@livekit/agents 0.7.8 → 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 -0
- package/dist/constants.cjs.map +1 -1
- package/dist/constants.d.cts +2 -0
- package/dist/constants.d.ts +2 -0
- package/dist/constants.d.ts.map +1 -1
- package/dist/constants.js +4 -0
- 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 -0
- 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 -451
- 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 -425
- 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 -849
- 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 -826
- 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 -555
- 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 -1185
- package/src/pipeline/speech_handle.ts +0 -201
|
@@ -1,451 +0,0 @@
|
|
|
1
|
-
"use strict";
|
|
2
|
-
var __create = Object.create;
|
|
3
|
-
var __defProp = Object.defineProperty;
|
|
4
|
-
var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
|
|
5
|
-
var __getOwnPropNames = Object.getOwnPropertyNames;
|
|
6
|
-
var __getProtoOf = Object.getPrototypeOf;
|
|
7
|
-
var __hasOwnProp = Object.prototype.hasOwnProperty;
|
|
8
|
-
var __export = (target, all) => {
|
|
9
|
-
for (var name in all)
|
|
10
|
-
__defProp(target, name, { get: all[name], enumerable: true });
|
|
11
|
-
};
|
|
12
|
-
var __copyProps = (to, from, except, desc) => {
|
|
13
|
-
if (from && typeof from === "object" || typeof from === "function") {
|
|
14
|
-
for (let key of __getOwnPropNames(from))
|
|
15
|
-
if (!__hasOwnProp.call(to, key) && key !== except)
|
|
16
|
-
__defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
|
|
17
|
-
}
|
|
18
|
-
return to;
|
|
19
|
-
};
|
|
20
|
-
var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
|
|
21
|
-
// If the importer is in node compatibility mode or this is not an ESM
|
|
22
|
-
// file that has been converted to a CommonJS file using a Babel-
|
|
23
|
-
// compatible transform (i.e. "__esModule" has not been set), then set
|
|
24
|
-
// "default" to the CommonJS "module.exports" for node compatibility.
|
|
25
|
-
isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
|
|
26
|
-
mod
|
|
27
|
-
));
|
|
28
|
-
var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
|
|
29
|
-
var multimodal_agent_exports = {};
|
|
30
|
-
__export(multimodal_agent_exports, {
|
|
31
|
-
AGENT_STATE_ATTRIBUTE: () => AGENT_STATE_ATTRIBUTE,
|
|
32
|
-
MultimodalAgent: () => MultimodalAgent,
|
|
33
|
-
RealtimeModel: () => RealtimeModel,
|
|
34
|
-
RealtimeSession: () => RealtimeSession
|
|
35
|
-
});
|
|
36
|
-
module.exports = __toCommonJS(multimodal_agent_exports);
|
|
37
|
-
var import_rtc_node = require("@livekit/rtc-node");
|
|
38
|
-
var import_node_events = require("node:events");
|
|
39
|
-
var import_audio = require("../audio.cjs");
|
|
40
|
-
var import_constants = require("../constants.cjs");
|
|
41
|
-
var llm = __toESM(require("../llm/index.cjs"), 1);
|
|
42
|
-
var import_log = require("../log.cjs");
|
|
43
|
-
var import_transcription = require("../transcription.cjs");
|
|
44
|
-
var import_utils = require("../utils.cjs");
|
|
45
|
-
var import_agent_playout = require("./agent_playout.cjs");
|
|
46
|
-
class RealtimeSession extends import_node_events.EventEmitter {
|
|
47
|
-
}
|
|
48
|
-
class RealtimeModel {
|
|
49
|
-
}
|
|
50
|
-
const AGENT_STATE_ATTRIBUTE = "lk.agent.state";
|
|
51
|
-
class MultimodalAgent extends import_node_events.EventEmitter {
|
|
52
|
-
model;
|
|
53
|
-
room = null;
|
|
54
|
-
linkedParticipant = null;
|
|
55
|
-
subscribedTrack = null;
|
|
56
|
-
readMicroTask = null;
|
|
57
|
-
#textResponseRetries = 0;
|
|
58
|
-
#maxTextResponseRetries;
|
|
59
|
-
constructor({
|
|
60
|
-
model,
|
|
61
|
-
chatCtx,
|
|
62
|
-
fncCtx,
|
|
63
|
-
maxTextResponseRetries = 5,
|
|
64
|
-
noiseCancellation
|
|
65
|
-
}) {
|
|
66
|
-
super();
|
|
67
|
-
this.model = model;
|
|
68
|
-
this.#chatCtx = chatCtx;
|
|
69
|
-
this.#fncCtx = fncCtx;
|
|
70
|
-
this.#maxTextResponseRetries = maxTextResponseRetries;
|
|
71
|
-
this.#noiseCancellation = noiseCancellation;
|
|
72
|
-
}
|
|
73
|
-
#participant = null;
|
|
74
|
-
#agentPublication = null;
|
|
75
|
-
#localTrackSid = null;
|
|
76
|
-
#localSource = null;
