream;
|
|
22
|
+
logger = log();
|
|
23
|
+
lastFinalTranscriptTime = 0;
|
|
24
|
+
audioTranscript = "";
|
|
25
|
+
audioInterimTranscript = "";
|
|
26
|
+
lastSpeakingTime = 0;
|
|
27
|
+
userTurnCommitted = false;
|
|
28
|
+
speaking = false;
|
|
29
|
+
sampleRate;
|
|
30
|
+
vadInputStream;
|
|
31
|
+
sttInputStream;
|
|
32
|
+
silenceAudioTransform = new IdentityTransform();
|
|
33
|
+
silenceAudioWriter;
|
|
34
|
+
// all cancellable tasks
|
|
35
|
+
bounceEOUTask;
|
|
36
|
+
commitUserTurnTask;
|
|
37
|
+
vadTask;
|
|
38
|
+
sttTask;
|
|
39
|
+
constructor(opts) {
|
|
40
|
+
this.hooks = opts.recognitionHooks;
|
|
41
|
+
this.stt = opts.stt;
|
|
42
|
+
this.vad = opts.vad;
|
|
43
|
+
this.turnDetector = opts.turnDetector;
|
|
44
|
+
this.turnDetectionMode = opts.turnDetectionMode;
|
|
45
|
+
this.minEndpointingDelay = opts.minEndpointingDelay;
|
|
46
|
+
this.maxEndpointingDelay = opts.maxEndpointingDelay;
|
|
47
|
+
this.lastLanguage = void 0;
|
|
48
|
+
this.deferredInputStream = new DeferredReadableStream();
|
|
49
|
+
const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();
|
|
50
|
+
this.vadInputStream = vadInputStream;
|
|
51
|
+
this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);
|
|
52
|
+
this.silenceAudioWriter = this.silenceAudioTransform.writable.getWriter();
|
|
53
|
+
}
|
|
54
|
+
/**
|
|
55
|
+
* Current transcript of the user's speech, including interim transcript if available.
|
|
56
|
+
*/
|
|
57
|
+
get currentTranscript() {
|
|
58
|
+
if (this.audioInterimTranscript) {
|
|
59
|
+
return `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();
|
|
60
|
+
}
|
|
61
|
+
return this.audioTranscript;
|
|
62
|
+
}
|
|
63
|
+
async start() {
|
|
64
|
+
this.vadTask = Task.from(({ signal }) => this.createVadTask(this.vad, signal));
|
|
65
|
+
this.vadTask.result.catch((err) => {
|
|
66
|
+
this.logger.error(`Error running VAD task: ${err}`);
|
|
67
|
+
});
|
|
68
|
+
this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));
|
|
69
|
+
this.sttTask.result.catch((err) => {
|
|
70
|
+
this.logger.error(`Error running STT task: ${err}`);
|
|
71
|
+
});
|
|
72
|
+
}
|
|
73
|
+
async onSTTEvent(ev) {
|
|
74
|
+
var _a, _b, _c, _d, _e, _f, _g, _h, _i;
|
|
75
|
+
if (this.turnDetectionMode === "manual" && this.userTurnCommitted && (this.bounceEOUTask === void 0 || this.bounceEOUTask.done || ev.type == SpeechEventType.INTERIM_TRANSCRIPT)) {
|
|
76
|
+
this.logger.debug(
|
|
77
|
+
{
|
|
78
|
+
userTurnCommitted: this.userTurnCommitted,
|
|
79
|
+
eouTaskDone: (_a = this.bounceEOUTask) == null ? void 0 : _a.done,
|
|
80
|
+
evType: ev.type,
|
|
81
|
+
turnDetectionMode: this.turnDetectionMode
|
|
82
|
+
},
|
|
83
|
+
"ignoring stt event"
|
|
84
|
+
);
|
|
85
|
+
return;
|
|
86
|
+
}
|
|
87
|
+
switch (ev.type) {
|
|
88
|
+
case SpeechEventType.FINAL_TRANSCRIPT:
|
|
89
|
+
this.hooks.onFinalTranscript(ev);
|
|
90
|
+
const transcript = (_c = (_b = ev.alternatives) == null ? void 0 : _b[0]) == null ? void 0 : _c.text;
|
|
91
|
+
this.lastLanguage = (_e = (_d = ev.alternatives) == null ? void 0 : _d[0]) == null ? void 0 : _e.language;
|
|
92
|
+
if (!transcript) {
|
|
93
|
+
return;
|
|
94
|
+
}
|
|
95
|
+
this.logger.debug(
|
|
96
|
+
{
|
|
97
|
+
user_transcript: transcript,
|
|
98
|
+
language: this.lastLanguage
|
|
99
|
+
},
|
|
100
|
+
"received user transcript"
|
|
101
|
+
);
|
|
102
|
+
this.lastFinalTranscriptTime = Date.now();
|
|
103
|
+
this.audioTranscript += ` ${transcript}`;
|
|
104
|
+
this.audioTranscript = this.audioTranscript.trimStart();
|
|
105
|
+
this.audioInterimTranscript = "";
|
|
106
|
+
if (!this.speaking) {
|
|
107
|
+
if (!this.vad) {
|
|
108
|
+
this.lastSpeakingTime = Date.now();
|
|
109
|
+
}
|
|
110
|
+
if (this.vadBaseTurnDetection || this.userTurnCommitted) {
|
|
111
|
+
const chatCtx = this.hooks.retrieveChatCtx();
|
|
112
|
+
this.logger.debug("running EOU detection on stt FINAL_TRANSCRIPT");
|
|
113
|
+
this.runEOUDetection(chatCtx);
|
|
114
|
+
}
|
|
115
|
+
}
|
|
116
|
+
break;
|
|
117
|
+
case SpeechEventType.INTERIM_TRANSCRIPT:
|
|
118
|
+
this.logger.debug({ transcript: (_g = (_f = ev.alternatives) == null ? void 0 : _f[0]) == null ? void 0 : _g.text }, "interim transcript");
|
|
119
|
+
this.hooks.onInterimTranscript(ev);
|
|
120
|
+
this.audioInterimTranscript = ((_i = (_h = ev.alternatives) == null ? void 0 : _h[0]) == null ? void 0 : _i.text) ?? "";
|
|
121
|
+
break;
|
|
122
|
+
case SpeechEventType.END_OF_SPEECH:
|
|
123
|
+
if (this.turnDetectionMode !== "stt") break;
|
|
124
|
+
this.userTurnCommitted = true;
|
|
125
|
+
if (!this.speaking) {
|
|
126
|
+
const chatCtx = this.hooks.retrieveChatCtx();
|
|
127
|
+
this.logger.debug("running EOU detection on stt END_OF_SPEECH");
|
|
128
|
+
this.runEOUDetection(chatCtx);
|
|
129
|
+
}
|
|
130
|
+
}
|
|
131
|
+
}
|
|
132
|
+
runEOUDetection(chatCtx) {
|
|
133
|
+
var _a;
|
|
134
|
+
this.logger.debug(
|
|
135
|
+
{
|
|
136
|
+
stt: this.stt,
|
|
137
|
+
audioTranscript: this.audioTranscript,
|
|
138
|
+
turnDetectionMode: this.turnDetectionMode
|
|
139
|
+
},
|
|
140
|
+
"running EOU detection"
|
|
141
|
+
);
|
|
142
|
+
if (this.stt && !this.audioTranscript && this.turnDetectionMode !== "manual") {
|
|
143
|
+
this.logger.debug("skipping EOU detection");
|
|
144
|
+
return;
|
|
145
|
+
}
|
|
146
|
+
chatCtx = chatCtx.copy();
|
|
147
|
+
chatCtx.addMessage({ role: "user", content: this.audioTranscript });
|
|
148
|
+
const turnDetector = (
|
|
149
|
+
// disable EOU model if manual turn detection enabled
|
|
150
|
+
this.audioTranscript && this.turnDetectionMode !== "manual" ? this.turnDetector : void 0
|
|
151
|
+
);
|
|
152
|
+
const bounceEOUTask = (lastSpeakingTime) => async (controller) => {
|
|
153
|
+
let endpointingDelay = this.minEndpointingDelay;
|
|
154
|
+
if (turnDetector) {
|
|
155
|
+
this.logger.debug("Running turn detector model");
|
|
156
|
+
if (!turnDetector.supportsLanguage(this.lastLanguage)) {
|
|
157
|
+
this.logger.debug(`Turn detector does not support language ${this.lastLanguage}`);
|
|
158
|
+
} else {
|
|
159
|
+
const endOfTurnProbability = await turnDetector.predictEndOfTurn(chatCtx);
