@livekit/agents 1.1.0 → 1.2.0
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/audio.cjs +89 -3
- package/dist/audio.cjs.map +1 -1
- package/dist/audio.d.cts +36 -1
- package/dist/audio.d.ts +36 -1
- package/dist/audio.d.ts.map +1 -1
- package/dist/audio.js +76 -2
- package/dist/audio.js.map +1 -1
- package/dist/beta/index.cjs +29 -0
- package/dist/beta/index.cjs.map +1 -0
- package/dist/beta/index.d.cts +2 -0
- package/dist/beta/index.d.ts +2 -0
- package/dist/beta/index.d.ts.map +1 -0
- package/dist/beta/index.js +7 -0
- package/dist/beta/index.js.map +1 -0
- package/dist/beta/workflows/index.cjs +29 -0
- package/dist/beta/workflows/index.cjs.map +1 -0
- package/dist/beta/workflows/index.d.cts +2 -0
- package/dist/beta/workflows/index.d.ts +2 -0
- package/dist/beta/workflows/index.d.ts.map +1 -0
- package/dist/beta/workflows/index.js +7 -0
- package/dist/beta/workflows/index.js.map +1 -0
- package/dist/beta/workflows/task_group.cjs +162 -0
- package/dist/beta/workflows/task_group.cjs.map +1 -0
- package/dist/beta/workflows/task_group.d.cts +32 -0
- package/dist/beta/workflows/task_group.d.ts +32 -0
- package/dist/beta/workflows/task_group.d.ts.map +1 -0
- package/dist/beta/workflows/task_group.js +138 -0
- package/dist/beta/workflows/task_group.js.map +1 -0
- package/dist/cli.cjs +44 -46
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.d.cts +3 -3
- package/dist/cli.d.ts +3 -3
- package/dist/cli.d.ts.map +1 -1
- package/dist/cli.js +45 -47
- package/dist/cli.js.map +1 -1
- package/dist/connection_pool.cjs +242 -0
- package/dist/connection_pool.cjs.map +1 -0
- package/dist/connection_pool.d.cts +123 -0
- package/dist/connection_pool.d.ts +123 -0
- package/dist/connection_pool.d.ts.map +1 -0
- package/dist/connection_pool.js +218 -0
- package/dist/connection_pool.js.map +1 -0
- package/dist/connection_pool.test.cjs +256 -0
- package/dist/connection_pool.test.cjs.map +1 -0
- package/dist/connection_pool.test.js +255 -0
- package/dist/connection_pool.test.js.map +1 -0
- package/dist/constants.cjs +30 -0
- package/dist/constants.cjs.map +1 -1
- package/dist/constants.d.cts +10 -0
- package/dist/constants.d.ts +10 -0
- package/dist/constants.d.ts.map +1 -1
- package/dist/constants.js +20 -0
- package/dist/constants.js.map +1 -1
- package/dist/cpu.cjs +189 -0
- package/dist/cpu.cjs.map +1 -0
- package/dist/cpu.d.cts +24 -0
- package/dist/cpu.d.ts +24 -0
- package/dist/cpu.d.ts.map +1 -0
- package/dist/cpu.js +152 -0
- package/dist/cpu.js.map +1 -0
- package/dist/cpu.test.cjs +227 -0
- package/dist/cpu.test.cjs.map +1 -0
- package/dist/cpu.test.js +204 -0
- package/dist/cpu.test.js.map +1 -0
- package/dist/http_server.cjs +9 -6
- package/dist/http_server.cjs.map +1 -1
- package/dist/http_server.d.cts +5 -1
- package/dist/http_server.d.ts +5 -1
- package/dist/http_server.d.ts.map +1 -1
- package/dist/http_server.js +9 -6
- package/dist/http_server.js.map +1 -1
- package/dist/index.cjs +24 -9
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +15 -11
- package/dist/index.d.ts +15 -11
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +18 -9
- package/dist/index.js.map +1 -1
- package/dist/inference/api_protos.cjs +70 -2
- package/dist/inference/api_protos.cjs.map +1 -1
- package/dist/inference/api_protos.d.cts +373 -32
- package/dist/inference/api_protos.d.ts +373 -32
- package/dist/inference/api_protos.d.ts.map +1 -1
- package/dist/inference/api_protos.js +62 -2
- package/dist/inference/api_protos.js.map +1 -1
- package/dist/inference/index.cjs +8 -0
- package/dist/inference/index.cjs.map +1 -1
- package/dist/inference/index.d.cts +3 -4
- package/dist/inference/index.d.ts +3 -4
- package/dist/inference/index.d.ts.map +1 -1