|
|
77
|
-
#agentPlayout = null;
|
|
78
|
-
#playingHandle = void 0;
|
|
79
|
-
#logger = (0, import_log.log)();
|
|
80
|
-
#session = null;
|
|
81
|
-
#fncCtx = void 0;
|
|
82
|
-
#chatCtx = void 0;
|
|
83
|
-
#noiseCancellation = void 0;
|
|
84
|
-
#_started = false;
|
|
85
|
-
#_pendingFunctionCalls = /* @__PURE__ */ new Set();
|
|
86
|
-
#_speaking = false;
|
|
87
|
-
get fncCtx() {
|
|
88
|
-
return this.#fncCtx;
|
|
89
|
-
}
|
|
90
|
-
set fncCtx(ctx) {
|
|
91
|
-
this.#fncCtx = ctx;
|
|
92
|
-
if (this.#session) {
|
|
93
|
-
this.#session.fncCtx = ctx;
|
|
94
|
-
}
|
|
95
|
-
}
|
|
96
|
-
get #pendingFunctionCalls() {
|
|
97
|
-
return this.#_pendingFunctionCalls;
|
|
98
|
-
}
|
|
99
|
-
set #pendingFunctionCalls(calls) {
|
|
100
|
-
this.#_pendingFunctionCalls = calls;
|
|
101
|
-
this.#updateState();
|
|
102
|
-
}
|
|
103
|
-
get #speaking() {
|
|
104
|
-
return this.#_speaking;
|
|
105
|
-
}
|
|
106
|
-
set #speaking(isSpeaking) {
|
|
107
|
-
this.#_speaking = isSpeaking;
|
|
108
|
-
this.#updateState();
|
|
109
|
-
}
|
|
110
|
-
get #started() {
|
|
111
|
-
return this.#_started;
|
|
112
|
-
}
|
|
113
|
-
set #started(started) {
|
|
114
|
-
this.#_started = started;
|
|
115
|
-
this.#updateState();
|
|
116
|
-
}
|
|
117
|
-
start(room, participant = null) {
|
|
118
|
-
return new Promise(async (resolve, reject) => {
|
|
119
|
-
var _a;
|
|
120
|
-
if (this.#started) {
|
|
121
|
-
reject(new Error("MultimodalAgent already started"));
|
|
122
|
-
}
|
|
123
|
-
this.#updateState();
|
|
124
|
-
room.on(import_rtc_node.RoomEvent.ParticipantConnected, (participant2) => {
|
|
125
|
-
if (this.linkedParticipant) {
|
|
126
|
-
return;
|
|
127
|
-
}
|
|
128
|
-
this.#linkParticipant(participant2.identity);
|
|
129
|
-
});
|
|
130
|
-
room.on(
|
|
131
|
-
import_rtc_node.RoomEvent.TrackPublished,
|
|
132
|
-
(trackPublication, participant2) => {
|
|
133
|
-
if (this.linkedParticipant && participant2.identity === this.linkedParticipant.identity && trackPublication.source === import_rtc_node.TrackSource.SOURCE_MICROPHONE && !trackPublication.subscribed) {
|
|
134
|
-
trackPublication.setSubscribed(true);
|
|
135
|
-
}
|
|
136
|
-
}
|
|
137
|
-
);
|
|
138
|
-
room.on(import_rtc_node.RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));
|
|
139
|
-
this.room = room;
|
|
140
|
-
this.#participant = participant;
|
|
141
|
-
this.#localSource = new import_rtc_node.AudioSource(this.model.sampleRate, this.model.numChannels);
|
|
142
|
-
this.#agentPlayout = new import_agent_playout.AgentPlayout(
|
|
143
|
-
this.#localSource,
|
|
144
|
-
this.model.sampleRate,
|
|
145
|
-
this.model.numChannels,
|
|
146
|
-
this.model.inFrameSize,
|
|
147
|
-
this.model.outFrameSize
|
|
148
|
-
);
|
|
149
|
-
const onPlayoutStarted = () => {
|
|
150
|
-
this.emit("agent_started_speaking");
|
|
151
|
-
this.#speaking = true;
|
|
152
|
-
};
|
|
153
|
-
const onPlayoutStopped = (interrupted) => {
|
|
154
|
-
this.emit("agent_stopped_speaking");
|
|
155
|
-
this.#speaking = false;
|
|
156
|
-
if (this.#playingHandle) {
|
|
157
|
-
let text = this.#playingHandle.synchronizer.playedText;
|
|
158
|
-
if (interrupted) {
|
|
159
|
-
text += "\u2026";
|
|
160
|
-
}
|
|
161
|
-
const msg = llm.ChatMessage.create({
|
|
162
|
-
role: llm.ChatRole.ASSISTANT,
|
|
163
|
-
text
|
|
164
|
-
});
|
|
165
|
-
if (interrupted) {
|
|
166
|
-
this.emit("agent_speech_interrupted", msg);
|
|
167
|
-
} else {
|
|
168
|
-
this.emit("agent_speech_committed", msg);
|
|
169
|
-
}
|
|
170
|
-
this.#logger.child({ transcription: text, interrupted }).debug("committed agent speech");
|
|
171
|
-
}
|
|
172
|
-
};
|
|
173
|
-
this.#agentPlayout.on("playout_started", onPlayoutStarted);
|
|
174
|
-
this.#agentPlayout.on("playout_stopped", onPlayoutStopped);
|
|
175
|
-
const track = import_rtc_node.LocalAudioTrack.createAudioTrack("assistant_voice", this.#localSource);
|
|
176
|
-
const options = new import_rtc_node.TrackPublishOptions();
|
|
177
|
-
options.source = import_rtc_node.TrackSource.SOURCE_MICROPHONE;
|
|
178
|
-
this.#agentPublication = await ((_a = room.localParticipant) == null ? void 0 : _a.publishTrack(track, options)) || null;
|
|
179
|
-
if (!this.#agentPublication) {