|
|
160
|
+
this.logger.debug(
|
|
161
|
+
{ endOfTurnProbability, language: this.lastLanguage },
|
|
162
|
+
"end of turn probability"
|
|
163
|
+
);
|
|
164
|
+
const unlikelyThreshold = await turnDetector.unlikelyThreshold(this.lastLanguage);
|
|
165
|
+
this.logger.debug(
|
|
166
|
+
{
|
|
167
|
+
unlikelyThreshold,
|
|
168
|
+
endOfTurnProbability,
|
|
169
|
+
language: this.lastLanguage,
|
|
170
|
+
transcript: this.audioTranscript
|
|
171
|
+
},
|
|
172
|
+
"EOU Detection"
|
|
173
|
+
);
|
|
174
|
+
if (unlikelyThreshold && endOfTurnProbability < unlikelyThreshold) {
|
|
175
|
+
endpointingDelay = this.maxEndpointingDelay;
|
|
176
|
+
}
|
|
177
|
+
}
|
|
178
|
+
}
|
|
179
|
+
const extraSleep = lastSpeakingTime + endpointingDelay - Date.now();
|
|
180
|
+
await delay(Math.max(extraSleep, 0), { signal: controller.signal });
|
|
181
|
+
this.logger.debug({ transcript: this.audioTranscript }, "end of user turn");
|
|
182
|
+
const committed = await this.hooks.onEndOfTurn({
|
|
183
|
+
newTranscript: this.audioTranscript,
|
|
184
|
+
transcriptionDelay: Math.max(this.lastFinalTranscriptTime - lastSpeakingTime, 0),
|
|
185
|
+
endOfUtteranceDelay: Date.now() - lastSpeakingTime
|
|
186
|
+
});
|
|
187
|
+
if (committed) {
|
|
188
|
+
this.audioTranscript = "";
|
|
189
|
+
}
|
|
190
|
+
this.userTurnCommitted = false;
|
|
191
|
+
};
|
|
192
|
+
(_a = this.bounceEOUTask) == null ? void 0 : _a.cancel();
|
|
193
|
+
this.bounceEOUTask = Task.from(bounceEOUTask(this.lastSpeakingTime));
|
|
194
|
+
this.bounceEOUTask.result.then(() => {
|
|
195
|
+
this.logger.debug("EOU detection task completed");
|
|
196
|
+
}).catch((err) => {
|
|
197
|
+
if (err instanceof Error && err.message.includes("This operation was aborted")) {
|
|
198
|
+
return;
|
|
199
|
+
}
|
|
200
|
+
this.logger.error(err, "Error in EOU detection task:");
|
|
201
|
+
});
|
|
202
|
+
}
|
|
203
|
+
async createSttTask(stt, signal) {
|
|
204
|
+
if (!stt) return;
|
|
205
|
+
this.logger.debug("createSttTask: create stt stream from stt node");
|
|
206
|
+
const sttStream = await stt(this.sttInputStream, {});
|
|
207
|
+
if (signal.aborted || sttStream === null) return;
|
|
208
|
+
if (sttStream instanceof ReadableStream) {
|
|
209
|
+
const reader = sttStream.getReader();
|
|
210
|
+
signal.addEventListener("abort", async () => {
|
|
211
|
+
try {
|
|
212
|
+
reader.releaseLock();
|
|
213
|
+
await (sttStream == null ? void 0 : sttStream.cancel());
|
|
214
|
+
} catch (e) {
|
|
215
|
+
this.logger.debug("createSttTask: error during abort handler:", e);
|
|
216
|
+
}
|
|
217
|
+
});
|
|
218
|
+
try {
|
|
219
|
+
while (true) {
|
|
220
|
+
if (signal.aborted) break;
|
|
221
|
+
const { done, value: ev } = await reader.read();
|
|
222
|
+
if (done) break;
|
|
223
|
+
if (typeof ev === "string") {
|
|
224
|
+
throw new Error("STT node must yield SpeechEvent");
|
|
225
|
+
} else {
|
|
226
|
+
await this.onSTTEvent(ev);
|
|
227
|
+
}
|
|
228
|
+
}
|
|
229
|
+
} catch (e) {
|
|
230
|
+
if (isStreamReaderReleaseError(e)) {
|
|
231
|
+
return;
|
|
232
|
+
}
|
|
233
|
+
this.logger.error({ error: e }, "createSttTask: error reading sttStream");
|
|
234
|
+
} finally {
|
|
235
|
+
reader.releaseLock();
|
|
236
|
+
try {
|
|
237
|
+
await sttStream.cancel();
|
|
238
|
+
} catch (e) {
|
|
239
|
+
this.logger.debug(
|
|
240
|
+
"createSttTask: error cancelling sttStream (may already be cancelled):",
|
|
241
|
+
e
|
|
242
|
+
);
|
|
243
|
+
}
|
|
244
|
+
}
|
|
245
|
+
}
|
|
246
|
+
}
|
|
247
|
+
async createVadTask(vad, signal) {
|
|
248
|
+
var _a;
|
|
249
|
+
if (!vad) return;
|
|
250
|
+
const vadStream = vad.stream();
|
|
251
|
+
vadStream.updateInputStream(this.vadInputStream);
|
|
252
|
+
const abortHandler = () => {
|
|
253
|
+
vadStream.detachInputStream();
|
|
254
|
+
vadStream.close();
|
|
255
|
+
signal.removeEventListener("abort", abortHandler);
|
|
256
|
+
};
|
|
257
|
+
signal.addEventListener("abort", abortHandler);
|
|
258
|
+
try {
|
|
259
|
+
for await (const ev of vadStream) {
|
|
260
|
+
if (signal.aborted) break;
|
|
261
|
+
switch (ev.type) {
|
|
262
|
+
case VADEventType.START_OF_SPEECH:
|
|
263
|
+
this.logger.debug("VAD task: START_OF_SPEECH");
|
|
264
|
+
this.hooks.onStartOfSpeech(ev);
|
|
265
|
+
this.speaking = true;
|
|
266
|
+
(_a = this.bounceEOUTask) == null ? void 0 : _a.cancel();
|
|
267
|
+
break;
|
|
268
|
+
case VADEventType.INFERENCE_DONE:
|
|
269
|
+
this.hooks.onVADInferenceDone(ev);
|
|
270
|
+
break;
|
|
271
|
+
case VADEventType.END_OF_SPEECH:
|
|
272
|
+
this.logger.debug("VAD task: END_OF_SPEECH");
|
|
273
|
+
this.hooks.onEndOfSpeech(ev);
|
|
274
|
+
this.speaking = false;
|
|
275
|
+
this.lastSpeakingTime = Date.now() - ev.silenceDuration;
|
|
276
|
+
if (this.vadBaseTurnDetection || this.turnDetectionMode === "stt" && this.userTurnCommitted) {
|
|
277
|
+
const chatCtx = this.hooks.retrieveChatCtx();
|
|
278
|
+
this.runEOUDetection(chatCtx);
|
|
279
|
+
}
|
|
280
|
+
break;
|
|
281
|
+
}
|
|
282
|
+
}
|
|
283
|
+
} catch (e) {
|
|
284
|
+
this.logger.error(e, "Error in VAD task");
|
|
285
|
+
} finally {
|
|
286
|
+
this.logger.debug("VAD task closed");
|
|
287
|
+
}
|
|
288
|
+
}
|
|
289
|
+
setInputAudioStream(audioStream) {
|
|
290
|
+
this.deferredInputStream.setSource(audioStream);
|
|
291
|
+
}
|
|
292
|
+
detachInputAudioStream() {
|
|
293
|
+
this.deferredInputStream.detachSource();
|
|
294
|
+
}
|
|
295
|
+
clearUserTurn() {
|
|
296
|
+
var _a;
|
|
297
|
+
this.audioTranscript = "";
|
|
298
|
+
this.audioInterimTranscript = "";
|
|
299
|
+
this.userTurnCommitted = false;
|
|
300
|
+
(_a = this.sttTask) == null ? void 0 : _a.cancelAndWait().finally(() => {
|
|
301
|
+
this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));
|
|
302
|
+
this.sttTask.result.catch((err) => {
|
|
303
|
+
this.logger.error(`Error running STT task: ${err}`);
|
|
304
|
+
});
|
|
305
|
+
});
|
|
306
|
+
}
|
|
307
|
+
commitUserTurn(audioDetached) {
|
|
308
|
+
var _a;
|
|
309
|
+
const commitUserTurnTask = (delayDuration = 500) => async (controller) => {
|
|
310
|
+
if (Date.now() - this.lastFinalTranscriptTime > delayDuration) {
|
|
311
|
+