- package/dist/inference/index.js +18 -3
- package/dist/inference/index.js.map +1 -1
- package/dist/inference/interruption/defaults.cjs +81 -0
- package/dist/inference/interruption/defaults.cjs.map +1 -0
- package/dist/inference/interruption/defaults.d.cts +19 -0
- package/dist/inference/interruption/defaults.d.ts +19 -0
- package/dist/inference/interruption/defaults.d.ts.map +1 -0
- package/dist/inference/interruption/defaults.js +46 -0
- package/dist/inference/interruption/defaults.js.map +1 -0
- package/dist/inference/interruption/errors.cjs +44 -0
- package/dist/inference/interruption/errors.cjs.map +1 -0
- package/dist/inference/interruption/errors.d.cts +12 -0
- package/dist/inference/interruption/errors.d.ts +12 -0
- package/dist/inference/interruption/errors.d.ts.map +1 -0
- package/dist/inference/interruption/errors.js +20 -0
- package/dist/inference/interruption/errors.js.map +1 -0
- package/dist/inference/interruption/http_transport.cjs +163 -0
- package/dist/inference/interruption/http_transport.cjs.map +1 -0
- package/dist/inference/interruption/http_transport.d.cts +63 -0
- package/dist/inference/interruption/http_transport.d.ts +63 -0
- package/dist/inference/interruption/http_transport.d.ts.map +1 -0
- package/dist/inference/interruption/http_transport.js +137 -0
- package/dist/inference/interruption/http_transport.js.map +1 -0
- package/dist/inference/interruption/interruption_cache_entry.cjs +58 -0
- package/dist/inference/interruption/interruption_cache_entry.cjs.map +1 -0
- package/dist/inference/interruption/interruption_cache_entry.d.cts +30 -0
- package/dist/inference/interruption/interruption_cache_entry.d.ts +30 -0
- package/dist/inference/interruption/interruption_cache_entry.d.ts.map +1 -0
- package/dist/inference/interruption/interruption_cache_entry.js +34 -0
- package/dist/inference/interruption/interruption_cache_entry.js.map +1 -0
- package/dist/inference/interruption/interruption_detector.cjs +198 -0
- package/dist/inference/interruption/interruption_detector.cjs.map +1 -0
- package/dist/inference/interruption/interruption_detector.d.cts +59 -0
- package/dist/inference/interruption/interruption_detector.d.ts +59 -0
- package/dist/inference/interruption/interruption_detector.d.ts.map +1 -0
- package/dist/inference/interruption/interruption_detector.js +164 -0
- package/dist/inference/interruption/interruption_detector.js.map +1 -0
- package/dist/inference/interruption/interruption_stream.cjs +368 -0
- package/dist/inference/interruption/interruption_stream.cjs.map +1 -0
- package/dist/inference/interruption/interruption_stream.d.cts +46 -0
- package/dist/inference/interruption/interruption_stream.d.ts +46 -0
- package/dist/inference/interruption/interruption_stream.d.ts.map +1 -0
- package/dist/inference/interruption/interruption_stream.js +344 -0
- package/dist/inference/interruption/interruption_stream.js.map +1 -0
- package/dist/inference/interruption/types.cjs +17 -0
- package/dist/inference/interruption/types.cjs.map +1 -0
- package/dist/inference/interruption/types.d.cts +66 -0
- package/dist/inference/interruption/types.d.ts +66 -0
- package/dist/inference/interruption/types.d.ts.map +1 -0
- package/dist/inference/interruption/types.js +1 -0
- package/dist/inference/interruption/types.js.map +1 -0
- package/dist/inference/interruption/utils.cjs +130 -0
- package/dist/inference/interruption/utils.cjs.map +1 -0
- package/dist/inference/interruption/utils.d.cts +41 -0
- package/dist/inference/interruption/utils.d.ts +41 -0
- package/dist/inference/interruption/utils.d.ts.map +1 -0
- package/dist/inference/interruption/utils.js +105 -0
- package/dist/inference/interruption/utils.js.map +1 -0