|
|
180
|
-
this.#logger.error("Failed to publish track");
|
|
181
|
-
reject(new Error("Failed to publish track"));
|
|
182
|
-
return;
|
|
183
|
-
}
|
|
184
|
-
await this.#agentPublication.waitForSubscription();
|
|
185
|
-
if (participant) {
|
|
186
|
-
if (typeof participant === "string") {
|
|
187
|
-
this.#linkParticipant(participant);
|
|
188
|
-
} else {
|
|
189
|
-
this.#linkParticipant(participant.identity);
|
|
190
|
-
}
|
|
191
|
-
} else {
|
|
192
|
-
for (const participant2 of room.remoteParticipants.values()) {
|
|
193
|
-
this.#linkParticipant(participant2.identity);
|
|
194
|
-
break;
|
|
195
|
-
}
|
|
196
|
-
}
|
|
197
|
-
this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });
|
|
198
|
-
this.#started = true;
|
|
199
|
-
this.#session.on("response_content_added", (message) => {
|
|
200
|
-
var _a2;
|
|
201
|
-
if (message.contentType === "text") return;
|
|
202
|
-
const synchronizer = new import_transcription.TextAudioSynchronizer(import_transcription.defaultTextSyncOptions);
|
|
203
|
-
synchronizer.on("textUpdated", async (text) => {
|
|
204
|
-
await this.#publishTranscription(
|
|
205
|
-
this.room.localParticipant.identity,
|
|
206
|
-
this.#getLocalTrackSid(),
|
|
207
|
-
text.text,
|
|
208
|
-
text.final,
|
|
209
|
-
text.id
|
|
210
|
-
);
|
|
211
|
-
});
|
|
212
|
-
const handle = (_a2 = this.#agentPlayout) == null ? void 0 : _a2.play(
|
|
213
|
-
message.itemId,
|
|
214
|
-
message.contentIndex,
|
|
215
|
-
synchronizer,
|
|
216
|
-
message.textStream,
|
|
217
|
-
message.audioStream
|
|
218
|
-
);
|
|
219
|
-
this.#playingHandle = handle;
|
|
220
|
-
});
|
|
221
|
-
this.#session.on("response_content_done", (message) => {
|
|
222
|
-
if (message.contentType === "text") {
|
|
223
|
-
if (this.#textResponseRetries >= this.#maxTextResponseRetries) {
|
|
224
|
-
throw new Error(
|
|
225
|
-
`The OpenAI Realtime API returned a text response after ${this.#maxTextResponseRetries} retries. Please try to reduce the number of text system or assistant messages in the chat context.`
|
|
226
|
-
);
|
|
227
|
-
}
|
|
228
|
-
this.#textResponseRetries++;
|
|
229
|
-
this.#logger.child({
|
|
230
|
-
itemId: message.itemId,
|
|
231
|
-
text: message.text,
|
|
232
|
-
retries: this.#textResponseRetries
|
|
233
|
-
}).warn(
|
|
234
|
-
"The OpenAI Realtime API returned a text response instead of audio. Attempting to recover to audio mode..."
|
|
235
|
-
);
|
|
236
|
-
this.#session.recoverFromTextResponse(message.itemId);
|
|
237
|
-
} else {
|
|
238
|
-
this.#textResponseRetries = 0;
|
|
239
|
-
}
|
|
240
|
-
});
|
|
241
|
-
this.#session.on("input_speech_committed", async (ev) => {
|
|
242
|
-
var _a2, _b;
|
|
243
|
-
const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
|
|
244
|
-
const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
|
|
245
|
-
if (participantIdentity && trackSid) {
|
|
246
|
-
await this.#publishTranscription(participantIdentity, trackSid, "\u2026", false, ev.itemId);
|
|
247
|
-
} else {
|
|
248
|
-
this.#logger.error("Participant or track not set");
|
|
249
|
-
}
|
|
250
|
-
});
|
|
251
|
-
this.#session.on("input_speech_transcription_completed", async (ev) => {
|
|
252
|
-
var _a2, _b;
|
|
253
|
-
const transcription = ev.transcript;
|
|
254
|
-
const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
|
|
255
|
-
const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
|
|
256
|
-
if (participantIdentity && trackSid) {
|
|
257
|
-
await this.#publishTranscription(
|
|
258
|
-
participantIdentity,
|
|
259
|
-
trackSid,
|
|
260
|
-
transcription,
|
|
261
|
-
true,
|
|
262
|
-
ev.itemId
|
|
263
|
-
);
|
|
264
|
-
} else {
|
|
265
|
-
this.#logger.error("Participant or track not set");
|
|
266
|
-
}
|
|
267
|
-
const userMsg = llm.ChatMessage.create({
|
|
268
|
-
role: llm.ChatRole.USER,
|
|
269
|
-
text: transcription
|
|
270
|
-
});
|
|
271
|
-
this.emit("user_speech_committed", userMsg);
|
|
272
|
-
this.#logger.child({ transcription }).debug("committed user speech");
|
|
273
|
-
});
|
|
274
|
-
this.#session.on("input_speech_started", async (ev) => {
|
|
275
|
-
var _a2, _b;
|
|
276
|
-
this.emit("user_started_speaking");
|
|
277
|
-