if (audioDetached && this.sampleRate !== void 0) {
|
|
312
|
+
const numSamples = Math.floor(this.sampleRate * 0.5);
|
|
313
|
+
const silence = new Int16Array(numSamples * 2);
|
|
314
|
+
const silenceFrame = new AudioFrame(silence, this.sampleRate, 1, numSamples);
|
|
315
|
+
this.silenceAudioWriter.write(silenceFrame);
|
|
316
|
+
}
|
|
317
|
+
await delay(delayDuration, { signal: controller.signal });
|
|
318
|
+
}
|
|
319
|
+
if (this.audioInterimTranscript) {
|
|
320
|
+
this.audioTranscript = `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();
|
|
321
|
+
}
|
|
322
|
+
this.audioInterimTranscript = "";
|
|
323
|
+
const chatCtx = this.hooks.retrieveChatCtx();
|
|
324
|
+
this.logger.debug("running EOU detection on commitUserTurn");
|
|
325
|
+
this.runEOUDetection(chatCtx);
|
|
326
|
+
this.userTurnCommitted = true;
|
|
327
|
+
};
|
|
328
|
+
(_a = this.commitUserTurnTask) == null ? void 0 : _a.cancel();
|
|
329
|
+
this.commitUserTurnTask = Task.from(commitUserTurnTask());
|
|
330
|
+
this.commitUserTurnTask.result.then(() => {
|
|
331
|
+
this.logger.debug("User turn committed");
|
|
332
|
+
}).catch((err) => {
|
|
333
|
+
this.logger.error(err, "Error in user turn commit task:");
|
|
334
|
+
});
|
|
335
|
+
}
|
|
336
|
+
async close() {
|
|
337
|
+
var _a, _b, _c, _d;
|
|
338
|
+
this.detachInputAudioStream();
|
|
339
|
+
await ((_a = this.commitUserTurnTask) == null ? void 0 : _a.cancelAndWait());
|
|
340
|
+
await ((_b = this.sttTask) == null ? void 0 : _b.cancelAndWait());
|
|
341
|
+
await ((_c = this.vadTask) == null ? void 0 : _c.cancelAndWait());
|
|
342
|
+
await ((_d = this.bounceEOUTask) == null ? void 0 : _d.cancelAndWait());
|
|
343
|
+
}
|
|
344
|
+
get vadBaseTurnDetection() {
|
|
345
|
+
return ["vad", void 0].includes(this.turnDetectionMode);
|
|
346
|
+
}
|
|
347
|
+
}
|
|
348
|
+
export {
|
|
349
|
+
AudioRecognition
|
|
350
|
+
};
|
|
351
|
+
//# sourceMappingURL=audio_recognition.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../../src/voice/audio_recognition.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioFrame } from '@livekit/rtc-node';\nimport { delay } from '@std/async';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ReadableStream } from 'node:stream/web';\nimport { type ChatContext } from '../llm/chat_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream, isStreamReaderReleaseError } from '../stream/deferred_stream.js';\nimport { IdentityTransform } from '../stream/identity_transform.js';\nimport { mergeReadableStreams } from '../stream/merge_readable_streams.js';\nimport { type SpeechEvent, SpeechEventType } from '../stt/stt.js';\nimport { Task } from '../utils.js';\nimport { type VAD, type VADEvent, VADEventType } from '../vad.js';\nimport type { TurnDetectionMode } from './agent_session.js';\nimport type { STTNode } from './io.js';\n\nexport interface EndOfTurnInfo {\n newTranscript: string;\n transcriptionDelay: number;\n endOfUtteranceDelay: number;\n}\n\nexport interface RecognitionHooks {\n onStartOfSpeech: (ev: VADEvent) => void;\n onVADInferenceDone: (ev: VADEvent) => void;\n onEndOfSpeech: (ev: VADEvent) => void;\n onInterimTranscript: (ev: SpeechEvent) => void;\n onFinalTranscript: (ev: SpeechEvent) => void;\n onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;\n\n retrieveChatCtx: () => ChatContext;\n}\n\nexport interface _TurnDetector {\n unlikelyThreshold: (language?: string) => Promise<number | undefined>;\n supportsLanguage: (language?: string) => Promise<boolean>;\n predictEndOfTurn(chatCtx: ChatContext): Promise<number>;\n}\n\nexport interface AudioRecognitionOptions {\n recognitionHooks: RecognitionHooks;\n stt?: STTNode;\n vad?: VAD;\n turnDetector?: _TurnDetector;\n turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n}\n\nexport class AudioRecognition {\n private hooks: RecognitionHooks;\n private stt?: STTNode;\n private vad?: VAD;\n private turnDetector?: _TurnDetector;\n private turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n private minEndpointingDelay: number;\n private maxEndpointingDelay: number;\n private lastLanguage?: string;\n\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private lastFinalTranscriptTime = 0;\n private audioTranscript = '';\n private audioInterimTranscript = '';\n private lastSpeakingTime = 0;\n private userTurnCommitted = false;\n private speaking = false;\n private sampleRate?: number;\n\n private vadInputStream: ReadableStream<AudioFrame>;\n private sttInputStream: ReadableStream<AudioFrame>;\n private silenceAudioTransform = new IdentityTransform<AudioFrame>();\n private silenceAudioWriter: WritableStreamDefaultWriter<AudioFrame>;\n\n // all cancellable tasks\n private bounceEOUTask?: Task<void>;\n private commitUserTurnTask?: Task<void>;\n private vadTask?: Task<void>;\n private sttTask?: Task<void>;\n\n constructor(opts: AudioRecognitionOptions) {\n this.hooks = opts.recognitionHooks;\n this.stt = opts.stt;\n this.vad = opts.vad;\n this.turnDetector = opts.turnDetector;\n this.turnDetectionMode = opts.turnDetectionMode;\n this.minEndpointingDelay = opts.minEndpointingDelay;\n this.maxEndpointingDelay = opts.maxEndpointingDelay;\n this.lastLanguage = undefined;\n\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();\n this.vadInputStream = vadInputStream;\n this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);\n this.silenceAudioWriter = this.silenceAudioTransform.writable.getWriter();\n }\n\n /**\n * Current transcript of the user's speech, including interim transcript if available.