- package/dist/inference/interruption/utils.test.cjs +105 -0
- package/dist/inference/interruption/utils.test.cjs.map +1 -0
- package/dist/inference/interruption/utils.test.js +104 -0
- package/dist/inference/interruption/utils.test.js.map +1 -0
- package/dist/inference/interruption/ws_transport.cjs +342 -0
- package/dist/inference/interruption/ws_transport.cjs.map +1 -0
- package/dist/inference/interruption/ws_transport.d.cts +33 -0
- package/dist/inference/interruption/ws_transport.d.ts +33 -0
- package/dist/inference/interruption/ws_transport.d.ts.map +1 -0
- package/dist/inference/interruption/ws_transport.js +308 -0
- package/dist/inference/interruption/ws_transport.js.map +1 -0
- package/dist/inference/llm.cjs +106 -66
- package/dist/inference/llm.cjs.map +1 -1
- package/dist/inference/llm.d.cts +65 -43
- package/dist/inference/llm.d.ts +65 -43
- package/dist/inference/llm.d.ts.map +1 -1
- package/dist/inference/llm.js +100 -66
- package/dist/inference/llm.js.map +1 -1
- package/dist/inference/stt.cjs +319 -170
- package/dist/inference/stt.cjs.map +1 -1
- package/dist/inference/stt.d.cts +64 -15
- package/dist/inference/stt.d.ts +64 -15
- package/dist/inference/stt.d.ts.map +1 -1
- package/dist/inference/stt.js +319 -170
- package/dist/inference/stt.js.map +1 -1
- package/dist/inference/stt.test.cjs +218 -0
- package/dist/inference/stt.test.cjs.map +1 -0
- package/dist/inference/stt.test.js +217 -0
- package/dist/inference/stt.test.js.map +1 -0
- package/dist/inference/tts.cjs +249 -71
- package/dist/inference/tts.cjs.map +1 -1
- package/dist/inference/tts.d.cts +55 -16
- package/dist/inference/tts.d.ts +55 -16
- package/dist/inference/tts.d.ts.map +1 -1
- package/dist/inference/tts.js +249 -77
- package/dist/inference/tts.js.map +1 -1
- package/dist/inference/tts.test.cjs +233 -0
- package/dist/inference/tts.test.cjs.map +1 -0
- package/dist/inference/tts.test.js +232 -0
- package/dist/inference/tts.test.js.map +1 -0
- package/dist/inference/utils.cjs +26 -7
- package/dist/inference/utils.cjs.map +1 -1
- package/dist/inference/utils.d.cts +14 -1
- package/dist/inference/utils.d.ts +14 -1
- package/dist/inference/utils.d.ts.map +1 -1
- package/dist/inference/utils.js +18 -2
- package/dist/inference/utils.js.map +1 -1
- package/dist/ipc/inference_proc_executor.cjs +6 -3
- package/dist/ipc/inference_proc_executor.cjs.map +1 -1
- package/dist/ipc/inference_proc_executor.d.ts.map +1 -1
- package/dist/ipc/inference_proc_executor.js +6 -3
- package/dist/ipc/inference_proc_executor.js.map +1 -1
- package/dist/ipc/inference_proc_lazy_main.cjs +13 -1
- package/dist/ipc/inference_proc_lazy_main.cjs.map +1 -1
- package/dist/ipc/inference_proc_lazy_main.js +13 -1
- package/dist/ipc/inference_proc_lazy_main.js.map +1 -1
- package/dist/ipc/job_proc_executor.cjs +6 -1
- package/dist/ipc/job_proc_executor.cjs.map +1 -1
- package/dist/ipc/job_proc_executor.d.ts.map +1 -1
- package/dist/ipc/job_proc_executor.js +6 -1
- package/dist/ipc/job_proc_executor.js.map +1 -1
- package/dist/ipc/job_proc_lazy_main.cjs +89 -17
- package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
- package/dist/ipc/job_proc_lazy_main.js +68 -18
- package/dist/ipc/job_proc_lazy_main.js.map +1 -1
- package/dist/ipc/supervised_proc.cjs +34 -8
- package/dist/ipc/supervised_proc.cjs.map +1 -1
- package/dist/ipc/supervised_proc.d.cts +8 -0
- package/dist/ipc/supervised_proc.d.ts +8 -0
- package/dist/ipc/supervised_proc.d.ts.map +1 -1
- package/dist/ipc/supervised_proc.js +34 -8
- package/dist/ipc/supervised_proc.js.map +1 -1
- package/dist/ipc/supervised_proc.test.cjs +145 -0
- package/dist/ipc/supervised_proc.test.cjs.map +1 -0
- package/dist/ipc/supervised_proc.test.js +122 -0