if (this.#playingHandle && !this.#playingHandle.done) {
|
|
278
|
-
this.#playingHandle.interrupt();
|
|
279
|
-
this.#session.conversation.item.truncate(
|
|
280
|
-
this.#playingHandle.itemId,
|
|
281
|
-
this.#playingHandle.contentIndex,
|
|
282
|
-
Math.floor(this.#playingHandle.audioSamples / 24e3 * 1e3)
|
|
283
|
-
);
|
|
284
|
-
this.#playingHandle = void 0;
|
|
285
|
-
}
|
|
286
|
-
const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
|
|
287
|
-
const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
|
|
288
|
-
if (participantIdentity && trackSid) {
|
|
289
|
-
await this.#publishTranscription(participantIdentity, trackSid, "\u2026", false, ev.itemId);
|
|
290
|
-
}
|
|
291
|
-
});
|
|
292
|
-
this.#session.on("input_speech_stopped", (ev) => {
|
|
293
|
-
this.emit("user_stopped_speaking");
|
|
294
|
-
});
|
|
295
|
-
this.#session.on("function_call_started", (ev) => {
|
|
296
|
-
this.#pendingFunctionCalls.add(ev.callId);
|
|
297
|
-
this.#updateState();
|
|
298
|
-
});
|
|
299
|
-
this.#session.on("function_call_completed", (ev) => {
|
|
300
|
-
this.#pendingFunctionCalls.delete(ev.callId);
|
|
301
|
-
this.#updateState();
|
|
302
|
-
});
|
|
303
|
-
this.#session.on("function_call_failed", (ev) => {
|
|
304
|
-
this.#pendingFunctionCalls.delete(ev.callId);
|
|
305
|
-
this.#updateState();
|
|
306
|
-
});
|
|
307
|
-
this.#session.on("metrics_collected", (metrics) => {
|
|
308
|
-
this.emit("metrics_collected", metrics);
|
|
309
|
-
});
|
|
310
|
-
resolve(this.#session);
|
|
311
|
-
});
|
|
312
|
-
}
|
|
313
|
-
#linkParticipant(participantIdentity) {
|
|
314
|
-
if (!this.room) {
|
|
315
|
-
this.#logger.error("Room is not set");
|
|
316
|
-
return;
|
|
317
|
-
}
|
|
318
|
-
this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;
|
|
319
|
-
if (!this.linkedParticipant) {
|
|
320
|
-
this.#logger.error(`Participant with identity ${participantIdentity} not found`);
|
|
321
|
-
return;
|
|
322
|
-
}
|
|
323
|
-
if (this.linkedParticipant.trackPublications.size > 0) {
|
|
324
|
-
this.#subscribeToMicrophone();
|
|
325
|
-
}
|
|
326
|
-
for (const publication of this.linkedParticipant.trackPublications.values()) {
|
|
327
|
-
if (publication.source === import_rtc_node.TrackSource.SOURCE_MICROPHONE && publication.track) {
|
|
328
|
-
this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);
|
|
329
|
-
break;
|
|
330
|
-
}
|
|
331
|
-
}
|
|
332
|
-
}
|
|
333
|
-
#subscribeToMicrophone() {
|
|
334
|
-
if (!this.linkedParticipant) {
|
|
335
|
-
this.#logger.error("Participant is not set");
|
|
336
|
-
return;
|
|
337
|
-
}
|
|
338
|
-
let microphonePublication = void 0;
|
|
339
|
-
for (const publication of this.linkedParticipant.trackPublications.values()) {
|
|
340
|
-
if (publication.source === import_rtc_node.TrackSource.SOURCE_MICROPHONE) {
|
|
341
|
-
microphonePublication = publication;
|
|
342
|
-
break;
|
|
343
|
-
}
|
|
344
|
-
}
|
|
345
|
-
if (!microphonePublication) {
|
|
346
|
-
return;
|
|
347
|
-
}
|
|
348
|
-
if (!microphonePublication.subscribed) {
|
|
349
|
-
microphonePublication.setSubscribed(true);
|
|
350
|
-
}
|
|
351
|
-
}
|
|
352
|
-
#handleTrackSubscription(track, publication, participant) {
|
|
353
|
-
var _a;
|
|
354
|
-
if (publication.source !== import_rtc_node.TrackSource.SOURCE_MICROPHONE || participant.identity !== ((_a = this.linkedParticipant) == null ? void 0 : _a.identity)) {
|
|
355
|
-
return;
|
|
356
|
-
}
|
|
357
|
-
const readAudioStreamTask = async (audioStream) => {
|
|
358
|
-
const bstream = new import_audio.AudioByteStream(
|
|
359
|
-
this.model.sampleRate,
|
|
360
|
-
this.model.numChannels,
|
|
361
|
-
this.model.inFrameSize
|
|
362
|
-
);
|
|
363
|
-
for await (const frame of audioStream) {
|
|
364
|
-
const audioData = frame.data;
|
|
365
|
-
for (const frame2 of bstream.write(audioData.buffer)) {
|
|
366
|
-
this.#session.inputAudioBuffer.append(frame2);
|
|
367
|
-
}
|
|
368
|
-
}
|
|
369
|
-
};
|
|
370
|
-
this.subscribedTrack = track;
|
|
371
|
-
this.readMicroTask = new Promise((resolve, reject) => {
|
|
372
|
-
const audioStreamOptions = {
|
|
373
|
-
sampleRate: this.model.sampleRate,
|
|
374
|
-
numChannels: this.model.numChannels,
|
|
375
|
-