\n */\n get currentTranscript(): string {\n if (this.audioInterimTranscript) {\n return `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n }\n return this.audioTranscript;\n }\n\n async start() {\n this.vadTask = Task.from(({ signal }) => this.createVadTask(this.vad, signal));\n this.vadTask.result.catch((err) => {\n this.logger.error(`Error running VAD task: ${err}`);\n });\n\n this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n this.sttTask.result.catch((err) => {\n this.logger.error(`Error running STT task: ${err}`);\n });\n }\n\n private async onSTTEvent(ev: SpeechEvent) {\n if (\n this.turnDetectionMode === 'manual' &&\n this.userTurnCommitted &&\n (this.bounceEOUTask === undefined ||\n this.bounceEOUTask.done ||\n ev.type == SpeechEventType.INTERIM_TRANSCRIPT)\n ) {\n // ignore stt event if user turn already committed and EOU task is done\n // or it's an interim transcript\n this.logger.debug(\n {\n userTurnCommitted: this.userTurnCommitted,\n eouTaskDone: this.bounceEOUTask?.done,\n evType: ev.type,\n turnDetectionMode: this.turnDetectionMode,\n },\n 'ignoring stt event',\n );\n return;\n }\n\n switch (ev.type) {\n case SpeechEventType.FINAL_TRANSCRIPT:\n this.hooks.onFinalTranscript(ev);\n const transcript = ev.alternatives?.[0]?.text;\n this.lastLanguage = ev.alternatives?.[0]?.language;\n\n if (!transcript) {\n // stt final transcript received but no transcript\n return;\n }\n\n this.logger.debug(\n {\n user_transcript: transcript,\n language: this.lastLanguage,\n },\n 'received user transcript',\n );\n\n this.lastFinalTranscriptTime = Date.now();\n this.audioTranscript += ` ${transcript}`;\n this.audioTranscript = this.audioTranscript.trimStart();\n this.audioInterimTranscript = '';\n\n if (!this.speaking) {\n if (!this.vad) {\n // Copied from python agents:\n // vad disabled, use stt timestamp\n // TODO: this would screw up transcription latency metrics\n // but we'll live with it for now.\n // the correct way is to ensure STT fires SpeechEventType.END_OF_SPEECH\n // and using that timestamp for _last_speaking_time\n this.lastSpeakingTime = Date.now();\n }\n\n if (this.vadBaseTurnDetection || this.userTurnCommitted) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on stt FINAL_TRANSCRIPT');\n this.runEOUDetection(chatCtx);\n }\n }\n break;\n case SpeechEventType.INTERIM_TRANSCRIPT:\n this.logger.debug({ transcript: ev.alternatives?.[0]?.text }, 'interim transcript');\n this.hooks.onInterimTranscript(ev);\n this.audioInterimTranscript = ev.alternatives?.[0]?.text ?? '';\n break;\n case SpeechEventType.END_OF_SPEECH:\n if (this.turnDetectionMode !== 'stt') break;\n this.userTurnCommitted = true;\n\n if (!this.speaking) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on stt END_OF_SPEECH');\n this.runEOUDetection(chatCtx);\n }\n }\n }\n\n private runEOUDetection(chatCtx: ChatContext) {\n this.logger.debug(\n {\n stt: this.stt,\n audioTranscript: this.audioTranscript,\n turnDetectionMode: this.turnDetectionMode,\n },\n 'running EOU detection',\n );\n\n if (this.stt && !this.audioTranscript && this.turnDetectionMode !== 'manual') {\n // stt enabled but no transcript yet\n this.logger.debug('skipping EOU detection');\n return;\n }\n\n chatCtx = chatCtx.copy();\n chatCtx.addMessage({ role: 'user', content: this.audioTranscript });\n\n const turnDetector =\n // disable EOU model if manual turn detection enabled\n this.audioTranscript && this.turnDetectionMode !== 'manual' ? this.turnDetector : undefined;\n\n const bounceEOUTask = (lastSpeakingTime: number) => async (controller: AbortController) => {\n let endpointingDelay = this.minEndpointingDelay;\n\n // TODO(AJS-74): need to support actual turn detection model plugins for following code to run\n if (turnDetector) {\n this.logger.debug('Running turn detector model');\n if (!turnDetector.supportsLanguage(this.lastLanguage)) {\n this.logger.debug(`Turn detector does not support language ${this.lastLanguage}`);\n } else {\n const endOfTurnProbability = await turnDetector.predictEndOfTurn(chatCtx);\n this.logger.debug(\n { endOfTurnProbability, language: this.lastLanguage },\n 'end of turn probability',\n );\n\n const unlikelyThreshold = await turnDetector.unlikelyThreshold(this.lastLanguage);\n this.logger.debug(\n {\n unlikelyThreshold,\n endOfTurnProbability,\n language: this.lastLanguage,\n transcript: this.audioTranscript,\n },\n 'EOU Detection',\n );\n\n if (unlikelyThreshold && endOfTurnProbability < unlikelyThreshold) {\n endpointingDelay = this.maxEndpointingDelay;\n }\n }\n }\n\n const extraSleep = lastSpeakingTime + endpointingDelay - Date.now();\n // add delay to see if there's a potential upcoming EOU task that cancels this one\n await delay(Math.max(extraSleep, 0), { signal: controller.signal });\n\n this.logger.debug({ transcript: this.audioTranscript }, 'end of user turn');\n\n const committed = await this.hooks.onEndOfTurn({\n newTranscript: this.audioTranscript,\n transcriptionDelay: Math.max(this.lastFinalTranscriptTime - lastSpeakingTime, 0),\n endOfUtteranceDelay: Date.now() - lastSpeakingTime,\n });\n\n if (committed) {\n // clear the transcript if the user turn was committed\n this.audioTranscript = '';\n }\n\n this.userTurnCommitted = false;\n };\n\n // cancel any existing EOU task\n this.bounceEOUTask?.cancel();\n this.bounceEOUTask = Task.from(bounceEOUTask(this.lastSpeakingTime));\n\n this.bounceEOUTask.result\n .then(() => {\n this.logger.debug('EOU detection task completed');\n })\n .catch((err: unknown) => {\n if (err instanceof Error && err.message.includes('This operation was aborted')) {\n // ignore aborted errors\n return;\n }\n this.logger.error(err, 'Error in EOU detection task:');\n });\n }\n\n private async createSttTask(stt: STTNode | undefined, signal: AbortSignal) {\n if (!stt) return;\n\n this.logger.debug('createSttTask: create stt stream from stt node');\n\n const sttStream = await stt(this.sttInputStream, {});\n\n if (signal.aborted || sttStream === null) return;\n\n if (sttStream instanceof ReadableStream) {\n const reader = sttStream.getReader();\n\n signal.addEventListener('abort', async () => {\n try {\n reader.releaseLock();\n await sttStream?.cancel();\n } catch (e) {\n this.logger.debug('createSttTask: error during abort handler:', e);\n }\n });\n\n try {\n while (true) {\n if (signal.aborted) break;\n\n const { done, value: ev } = await reader.read();\n if (done) break;\n\n if (typeof ev === 'string') {\n throw new Error('STT node must yield SpeechEvent');\n } else {\n await this.onSTTEvent(ev);\n }\n }\n } catch (e) {\n if (isStreamReaderReleaseError(e)) {\n return;\n }\n this.logger.error({ error: e }, 'createSttTask: error reading sttStream');\n } finally {\n reader.releaseLock();\n try {\n await