- package/dist/ipc/supervised_proc.test.js.map +1 -0
- package/dist/job.cjs +109 -1
- package/dist/job.cjs.map +1 -1
- package/dist/job.d.cts +14 -0
- package/dist/job.d.ts +14 -0
- package/dist/job.d.ts.map +1 -1
- package/dist/job.js +99 -1
- package/dist/job.js.map +1 -1
- package/dist/language.cjs +394 -0
- package/dist/language.cjs.map +1 -0
- package/dist/language.d.cts +15 -0
- package/dist/language.d.ts +15 -0
- package/dist/language.d.ts.map +1 -0
- package/dist/language.js +363 -0
- package/dist/language.js.map +1 -0
- package/dist/language.test.cjs +43 -0
- package/dist/language.test.cjs.map +1 -0
- package/dist/language.test.js +49 -0
- package/dist/language.test.js.map +1 -0
- package/dist/llm/chat_context.cjs +274 -3
- package/dist/llm/chat_context.cjs.map +1 -1
- package/dist/llm/chat_context.d.cts +86 -2
- package/dist/llm/chat_context.d.ts +86 -2
- package/dist/llm/chat_context.d.ts.map +1 -1
- package/dist/llm/chat_context.js +273 -3
- package/dist/llm/chat_context.js.map +1 -1
- package/dist/llm/chat_context.test.cjs +574 -0
- package/dist/llm/chat_context.test.cjs.map +1 -1
- package/dist/llm/chat_context.test.js +574 -0
- package/dist/llm/chat_context.test.js.map +1 -1
- package/dist/llm/fallback_adapter.cjs +278 -0
- package/dist/llm/fallback_adapter.cjs.map +1 -0
- package/dist/llm/fallback_adapter.d.cts +73 -0
- package/dist/llm/fallback_adapter.d.ts +73 -0
- package/dist/llm/fallback_adapter.d.ts.map +1 -0
- package/dist/llm/fallback_adapter.js +254 -0
- package/dist/llm/fallback_adapter.js.map +1 -0
- package/dist/llm/fallback_adapter.test.cjs +176 -0
- package/dist/llm/fallback_adapter.test.cjs.map +1 -0
- package/dist/llm/fallback_adapter.test.js +175 -0
- package/dist/llm/fallback_adapter.test.js.map +1 -0
- package/dist/llm/index.cjs +9 -0
- package/dist/llm/index.cjs.map +1 -1
- package/dist/llm/index.d.cts +4 -3
- package/dist/llm/index.d.ts +4 -3
- package/dist/llm/index.d.ts.map +1 -1
- package/dist/llm/index.js +11 -1
- package/dist/llm/index.js.map +1 -1
- package/dist/llm/llm.cjs +65 -11
- package/dist/llm/llm.cjs.map +1 -1
- package/dist/llm/llm.d.cts +13 -2
- package/dist/llm/llm.d.ts +13 -2
- package/dist/llm/llm.d.ts.map +1 -1
- package/dist/llm/llm.js +65 -11
- package/dist/llm/llm.js.map +1 -1
- package/dist/llm/provider_format/google.cjs +6 -2
- package/dist/llm/provider_format/google.cjs.map +1 -1
- package/dist/llm/provider_format/google.d.cts +1 -1
- package/dist/llm/provider_format/google.d.ts +1 -1
- package/dist/llm/provider_format/google.d.ts.map +1 -1
- package/dist/llm/provider_format/google.js +6 -2
- package/dist/llm/provider_format/google.js.map +1 -1
- package/dist/llm/provider_format/google.test.cjs +48 -0
- package/dist/llm/provider_format/google.test.cjs.map +1 -1
- package/dist/llm/provider_format/google.test.js +54 -1
- package/dist/llm/provider_format/google.test.js.map +1 -1
- package/dist/llm/provider_format/index.cjs +2 -0
- package/dist/llm/provider_format/index.cjs.map +1 -1
- package/dist/llm/provider_format/index.d.cts +2 -2
- package/dist/llm/provider_format/index.d.ts +2 -2
- package/dist/llm/provider_format/index.d.ts.map +1 -1
- package/dist/llm/provider_format/index.js +6 -1
- package/dist/llm/provider_format/index.js.map +1 -1
- package/dist/llm/provider_format/openai.cjs +126 -24
- package/dist/llm/provider_format/openai.cjs.map +1 -1
- package/dist/llm/provider_format/openai.d.cts +1 -0
- package/dist/llm/provider_format/openai.d.ts +1 -0
- package/dist/llm/provider_format/openai.d.ts.map +1 -1
- package/dist/llm/provider_format/openai.js +124 -23
- package/dist/llm/provider_format/openai.js.map +1 -1
- package/dist/llm/provider_format/openai.test.cjs +393 -0