...this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}
|
|
376
|
-
};
|
|
377
|
-
readAudioStreamTask(new import_rtc_node.AudioStream(track, audioStreamOptions)).then(resolve).catch(reject);
|
|
378
|
-
});
|
|
379
|
-
}
|
|
380
|
-
#getLocalTrackSid() {
|
|
381
|
-
if (!this.#localTrackSid && this.room && this.room.localParticipant) {
|
|
382
|
-
this.#localTrackSid = (0, import_utils.findMicroTrackId)(this.room, this.room.localParticipant.identity);
|
|
383
|
-
}
|
|
384
|
-
return this.#localTrackSid;
|
|
385
|
-
}
|
|
386
|
-
async #publishTranscription(participantIdentity, trackSid, text, isFinal, id) {
|
|
387
|
-
var _a;
|
|
388
|
-
this.#logger.debug(
|
|
389
|
-
`Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`
|
|
390
|
-
);
|
|
391
|
-
if (!((_a = this.room) == null ? void 0 : _a.localParticipant)) {
|
|
392
|
-
this.#logger.error("Room or local participant not set");
|
|
393
|
-
return;
|
|
394
|
-
}
|
|
395
|
-
this.room.localParticipant.publishTranscription({
|
|
396
|
-
participantIdentity,
|
|
397
|
-
trackSid,
|
|
398
|
-
segments: [
|
|
399
|
-
{
|
|
400
|
-
text,
|
|
401
|
-
final: isFinal,
|
|
402
|
-
id,
|
|
403
|
-
startTime: BigInt(0),
|
|
404
|
-
endTime: BigInt(0),
|
|
405
|
-
language: ""
|
|
406
|
-
}
|
|
407
|
-
]
|
|
408
|
-
});
|
|
409
|
-
const stream = await this.room.localParticipant.streamText({
|
|
410
|
-
topic: import_constants.TOPIC_TRANSCRIPTION,
|
|
411
|
-
senderIdentity: participantIdentity,
|
|
412
|
-
attributes: {
|
|
413
|
-
[import_constants.ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,
|
|
414
|
-
[import_constants.ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString()
|
|
415
|
-
}
|
|
416
|
-
});
|
|
417
|
-
await stream.write(text);
|
|
418
|
-
await stream.close();
|
|
419
|
-
}
|
|
420
|
-
#updateState() {
|
|
421
|
-
let newState = "initializing";
|
|
422
|
-
if (this.#pendingFunctionCalls.size > 0) {
|
|
423
|
-
newState = "thinking";
|
|
424
|
-
} else if (this.#speaking) {
|
|
425
|
-
newState = "speaking";
|
|
426
|
-
} else if (this.#started) {
|
|
427
|
-
newState = "listening";
|
|
428
|
-
}
|
|
429
|
-
this.#setState(newState);
|
|
430
|
-
}
|
|
431
|
-
#setState(state) {
|
|
432
|
-
var _a;
|
|
433
|
-
if (((_a = this.room) == null ? void 0 : _a.isConnected) && this.room.localParticipant) {
|
|
434
|
-
const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];
|
|
435
|
-
if (currentState !== state) {
|
|
436
|
-
this.room.localParticipant.setAttributes({
|
|
437
|
-
[AGENT_STATE_ATTRIBUTE]: state
|
|
438
|
-
});
|
|
439
|
-
this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);
|
|
440
|
-
}
|
|
441
|
-
}
|
|
442
|
-
}
|
|
443
|
-
}
|
|
444
|
-
// Annotate the CommonJS export names for ESM import in node:
|
|
445
|
-
0 && (module.exports = {
|
|
446
|
-
AGENT_STATE_ATTRIBUTE,
|
|
447
|
-
MultimodalAgent,
|
|
448
|
-
RealtimeModel,
|
|
449
|
-
RealtimeSession
|
|
450
|
-
});
|
|
451
|
-
//# sourceMappingURL=multimodal_agent.cjs.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n NoiseCancellationOptions,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport {\n ATTRIBUTE_TRANSCRIPTION_FINAL,\n ATTRIBUTE_TRANSCRIPTION_TRACK_ID,\n TOPIC_TRANSCRIPTION,\n} from '../constants.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { TextAudioSynchronizer, defaultTextSyncOptions } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n abstract recoverFromTextResponse(itemId: string): void;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n #textResponseRetries = 0;\n #maxTextResponseRetries: number;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n maxTextResponseRetries = 5,\n noiseCancellation,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n maxTextResponseRetries?: number;\n noiseCancellation?: NoiseCancellationOptions;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n this.#maxTextResponseRetries = maxTextResponseRetries;\n this.#noiseCancellation = noiseCancellation;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n #noiseCancellation: NoiseCancellationOptions | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity!);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.synchronizer.playedText;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity!);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') return;\n\n const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);\n synchronizer.on('textUpdated', async (text) => {\n await this.