sttStream.cancel();\n } catch (e) {\n this.logger.debug(\n 'createSttTask: error cancelling sttStream (may already be cancelled):',\n e,\n );\n }\n }\n }\n }\n\n private async createVadTask(vad: VAD | undefined, signal: AbortSignal) {\n if (!vad) return;\n\n const vadStream = vad.stream();\n vadStream.updateInputStream(this.vadInputStream);\n\n const abortHandler = () => {\n vadStream.detachInputStream();\n vadStream.close();\n signal.removeEventListener('abort', abortHandler);\n };\n signal.addEventListener('abort', abortHandler);\n\n try {\n for await (const ev of vadStream) {\n if (signal.aborted) break;\n\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.logger.debug('VAD task: START_OF_SPEECH');\n this.hooks.onStartOfSpeech(ev);\n this.speaking = true;\n\n this.bounceEOUTask?.cancel();\n break;\n case VADEventType.INFERENCE_DONE:\n this.hooks.onVADInferenceDone(ev);\n break;\n case VADEventType.END_OF_SPEECH:\n this.logger.debug('VAD task: END_OF_SPEECH');\n this.hooks.onEndOfSpeech(ev);\n this.speaking = false;\n // when VAD fires END_OF_SPEECH, it already waited for the silence_duration\n this.lastSpeakingTime = Date.now() - ev.silenceDuration;\n\n if (\n this.vadBaseTurnDetection ||\n (this.turnDetectionMode === 'stt' && this.userTurnCommitted)\n ) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.runEOUDetection(chatCtx);\n }\n break;\n }\n }\n } catch (e) {\n this.logger.error(e, 'Error in VAD task');\n } finally {\n this.logger.debug('VAD task closed');\n }\n }\n\n setInputAudioStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputAudioStream() {\n this.deferredInputStream.detachSource();\n }\n\n clearUserTurn() {\n this.audioTranscript = '';\n this.audioInterimTranscript = '';\n this.userTurnCommitted = false;\n\n this.sttTask?.cancelAndWait().finally(() => {\n this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n this.sttTask.result.catch((err) => {\n this.logger.error(`Error running STT task: ${err}`);\n });\n });\n }\n\n commitUserTurn(audioDetached: boolean) {\n const commitUserTurnTask =\n (delayDuration: number = 500) =>\n async (controller: AbortController) => {\n if (Date.now() - this.lastFinalTranscriptTime > delayDuration) {\n // flush the stt by pushing silence\n if (audioDetached && this.sampleRate !== undefined) {\n const numSamples = Math.floor(this.sampleRate * 0.5);\n const silence = new Int16Array(numSamples * 2);\n const silenceFrame = new AudioFrame(silence, this.sampleRate, 1, numSamples);\n this.silenceAudioWriter.write(silenceFrame);\n }\n\n // wait for the final transcript to be available\n await delay(delayDuration, { signal: controller.signal });\n }\n\n if (this.audioInterimTranscript) {\n // append interim transcript in case the final transcript is not ready\n this.audioTranscript = `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n }\n this.audioInterimTranscript = '';\n\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on commitUserTurn');\n this.runEOUDetection(chatCtx);\n this.userTurnCommitted = true;\n };\n\n // cancel any existing commit user turn task\n this.commitUserTurnTask?.cancel();\n this.commitUserTurnTask = Task.from(commitUserTurnTask());\n\n this.commitUserTurnTask.result\n .then(() => {\n this.logger.debug('User turn committed');\n })\n .catch((err: unknown) => {\n this.logger.error(err, 'Error in user turn commit task:');\n });\n }\n\n async close() {\n this.detachInputAudioStream();\n await this.commitUserTurnTask?.cancelAndWait();\n await this.sttTask?.cancelAndWait();\n await this.vadTask?.cancelAndWait();\n await this.bounceEOUTask?.cancelAndWait();\n }\n\n private get vadBaseTurnDetection() {\n return ['vad', undefined].includes(this.turnDetectionMode);\n }\n}\n"],"mappings":"AAGA,SAAS,kBAAkB;AAC3B,SAAS,aAAa;AAEtB,SAAS,sBAAsB;AAC/B,eAAiC;AACjC,SAAS,WAAW;AACpB,SAAS,wBAAwB,kCAAkC;AACnE,SAAS,yBAAyB;AAClC,SAAS,4BAA4B;AACrC,SAA2B,uBAAuB;AAClD,SAAS,YAAY;AACrB,SAAkC,oBAAoB;AAqC/C,MAAM,iBAAiB;AAAA,EACpB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA,SAAS,IAAI;AAAA,EACb,0BAA0B;AAAA,EAC1B,kBAAkB;AAAA,EAClB,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EACpB,WAAW;AAAA,EACX;AAAA,EAEA;AAAA,EACA;AAAA,EACA,wBAAwB,IAAI,kBAA8B;AAAA,EAC1D;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,MAA+B;AACzC,SAAK,QAAQ,KAAK;AAClB,SAAK,MAAM,KAAK;AAChB,SAAK,MAAM,KAAK;AAChB,SAAK,eAAe,KAAK;AACzB,SAAK,oBAAoB,KAAK;AAC9B,SAAK,sBAAsB,KAAK;AAChC,SAAK,sBAAsB,KAAK;AAChC,SAAK,eAAe;AAEpB,SAAK,sBAAsB,IAAI,uBAAmC;AAClE,UAAM,CAAC,gBAAgB,cAAc,IAAI,KAAK,oBAAoB,OAAO,IAAI;AAC7E,SAAK,iBAAiB;AACtB,SAAK,iBAAiB,qBAAqB,gBAAgB,KAAK,sBAAsB,QAAQ;AAC9F,SAAK,qBAAqB,KAAK,sBAAsB,SAAS,UAAU;AAAA,EAC1E;AAAA;AAAA;AAAA;AAAA,EAKA,IAAI,oBAA4B;AAC9B,QAAI,KAAK,wBAAwB;AAC/B,aAAO,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,IACvE;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,UAAU,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAED,SAAK,UAAU,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,WAAW,IAAiB;AAzH5C;AA0HI,QACE,KAAK,sBAAsB,YAC3B,KAAK,sBACJ,KAAK,kBAAkB,UACtB,KAAK,cAAc,QACnB,GAAG,QAAQ,gBAAgB,qBAC7B;AAGA,WAAK,OAAO;AAAA,QACV;AAAA,UACE,mBAAmB,KAAK;AAAA,UACxB,cAAa,UAAK,kBAAL,mBAAoB;AAAA,UACjC,QAAQ,GAAG;AAAA,UACX,mBAAmB,KAAK;AAAA,QAC1B;AAAA,QACA;AAAA,MACF;AACA;AAAA,IACF;AAEA,YAAQ,GAAG,MAAM;AAAA,MACf,KAAK,gBAAgB;AACnB,aAAK,MAAM,kBAAkB,EAAE;AAC/B,cAAM,cAAa,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AACzC,aAAK,gBAAe,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AAE1C,YAAI,CAAC,YAAY;AAEf;AAAA,QACF;AAEA,aAAK,OAAO;AAAA,UACV;AAAA,YACE,iBAAiB;AAAA,YACjB,UAAU,KAAK;AAAA,UACjB;AAAA,UACA;AAAA,QACF;AAEA,aAAK,0BAA0B,KAAK,IAAI;AACxC,aAAK,mBAAmB,IAAI,UAAU;AACtC,aAAK,kBAAkB,KAAK,gBAAgB,UAAU;AACtD,aAAK,yBAAyB;AAE9B,YAAI,CAAC,KAAK,UAAU;AAClB,cAAI,CAAC,KAAK,KAAK;AAOb,iBAAK,mBAAmB,KAAK,IAAI;AAAA,UACnC;AAEA,cAAI,KAAK,wBAAwB,KAAK,mBAAmB;AACvD,kBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,iBAAK,OAAO,MAAM,+CAA+C;AACjE,iBAAK,gBAAgB,OAAO;AAAA,UAC9B;AAAA,QACF;AACA;AAAA,MACF,KAAK,gBAAgB;AACnB,aAAK,OAAO,MAAM,EAAE,aAAY,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,KAAK,GAAG,oBAAoB;AAClF,aAAK,MAAM,oBAAoB,EAAE;AACjC,aAAK,2BAAyB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