#publishTranscription(\n this.room!.localParticipant!.identity!,\n this.#getLocalTrackSid()!,\n text.text,\n text.final,\n text.id,\n );\n });\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n synchronizer,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_done', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') {\n if (this.#textResponseRetries >= this.#maxTextResponseRetries) {\n throw new Error(\n 'The OpenAI Realtime API returned a text response ' +\n `after ${this.#maxTextResponseRetries} retries. ` +\n 'Please try to reduce the number of text system or ' +\n 'assistant messages in the chat context.',\n );\n }\n\n this.#textResponseRetries++;\n this.#logger\n .child({\n itemId: message.itemId,\n text: message.text,\n retries: this.#textResponseRetries,\n })\n .warn(\n 'The OpenAI Realtime API returned a text response instead of audio. ' +\n 'Attempting to recover to audio mode...',\n );\n this.#session!.recoverFromTextResponse(message.itemId);\n } else {\n this.#textResponseRetries = 0;\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', async (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n await this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', async (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n await this.#publishTranscription(\n participantIdentity,\n trackSid,\n transcription,\n true,\n ev.itemId,\n );\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', async (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n await this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n const audioStreamOptions = {\n sampleRate: this.model.sampleRate,\n numChannels: this.model.numChannels,\n ...(this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}),\n };\n readAudioStreamTask(new AudioStream(track, audioStreamOptions)).then(resolve).catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant!.identity!);\n }\n return this.#localTrackSid;\n }\n\n async #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): Promise<void> {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n\n const stream = await this.room.localParticipant.streamText({\n topic: TOPIC_TRANSCRIPTION,\n senderIdentity: participantIdentity,\n attributes: {\n [ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,\n [ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString(),\n },\n });\n await stream.write(text);\n await stream.close();\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes![AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAYA,sBAOO;AACP,yBAA6B;AAC7B,mBAAgC;AAChC,uBAIO;AACP,UAAqB;AACrB,iBAAoB;AAEpB,2BAA8D;AAC9D,mBAAiC;AACjC,2BAAiD;AAM1C,MAAe,wBAAwB,gCAAa;AAO3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,gCAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,uBAAuB;AAAA,EACvB;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA,yBAAyB;AAAA,IACzB;AAAA,EACF,GAMG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,0BAA0B;AAC/B,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,cAAU,gBAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EACxC,qBAA2D;AAAA,EAE3D,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AA1JlD;AA2JM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,0BAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAS;AAAA,MAC7C,CAAC;AACD,WAAK;AAAA,QACH,0BAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,4BAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,0BAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,4BAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,aAAa;AAC5C,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gCAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oCAAoB;AACxC,cAAQ,SAAS,4BAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAS;AAAA,QAC7C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAS;AAC3C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AA7PnE,YAAAC;AA+PQ,YAAI,QAAQ,gBAAgB,OAAQ;AAEpC,cAAM,eAAe,IAAI,2CAAsB,2CAAsB;AACrE,qBAAa,GAAG,eAAe,OAAO,SAAS;AAC7C