,SAAQ;AAC5D;AAAA,MACF,KAAK,gBAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC,aAAK,oBAAoB;AAEzB,YAAI,CAAC,KAAK,UAAU;AAClB,gBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,eAAK,OAAO,MAAM,4CAA4C;AAC9D,eAAK,gBAAgB,OAAO;AAAA,QAC9B;AAAA,IACJ;AAAA,EACF;AAAA,EAEQ,gBAAgB,SAAsB;AA1MhD;AA2MI,SAAK,OAAO;AAAA,MACV;AAAA,QACE,KAAK,KAAK;AAAA,QACV,iBAAiB,KAAK;AAAA,QACtB,mBAAmB,KAAK;AAAA,MAC1B;AAAA,MACA;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,CAAC,KAAK,mBAAmB,KAAK,sBAAsB,UAAU;AAE5E,WAAK,OAAO,MAAM,wBAAwB;AAC1C;AAAA,IACF;AAEA,cAAU,QAAQ,KAAK;AACvB,YAAQ,WAAW,EAAE,MAAM,QAAQ,SAAS,KAAK,gBAAgB,CAAC;AAElE,UAAM;AAAA;AAAA,MAEJ,KAAK,mBAAmB,KAAK,sBAAsB,WAAW,KAAK,eAAe;AAAA;AAEpF,UAAM,gBAAgB,CAAC,qBAA6B,OAAO,eAAgC;AACzF,UAAI,mBAAmB,KAAK;AAG5B,UAAI,cAAc;AAChB,aAAK,OAAO,MAAM,6BAA6B;AAC/C,YAAI,CAAC,aAAa,iBAAiB,KAAK,YAAY,GAAG;AACrD,eAAK,OAAO,MAAM,2CAA2C,KAAK,YAAY,EAAE;AAAA,QAClF,OAAO;AACL,gBAAM,uBAAuB,MAAM,aAAa,iBAAiB,OAAO;AACxE,eAAK,OAAO;AAAA,YACV,EAAE,sBAAsB,UAAU,KAAK,aAAa;AAAA,YACpD;AAAA,UACF;AAEA,gBAAM,oBAAoB,MAAM,aAAa,kBAAkB,KAAK,YAAY;AAChF,eAAK,OAAO;AAAA,YACV;AAAA,cACE;AAAA,cACA;AAAA,cACA,UAAU,KAAK;AAAA,cACf,YAAY,KAAK;AAAA,YACnB;AAAA,YACA;AAAA,UACF;AAEA,cAAI,qBAAqB,uBAAuB,mBAAmB;AACjE,+BAAmB,KAAK;AAAA,UAC1B;AAAA,QACF;AAAA,MACF;AAEA,YAAM,aAAa,mBAAmB,mBAAmB,KAAK,IAAI;AAElE,YAAM,MAAM,KAAK,IAAI,YAAY,CAAC,GAAG,EAAE,QAAQ,WAAW,OAAO,CAAC;AAElE,WAAK,OAAO,MAAM,EAAE,YAAY,KAAK,gBAAgB,GAAG,kBAAkB;AAE1E,YAAM,YAAY,MAAM,KAAK,MAAM,YAAY;AAAA,QAC7C,eAAe,KAAK;AAAA,QACpB,oBAAoB,KAAK,IAAI,KAAK,0BAA0B,kBAAkB,CAAC;AAAA,QAC/E,qBAAqB,KAAK,IAAI,IAAI;AAAA,MACpC,CAAC;AAED,UAAI,WAAW;AAEb,aAAK,kBAAkB;AAAA,MACzB;AAEA,WAAK,oBAAoB;AAAA,IAC3B;AAGA,eAAK,kBAAL,mBAAoB;AACpB,SAAK,gBAAgB,KAAK,KAAK,cAAc,KAAK,gBAAgB,CAAC;AAEnE,SAAK,cAAc,OAChB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,8BAA8B;AAAA,IAClD,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,UAAI,eAAe,SAAS,IAAI,QAAQ,SAAS,4BAA4B,GAAG;AAE9E;AAAA,MACF;AACA,WAAK,OAAO,MAAM,KAAK,8BAA8B;AAAA,IACvD,CAAC;AAAA,EACL;AAAA,EAEA,MAAc,cAAc,KAA0B,QAAqB;AACzE,QAAI,CAAC,IAAK;AAEV,SAAK,OAAO,MAAM,gDAAgD;AAElE,UAAM,YAAY,MAAM,IAAI,KAAK,gBAAgB,CAAC,CAAC;AAEnD,QAAI,OAAO,WAAW,cAAc,KAAM;AAE1C,QAAI,qBAAqB,gBAAgB;AACvC,YAAM,SAAS,UAAU,UAAU;AAEnC,aAAO,iBAAiB,SAAS,YAAY;AAC3C,YAAI;AACF,iBAAO,YAAY;AACnB,iBAAM,uCAAW;AAAA,QACnB,SAAS,GAAG;AACV,eAAK,OAAO,MAAM,8CAA8C,CAAC;AAAA,QACnE;AAAA,MACF,CAAC;AAED,UAAI;AACF,eAAO,MAAM;AACX,cAAI,OAAO,QAAS;AAEpB,gBAAM,EAAE,MAAM,OAAO,GAAG,IAAI,MAAM,OAAO,KAAK;AAC9C,cAAI,KAAM;AAEV,cAAI,OAAO,OAAO,UAAU;AAC1B,kBAAM,IAAI,MAAM,iCAAiC;AAAA,UACnD,OAAO;AACL,kBAAM,KAAK,WAAW,EAAE;AAAA,UAC1B;AAAA,QACF;AAAA,MACF,SAAS,GAAG;AACV,YAAI,2BAA2B,CAAC,GAAG;AACjC;AAAA,QACF;AACA,aAAK,OAAO,MAAM,EAAE,OAAO,EAAE,GAAG,wCAAwC;AAAA,MAC1E,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,UAAU,OAAO;AAAA,QACzB,SAAS,GAAG;AACV,eAAK,OAAO;AAAA,YACV;AAAA,YACA;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,cAAc,KAAsB,QAAqB;AA3VzE;AA4VI,QAAI,CAAC,IAAK;AAEV,UAAM,YAAY,IAAI,OAAO;AAC7B,cAAU,kBAAkB,KAAK,cAAc;AAE/C,UAAM,eAAe,MAAM;AACzB,gBAAU,kBAAkB;AAC5B,gBAAU,MAAM;AAChB,aAAO,oBAAoB,SAAS,YAAY;AAAA,IAClD;AACA,WAAO,iBAAiB,SAAS,YAAY;AAE7C,QAAI;AACF,uBAAiB,MAAM,WAAW;AAChC,YAAI,OAAO,QAAS;AAEpB,gBAAQ,GAAG,MAAM;AAAA,UACf,KAAK,aAAa;AAChB,iBAAK,OAAO,MAAM,2BAA2B;AAC7C,iBAAK,MAAM,gBAAgB,EAAE;AAC7B,iBAAK,WAAW;AAEhB,uBAAK,kBAAL,mBAAoB;AACpB;AAAA,UACF,KAAK,aAAa;AAChB,iBAAK,MAAM,mBAAmB,EAAE;AAChC;AAAA,UACF,KAAK,aAAa;AAChB,iBAAK,OAAO,MAAM,yBAAyB;AAC3C,iBAAK,MAAM,cAAc,EAAE;AAC3B,iBAAK,WAAW;AAEhB,iBAAK,mBAAmB,KAAK,IAAI,IAAI,GAAG;AAExC,gBACE,KAAK,wBACJ,KAAK,sBAAsB,SAAS,KAAK,mBAC1C;AACA,oBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,mBAAK,gBAAgB,OAAO;AAAA,YAC9B;AACA;AAAA,QACJ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,GAAG,mBAAmB;AAAA,IAC1C,UAAE;AACA,WAAK,OAAO,MAAM,iBAAiB;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,oBAAoB,aAAyC;AAC3D,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,yBAAyB;AACvB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA,EAEA,gBAAgB;AAvZlB;AAwZI,SAAK,kBAAkB;AACvB,SAAK,yBAAyB;AAC9B,SAAK,oBAAoB;AAEzB,eAAK,YAAL,mBAAc,gBAAgB,QAAQ,MAAM;AAC1C,WAAK,UAAU,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,WAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,aAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,MACpD,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,eAAe,eAAwB;AApazC;AAqaI,UAAM,qBACJ,CAAC,gBAAwB,QACzB,OAAO,eAAgC;AACrC,UAAI,KAAK,IAAI,IAAI,KAAK,0BAA0B,eAAe;AAE7D,YAAI,iBAAiB,KAAK,eAAe,QAAW;AAClD,gBAAM,aAAa,KAAK,MAAM,KAAK,aAAa,GAAG;AACnD,gBAAM,UAAU,IAAI,WAAW,aAAa,CAAC;AAC7C,gBAAM,eAAe,IAAI,WAAW,SAAS,KAAK,YAAY,GAAG,UAAU;AAC3E,eAAK,mBAAmB,MAAM,YAAY;AAAA,QAC5C;AAGA,cAAM,MAAM,eAAe,EAAE,QAAQ,WAAW,OAAO,CAAC;AAAA,MAC1D;AAEA,UAAI,KAAK,wBAAwB;AAE/B,aAAK,kBAAkB,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,MACvF;AACA,WAAK,yBAAyB;AAE9B,YAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,WAAK,OAAO,MAAM,yCAAyC;AAC3D,WAAK,gBAAgB,OAAO;AAC5B,WAAK,oBAAoB;AAAA,IAC3B;AAGF,eAAK,uBAAL,mBAAyB;AACzB,SAAK,qBAAqB,KAAK,KAAK,mBAAmB,CAAC;AAExD,SAAK,mBAAmB,OACrB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,qBAAqB;AAAA,IACzC,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,WAAK,OAAO,MAAM,KAAK,iCAAiC;AAAA,IAC1D,CAAC;AAAA,EACL;AAAA,EAEA,MAAM,QAAQ;AA9chB;AA+cI,SAAK,uBAAuB;AAC5B,YAAM,UAAK,uBAAL,mBAAyB;AAC/B,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,kBAAL,mBAAoB;AAAA,EAC5B;AAAA,EAEA,IAAY,uBAAuB;AACjC,WAAO,CAAC,OAAO,MAAS,EAAE,SAAS,KAAK,iBAAiB;AAAA,EAC3D;AACF;","names":[]}
|
|
@@ -0,0 +1,145 @@
|
|
|
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 events_exports = {};
|
|
20
|
+
__export(events_exports, {
|
|
21
|
+
AgentSessionEventTypes: () => AgentSessionEventTypes,
|
|
22
|
+
CloseReason: () => CloseReason,
|
|
23
|
+
createAgentStateChangedEvent: () => createAgentStateChangedEvent,
|
|
24
|
+
createCloseEvent: () => createCloseEvent,
|
|
25
|
+
createConversationItemAddedEvent: () => createConversationItemAddedEvent,
|
|
26
|
+
createErrorEvent: () => createErrorEvent,
|
|
27
|
+
createFunctionToolsExecutedEvent: () => createFunctionToolsExecutedEvent,
|
|
28
|
+
createMetricsCollectedEvent: () => createMetricsCollectedEvent,
|
|
29
|
+
createSpeechCreatedEvent: () => createSpeechCreatedEvent,
|
|
30
|
+
createUserInputTranscribedEvent: () => createUserInputTranscribedEvent,
|
|
31
|
+
createUserStateChangedEvent: () => createUserStateChangedEvent,
|
|
32
|
+
zipFunctionCallsAndOutputs: () => zipFunctionCallsAndOutputs
|
|
33
|
+
});
|
|
34
|
+
module.exports = __toCommonJS(events_exports);
|
|
35
|
+
var AgentSessionEventTypes = /* @__PURE__ */ ((AgentSessionEventTypes2) => {
|
|
36
|
+
AgentSessionEventTypes2["UserInputTranscribed"] = "user_input_transcribed";
|
|
37
|
+
AgentSessionEventTypes2["AgentStateChanged"] = "agent_state_changed";
|
|
38
|
+
AgentSessionEventTypes2["UserStateChanged"] = "user_state_changed";
|
|
39
|
+
AgentSessionEventTypes2["ConversationItemAdded"] = "conversation_item_added";
|
|
40
|
+
AgentSessionEventTypes2["FunctionToolsExecuted"] = "function_tools_executed";
|
|
41
|
+
AgentSessionEventTypes2["MetricsCollected"] = "metrics_collected";
|
|
42
|
+
AgentSessionEventTypes2["SpeechCreated"] = "speech_created";
|
|
43
|
+
AgentSessionEventTypes2["Error"] = "error";
|
|
44
|
+
AgentSessionEventTypes2["Close"] = "close";
|
|
45
|
+
return AgentSessionEventTypes2;
|
|
46
|
+
})(AgentSessionEventTypes || {});
|
|
47
|
+
var CloseReason = /* @__PURE__ */ ((CloseReason2) => {
|
|
48
|
+
CloseReason2["ERROR"] = "error";
|
|
49
|
+
CloseReason2["JOB_SHUTDOWN"] = "job_shutdown";
|
|
50
|
+
CloseReason2["PARTICIPANT_DISCONNECTED"] = "participant_disconnected";
|
|
51
|
+
CloseReason2["USER_INITIATED"] = "user_initiated";
|
|
52
|
+
return CloseReason2;
|
|
53
|
+
})(CloseReason || {});
|
|
54
|
+
const createUserStateChangedEvent = (oldState, newState, createdAt = Date.now()) => ({
|
|
55
|
+
type: "user_state_changed",
|
|
56
|
+
oldState,
|
|
57
|
+
newState,
|
|
58
|
+
createdAt
|
|
59
|
+
});
|
|
60
|
+
const createAgentStateChangedEvent = (oldState, newState, createdAt = Date.now()) => ({
|
|
61
|
+
type: "agent_state_changed",
|
|
62
|
+
oldState,
|
|
63
|
+
newState,
|
|
64
|
+
createdAt
|
|
65
|
+
});
|
|
66
|
+
const createUserInputTranscribedEvent = ({
|
|
67
|
+
transcript,
|
|
68
|
+
isFinal,
|
|
69
|
+
speakerId = null,
|
|
70
|
+
createdAt = Date.now()
|
|
71
|
+
}) => ({
|
|
72
|
+
type: "user_input_transcribed",
|
|
73
|
+
transcript,
|
|
74
|
+
isFinal,
|
|
75
|
+
speakerId,
|
|
76
|
+
createdAt
|
|
77
|
+
});
|
|
78
|
+
const createMetricsCollectedEvent = ({
|
|
79
|
+
metrics,
|
|
80
|
+
createdAt = Date.now()
|
|
81
|
+
}) => ({
|
|
82
|
+
type: "metrics_collected",
|
|
83
|
+
metrics,
|
|
84
|
+
createdAt
|
|
85
|
+
});
|
|
86
|
+
const createConversationItemAddedEvent = (item, createdAt = Date.now()) => ({
|
|
87
|
+
type: "conversation_item_added",
|
|
88
|
+
item,
|
|
89
|
+
createdAt
|
|
90
|
+
});
|
|
91
|
+
const createFunctionToolsExecutedEvent = ({
|
|
92
|
+
functionCalls,
|
|
93
|
+
functionCallOutputs,
|
|
94
|
+
createdAt = Date.now()
|
|
95
|
+
}) => {
|
|
96
|
+
return {
|
|
97
|
+
type: "function_tools_executed",
|
|
98
|
+
functionCalls,
|
|
99
|
+
functionCallOutputs,
|
|
100
|
+
createdAt
|
|
101
|
+
};
|
|
102
|
+
};
|
|
103
|
+
const zipFunctionCallsAndOutputs = (event) => {
|
|
104
|
+
return event.functionCalls.map((call, index) => [call, event.functionCallOutputs[index]]);
|
|
105
|
+
};
|
|
106
|
+
const createSpeechCreatedEvent = ({
|
|
107
|
+
userInitiated,
|
|
108
|
+
source,
|
|
109
|
+
speechHandle,
|
|
110
|
+
createdAt = Date.now()
|
|
111
|
+
}) => ({
|
|
112
|
+
type: "speech_created",
|
|
113
|
+
userInitiated,
|
|
114
|
+
source,
|
|
115
|
+
speechHandle,
|
|
116
|
+
createdAt
|
|
117
|
+
});
|
|
118
|
+
const createErrorEvent = (error, source, createdAt = Date.now()) => ({
|
|
119
|
+
type: "error",
|
|
120
|
+
error,
|
|
121
|
+
source,
|
|
122
|
+
createdAt
|
|
123
|
+
});
|
|
124
|
+
const createCloseEvent = (reason, error = null, createdAt = Date.now()) => ({
|
|
125
|
+
type: "close",
|
|
126
|
+
error,
|
|
127
|
+
reason,
|
|
128
|
+
createdAt
|
|
129
|
+
});
|
|
130
|
+
// Annotate the CommonJS export names for ESM import in node:
|
|
131
|
+
0 && (module.exports = {
|
|
132
|
+
AgentSessionEventTypes,
|
|
133
|
+
CloseReason,
|
|
134
|
+
createAgentStateChangedEvent,
|
|
135
|
+
createCloseEvent,
|
|
136
|
+
createConversationItemAddedEvent,
|
|
137
|
+
createErrorEvent,
|
|
138
|
+
createFunctionToolsExecutedEvent,
|
|
139
|
+
createMetricsCollectedEvent,
|
|
140
|
+
createSpeechCreatedEvent,
|
|
141
|
+
createUserInputTranscribedEvent,
|
|
142
|
+
createUserStateChangedEvent,
|
|
143
|
+
zipFunctionCallsAndOutputs
|
|
144
|
+
});
|
|
145
|
+
//# sourceMappingURL=events.cjs.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../../src/voice/events.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n ChatMessage,\n FunctionCall,\n FunctionCallOutput,\n RealtimeModelError,\n} from '../llm/index.js';\nimport type { LLM, RealtimeModel } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport type { AgentMetrics } from '../metrics/base.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS } from '../tts/index.js';\nimport type { TTSError } from '../tts/tts.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport enum AgentSessionEventTypes {\n UserInputTranscribed = 'user_input_transcribed',\n AgentStateChanged = 'agent_state_changed',\n UserStateChanged = 'user_state_changed',\n ConversationItemAdded = 'conversation_item_added',\n FunctionToolsExecuted = 'function_tools_executed',\n MetricsCollected = 'metrics_collected',\n SpeechCreated = 'speech_created',\n Error = 'error',\n Close = 'close',\n}\n\nexport type UserState = 'speaking' | 'listening' | 'away';\nexport type AgentState = 'initializing' | 'idle' | 'listening' | 'thinking' | 'speaking';\n\nexport enum CloseReason {\n ERROR = 'error',\n JOB_SHUTDOWN = 'job_shutdown',\n PARTICIPANT_DISCONNECTED = 'participant_disconnected',\n USER_INITIATED = 'user_initiated',\n}\n\nexport type SpeechSource = 'say' | 'generate_reply' | 'tool_response';\n\nexport type UserStateChangedEvent = {\n type: 'user_state_changed';\n oldState: UserState;\n newState: UserState;\n createdAt: number;\n};\n\nexport const createUserStateChangedEvent = (\n oldState: UserState,\n