,gBAAM,KAAK;AAAA,YACT,KAAK,KAAM,iBAAkB;AAAA,YAC7B,KAAK,kBAAkB;AAAA,YACvB,KAAK;AAAA,YACL,KAAK;AAAA,YACL,KAAK;AAAA,UACP;AAAA,QACF,CAAC;AAED,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,YAAiB;AAE1D,YAAI,QAAQ,gBAAgB,QAAQ;AAClC,cAAI,KAAK,wBAAwB,KAAK,yBAAyB;AAC7D,kBAAM,IAAI;AAAA,cACR,0DACW,KAAK,uBAAuB;AAAA,YAGzC;AAAA,UACF;AAEA,eAAK;AACL,eAAK,QACF,MAAM;AAAA,YACL,QAAQ,QAAQ;AAAA,YAChB,MAAM,QAAQ;AAAA,YACd,SAAS,KAAK;AAAA,UAChB,CAAC,EACA;AAAA,YACC;AAAA,UAEF;AACF,eAAK,SAAU,wBAAwB,QAAQ,MAAM;AAAA,QACvD,OAAO;AACL,eAAK,uBAAuB;AAAA,QAC9B;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,OAAO,OAAY;AArTpE,YAAAA,KAAA;AAuTQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,gBAAM,KAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACvF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,OAAO,OAAY;AAjUlF,YAAAA,KAAA;AAmUQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,gBAAM,KAAK;AAAA,YACT;AAAA,YACA;AAAA,YACA;AAAA,YACA;AAAA,YACA,GAAG;AAAA,UACL;AAAA,QACF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,OAAO,OAAY;AAzVlE,YAAAA,KAAA;AA0VQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,gBAAM,KAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACvF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AAhcJ;AAicI,QACE,YAAY,WAAW,4BAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,YAAM,qBAAqB;AAAA,QACzB,YAAY,KAAK,MAAM;AAAA,QACvB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,qBAAqB,EAAE,mBAAmB,KAAK,mBAAmB,IAAI,CAAC;AAAA,MAClF;AACA,0BAAoB,IAAI,4BAAY,OAAO,kBAAkB,CAAC,EAAE,KAAK,OAAO,EAAE,MAAM,MAAM;AAAA,IAC5F,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AACjC,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,qBAAiB,+BAAiB,KAAK,MAAM,KAAK,KAAK,iBAAkB,QAAS;AAAA,IACzF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,sBACJ,qBACA,UACA,MACA,SACA,IACe;AA9enB;AA+eI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAED,UAAM,SAAS,MAAM,KAAK,KAAK,iBAAiB,WAAW;AAAA,MACzD,OAAO;AAAA,MACP,gBAAgB;AAAA,MAChB,YAAY;AAAA,QACV,CAAC,iDAAgC,GAAG;AAAA,QACpC,CAAC,8CAA6B,GAAG,QAAQ,SAAS;AAAA,MACpD;AAAA,IACF,CAAC;AACD,UAAM,OAAO,MAAM,IAAI;AACvB,UAAM,OAAO,MAAM;AAAA,EACrB;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AA/hB/B;AAgiBI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAY,qBAAqB;AACjF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
@@ -1,48 +0,0 @@
|
|
|
1
|
-
/// <reference types="node" resolution-mode="require"/>
|
|
2
|
-
import type { NoiseCancellationOptions, RemoteAudioTrack, RemoteParticipant, Room } from '@livekit/rtc-node';
|
|
3
|
-
import { EventEmitter } from 'node:events';
|
|
4
|
-
import * as llm from '../llm/index.js';
|
|
5
|
-
/**
|
|
6
|
-
* @internal
|
|
7
|
-
* @beta
|
|
8
|
-
*/
|
|
9
|
-
export declare abstract class RealtimeSession extends EventEmitter {
|
|
10
|
-
abstract conversation: any;
|
|
11
|
-
abstract inputAudioBuffer: any;
|
|
12
|
-
abstract fncCtx: llm.FunctionContext | undefined;
|
|
13
|
-
abstract recoverFromTextResponse(itemId: string): void;
|
|
14
|
-
}
|
|
15
|
-
/**
|
|
16
|
-
* @internal
|
|
17
|
-
* @beta
|
|
18
|
-
*/
|
|
19
|
-
export declare abstract class RealtimeModel {
|
|
20
|
-
abstract session(options: any): RealtimeSession;
|
|
21
|
-
abstract close(): Promise<void>;
|
|
22
|
-
abstract sampleRate: number;
|
|
23
|
-
abstract numChannels: number;
|
|
24
|
-
abstract inFrameSize: number;
|
|
25
|
-
abstract outFrameSize: number;
|
|
26
|
-
}
|
|
27
|
-
export type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';
|
|
28
|
-
export declare const AGENT_STATE_ATTRIBUTE = "lk.agent.state";
|
|
29
|
-
/** @beta */
|
|
30
|
-
export declare class MultimodalAgent extends EventEmitter {
|
|
31
|
-
#private;
|
|
32
|
-
model: RealtimeModel;
|
|
33
|
-
room: Room | null;
|
|
34
|
-
linkedParticipant: RemoteParticipant | null;
|
|
35
|
-
subscribedTrack: RemoteAudioTrack | null;
|
|
36
|
-
readMicroTask: Promise<void> | null;
|
|
37
|
-
constructor({ model, chatCtx, fncCtx, maxTextResponseRetries, noiseCancellation, }: {
|
|
38
|
-
model: RealtimeModel;
|
|
39
|
-
chatCtx?: llm.ChatContext;
|
|
40
|
-
fncCtx?: llm.FunctionContext;
|
|
41
|
-
maxTextResponseRetries?: number;
|
|
42
|
-
noiseCancellation?: NoiseCancellationOptions;
|
|
43
|
-
});
|