newState: UserState,\n createdAt: number = Date.now(),\n): UserStateChangedEvent => ({\n type: 'user_state_changed',\n oldState,\n newState,\n createdAt,\n});\n\nexport type AgentStateChangedEvent = {\n type: 'agent_state_changed';\n oldState: AgentState;\n newState: AgentState;\n createdAt: number;\n};\n\nexport const createAgentStateChangedEvent = (\n oldState: AgentState,\n newState: AgentState,\n createdAt: number = Date.now(),\n): AgentStateChangedEvent => ({\n type: 'agent_state_changed',\n oldState,\n newState,\n createdAt,\n});\n\nexport type UserInputTranscribedEvent = {\n type: 'user_input_transcribed';\n transcript: string;\n isFinal: boolean;\n speakerId: string | null;\n createdAt: number;\n};\n\nexport const createUserInputTranscribedEvent = ({\n transcript,\n isFinal,\n speakerId = null,\n createdAt = Date.now(),\n}: {\n transcript: string;\n isFinal: boolean;\n speakerId?: string | null;\n createdAt?: number;\n}): UserInputTranscribedEvent => ({\n type: 'user_input_transcribed',\n transcript,\n isFinal,\n speakerId,\n createdAt,\n});\n\nexport type MetricsCollectedEvent = {\n type: 'metrics_collected';\n metrics: AgentMetrics;\n createdAt: number;\n};\n\nexport const createMetricsCollectedEvent = ({\n metrics,\n createdAt = Date.now(),\n}: {\n metrics: AgentMetrics;\n createdAt?: number;\n}): MetricsCollectedEvent => ({\n type: 'metrics_collected',\n metrics,\n createdAt,\n});\n\nexport type ConversationItemAddedEvent = {\n type: 'conversation_item_added';\n item: ChatMessage;\n createdAt: number;\n};\n\nexport const createConversationItemAddedEvent = (\n item: ChatMessage,\n createdAt: number = Date.now(),\n): ConversationItemAddedEvent => ({\n type: 'conversation_item_added',\n item,\n createdAt,\n});\n\nexport type FunctionToolsExecutedEvent = {\n type: 'function_tools_executed';\n functionCalls: FunctionCall[];\n functionCallOutputs: FunctionCallOutput[];\n createdAt: number;\n};\n\nexport const createFunctionToolsExecutedEvent = ({\n functionCalls,\n functionCallOutputs,\n createdAt = Date.now(),\n}: {\n functionCalls: FunctionCall[];\n functionCallOutputs: FunctionCallOutput[];\n createdAt?: number;\n}): FunctionToolsExecutedEvent => {\n return {\n type: 'function_tools_executed',\n functionCalls,\n functionCallOutputs,\n createdAt,\n };\n};\n\nexport const zipFunctionCallsAndOutputs = (\n event: FunctionToolsExecutedEvent,\n): Array<[FunctionCall, FunctionCallOutput]> => {\n return event.functionCalls.map((call, index) => [call, event.functionCallOutputs[index]!]);\n};\n\nexport type SpeechCreatedEvent = {\n type: 'speech_created';\n /**\n * True if the speech was created using public methods like `say` or `generate_reply`\n */\n userInitiated: boolean;\n /**\n * Source indicating how the speech handle was created\n */\n source: SpeechSource;\n /**\n * The speech handle that was created\n */\n // TODO(shubhra): we need to make sure this doesn't get serialized\n speechHandle: SpeechHandle;\n /**\n * The timestamp when the speech handle was created\n */\n createdAt: number;\n};\n\nexport const createSpeechCreatedEvent = ({\n userInitiated,\n source,\n speechHandle,\n createdAt = Date.now(),\n}: {\n userInitiated: boolean;\n source: SpeechSource;\n speechHandle: SpeechHandle;\n createdAt?: number;\n}): SpeechCreatedEvent => ({\n type: 'speech_created',\n userInitiated,\n source,\n speechHandle,\n createdAt,\n});\n\nexport type ErrorEvent = {\n type: 'error';\n error: RealtimeModelError | STTError | TTSError | LLMError | unknown;\n source: LLM | STT | TTS | RealtimeModel | unknown;\n createdAt: number;\n};\n\nexport const createErrorEvent = (\n error: RealtimeModelError | STTError | TTSError | LLMError | unknown,\n source: LLM | STT | TTS | RealtimeModel | unknown,\n createdAt: number = Date.now(),\n): ErrorEvent => ({\n type: 'error',\n error,\n source,\n createdAt,\n});\n\nexport type CloseEvent = {\n type: 'close';\n error: RealtimeModelError | STTError | TTSError | LLMError | null;\n reason: CloseReason;\n createdAt: number;\n};\n\nexport const createCloseEvent = (\n reason: CloseReason,\n error: RealtimeModelError | STTError | TTSError | LLMError | null = null,\n createdAt: number = Date.now(),\n): CloseEvent => ({\n type: 'close',\n error,\n reason,\n createdAt,\n});\n\nexport type AgentEvent =\n | UserInputTranscribedEvent\n | UserStateChangedEvent\n | AgentStateChangedEvent\n | MetricsCollectedEvent\n | ConversationItemAddedEvent\n | FunctionToolsExecutedEvent\n | SpeechCreatedEvent\n | ErrorEvent\n | CloseEvent;\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAkBO,IAAK,yBAAL,kBAAKA,4BAAL;AACL,EAAAA,wBAAA,0BAAuB;AACvB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,mBAAgB;AAChB,EAAAA,wBAAA,WAAQ;AACR,EAAAA,wBAAA,WAAQ;AATE,SAAAA;AAAA,GAAA;AAeL,IAAK,cAAL,kBAAKC,iBAAL;AACL,EAAAA,aAAA,WAAQ;AACR,EAAAA,aAAA,kBAAe;AACf,EAAAA,aAAA,8BAA2B;AAC3B,EAAAA,aAAA,oBAAiB;AAJP,SAAAA;AAAA,GAAA;AAgBL,MAAM,8BAA8B,CACzC,UACA,UACA,YAAoB,KAAK,IAAI,OACF;AAAA,EAC3B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,+BAA+B,CAC1C,UACA,UACA,YAAoB,KAAK,IAAI,OACD;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AAUO,MAAM,kCAAkC,CAAC;AAAA,EAC9C;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,YAAY,KAAK,IAAI;AACvB,OAKkC;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAQO,MAAM,8BAA8B,CAAC;AAAA,EAC1C;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAG8B;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AACF;AAQO,MAAM,mCAAmC,CAC9C,MACA,YAAoB,KAAK,IAAI,OACG;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AACF;AASO,MAAM,mCAAmC,CAAC;AAAA,EAC/C;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,MAIkC;AAChC,SAAO;AAAA,IACL,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,MAAM,6BAA6B,CACxC,UAC8C;AAC9C,SAAO,MAAM,cAAc,IAAI,CAAC,MAAM,UAAU,CAAC,MAAM,MAAM,oBAAoB,KAAK,CAAE,CAAC;AAC3F;AAuBO,MAAM,2BAA2B,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAK2B;AAAA,EACzB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,OACA,QACA,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,QACA,QAAoE,MACpE,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;","names":["AgentSessionEventTypes","CloseReason"]}
|