|
44
|
-
get fncCtx(): llm.FunctionContext | undefined;
|
|
45
|
-
set fncCtx(ctx: llm.FunctionContext | undefined);
|
|
46
|
-
start(room: Room, participant?: RemoteParticipant | string | null): Promise<RealtimeSession>;
|
|
47
|
-
}
|
|
48
|
-
//# sourceMappingURL=multimodal_agent.d.ts.map
|
|
@@ -1,48 +0,0 @@
|
|
|
1
|
-
/// <reference types="node" resolution-mode="require"/>
|
|
2
|
-
import type { NoiseCancellationOptions, RemoteAudioTrack, RemoteParticipant, Room } from '@livekit/rtc-node';
|
|
3
|
-
import { EventEmitter } from 'node:events';
|
|
4
|
-
import * as llm from '../llm/index.js';
|
|
5
|
-
/**
|
|
6
|
-
* @internal
|
|
7
|
-
* @beta
|
|
8
|
-
*/
|
|
9
|
-
export declare abstract class RealtimeSession extends EventEmitter {
|
|
10
|
-
abstract conversation: any;
|
|
11
|
-
abstract inputAudioBuffer: any;
|
|
12
|
-
abstract fncCtx: llm.FunctionContext | undefined;
|
|
13
|
-
abstract recoverFromTextResponse(itemId: string): void;
|
|
14
|
-
}
|
|
15
|
-
/**
|
|
16
|
-
* @internal
|
|
17
|
-
* @beta
|
|
18
|
-
*/
|
|
19
|
-
export declare abstract class RealtimeModel {
|
|
20
|
-
abstract session(options: any): RealtimeSession;
|
|
21
|
-
abstract close(): Promise<void>;
|
|
22
|
-
abstract sampleRate: number;
|
|
23
|
-
abstract numChannels: number;
|
|
24
|
-
abstract inFrameSize: number;
|
|
25
|
-
abstract outFrameSize: number;
|
|
26
|
-
}
|
|
27
|
-
export type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';
|
|
28
|
-
export declare const AGENT_STATE_ATTRIBUTE = "lk.agent.state";
|
|
29
|
-
/** @beta */
|
|
30
|
-
export declare class MultimodalAgent extends EventEmitter {
|
|
31
|
-
#private;
|
|
32
|
-
model: RealtimeModel;
|
|
33
|
-
room: Room | null;
|
|
34
|
-
linkedParticipant: RemoteParticipant | null;
|
|
35
|
-
subscribedTrack: RemoteAudioTrack | null;
|
|
36
|
-
readMicroTask: Promise<void> | null;
|
|
37
|
-
constructor({ model, chatCtx, fncCtx, maxTextResponseRetries, noiseCancellation, }: {
|
|
38
|
-
model: RealtimeModel;
|
|
39
|
-
chatCtx?: llm.ChatContext;
|
|
40
|
-
fncCtx?: llm.FunctionContext;
|
|
41
|
-
maxTextResponseRetries?: number;
|
|
42
|
-
noiseCancellation?: NoiseCancellationOptions;
|
|
43
|
-
});
|
|
44
|
-
get fncCtx(): llm.FunctionContext | undefined;
|
|
45
|
-
set fncCtx(ctx: llm.FunctionContext | undefined);
|
|
46
|
-
start(room: Room, participant?: RemoteParticipant | string | null): Promise<RealtimeSession>;
|
|
47
|
-
}
|
|
48
|
-
//# sourceMappingURL=multimodal_agent.d.ts.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"multimodal_agent.d.ts","sourceRoot":"","sources":["../../src/multimodal/multimodal_agent.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAEV,wBAAwB,EACxB,gBAAgB,EAChB,iBAAiB,EAGjB,IAAI,EACL,MAAM,mBAAmB,CAAC;AAS3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAO3C,OAAO,KAAK,GAAG,MAAM,iBAAiB,CAAC;AAOvC;;;GAGG;AACH,8BAAsB,eAAgB,SAAQ,YAAY;IAExD,QAAQ,CAAC,YAAY,EAAE,GAAG,CAAC;IAE3B,QAAQ,CAAC,gBAAgB,EAAE,GAAG,CAAC;IAC/B,QAAQ,CAAC,MAAM,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,CAAC;IACjD,QAAQ,CAAC,uBAAuB,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;CACvD;AAED;;;GAGG;AACH,8BAAsB,aAAa;IAEjC,QAAQ,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,GAAG,eAAe;IAC/C,QAAQ,CAAC,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAC/B,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;CAC/B;AAED,MAAM,MAAM,UAAU,GAAG,cAAc,GAAG,UAAU,GAAG,WAAW,GAAG,UAAU,CAAC;AAChF,eAAO,MAAM,qBAAqB,mBAAmB,CAAC;AAEtD,YAAY;AACZ,qBAAa,eAAgB,SAAQ,YAAY;;IAC/C,KAAK,EAAE,aAAa,CAAC;IACrB,IAAI,EAAE,IAAI,GAAG,IAAI,CAAQ;IACzB,iBAAiB,EAAE,iBAAiB,GAAG,IAAI,CAAQ;IACnD,eAAe,EAAE,gBAAgB,GAAG,IAAI,CAAQ;IAChD,aAAa,EAAE,OAAO,CAAC,IAAI,CAAC,GAAG,IAAI,CAAQ;gBAK/B,EACV,KAAK,EACL,OAAO,EACP,MAAM,EACN,sBAA0B,EAC1B,iBAAiB,GAClB,EAAE;QACD,KAAK,EAAE,aAAa,CAAC;QACrB,OAAO,CAAC,EAAE,GAAG,CAAC,WAAW,CAAC;QAC1B,MAAM,CAAC,EAAE,GAAG,CAAC,eAAe,CAAC;QAC7B,sBAAsB,CAAC,EAAE,MAAM,CAAC;QAChC,iBAAiB,CAAC,EAAE,wBAAwB,CAAC;KAC9C;IAyBD,IAAI,MAAM,IAAI,GAAG,CAAC,eAAe,GAAG,SAAS,CAE5C;IAED,IAAI,MAAM,CAAC,GAAG,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,EAK9C;IA6BD,KAAK,CACH,IAAI,EAAE,IAAI,EACV,WAAW,GAAE,iBAAiB,GAAG,MAAM,GAAG,IAAW,GACpD,OAAO,CAAC,eAAe,CAAC;CAiZ5B"}
|