- package/dist/llm/provider_format/openai.test.cjs.map +1 -1
- package/dist/llm/provider_format/openai.test.js +400 -2
- package/dist/llm/provider_format/openai.test.js.map +1 -1
- package/dist/llm/provider_format/utils.cjs +5 -4
- package/dist/llm/provider_format/utils.cjs.map +1 -1
- package/dist/llm/provider_format/utils.d.ts.map +1 -1
- package/dist/llm/provider_format/utils.js +5 -4
- package/dist/llm/provider_format/utils.js.map +1 -1
- package/dist/llm/realtime.cjs +3 -0
- package/dist/llm/realtime.cjs.map +1 -1
- package/dist/llm/realtime.d.cts +15 -1
- package/dist/llm/realtime.d.ts +15 -1
- package/dist/llm/realtime.d.ts.map +1 -1
- package/dist/llm/realtime.js +3 -0
- package/dist/llm/realtime.js.map +1 -1
- package/dist/llm/remote_chat_context.cjs.map +1 -1
- package/dist/llm/remote_chat_context.d.cts +2 -0
- package/dist/llm/remote_chat_context.d.ts +2 -0
- package/dist/llm/remote_chat_context.d.ts.map +1 -1
- package/dist/llm/remote_chat_context.js.map +1 -1
- package/dist/llm/tool_context.cjs +50 -2
- package/dist/llm/tool_context.cjs.map +1 -1
- package/dist/llm/tool_context.d.cts +47 -11
- package/dist/llm/tool_context.d.ts +47 -11
- package/dist/llm/tool_context.d.ts.map +1 -1
- package/dist/llm/tool_context.js +48 -3
- package/dist/llm/tool_context.js.map +1 -1
- package/dist/llm/tool_context.test.cjs +197 -0
- package/dist/llm/tool_context.test.cjs.map +1 -1
- package/dist/llm/tool_context.test.js +175 -0
- package/dist/llm/tool_context.test.js.map +1 -1
- package/dist/llm/utils.cjs +18 -12
- package/dist/llm/utils.cjs.map +1 -1
- package/dist/llm/utils.d.cts +2 -3
- package/dist/llm/utils.d.ts +2 -3
- package/dist/llm/utils.d.ts.map +1 -1
- package/dist/llm/utils.js +18 -12
- package/dist/llm/utils.js.map +1 -1
- package/dist/llm/zod-utils.cjs +102 -0
- package/dist/llm/zod-utils.cjs.map +1 -0
- package/dist/llm/zod-utils.d.cts +65 -0
- package/dist/llm/zod-utils.d.ts +65 -0
- package/dist/llm/zod-utils.d.ts.map +1 -0
- package/dist/llm/zod-utils.js +64 -0
- package/dist/llm/zod-utils.js.map +1 -0
- package/dist/llm/zod-utils.test.cjs +472 -0
- package/dist/llm/zod-utils.test.cjs.map +1 -0
- package/dist/llm/zod-utils.test.js +455 -0
- package/dist/llm/zod-utils.test.js.map +1 -0
- package/dist/log.cjs +45 -14
- package/dist/log.cjs.map +1 -1
- package/dist/log.d.cts +8 -1
- package/dist/log.d.ts +8 -1
- package/dist/log.d.ts.map +1 -1
- package/dist/log.js +45 -15
- package/dist/log.js.map +1 -1
- package/dist/metrics/base.cjs.map +1 -1
- package/dist/metrics/base.d.cts +75 -19
- package/dist/metrics/base.d.ts +75 -19
- package/dist/metrics/base.d.ts.map +1 -1
- package/dist/metrics/index.cjs +5 -0
- package/dist/metrics/index.cjs.map +1 -1
- package/dist/metrics/index.d.cts +2 -1
- package/dist/metrics/index.d.ts +2 -1
- package/dist/metrics/index.d.ts.map +1 -1
- package/dist/metrics/index.js +6 -0
- package/dist/metrics/index.js.map +1 -1
- package/dist/metrics/model_usage.cjs +189 -0
- package/dist/metrics/model_usage.cjs.map +1 -0
- package/dist/metrics/model_usage.d.cts +92 -0
- package/dist/metrics/model_usage.d.ts +92 -0
- package/dist/metrics/model_usage.d.ts.map +1 -0
- package/dist/metrics/model_usage.js +164 -0
- package/dist/metrics/model_usage.js.map +1 -0
- package/dist/metrics/model_usage.test.cjs +474 -0
- package/dist/metrics/model_usage.test.cjs.map +1 -0
- package/dist/metrics/model_usage.test.js +476 -0
- package/dist/metrics/model_usage.test.js.map +1 -0
- package/dist/metrics/usage_collector.cjs +5 -2
- package/dist/metrics/usage_collector.cjs.map +1 -1
- package/dist/metrics/usage_collector.d.cts +10 -1
- package/dist/metrics/usage_collector.d.ts +10 -1
- package/dist/metrics/usage_collector.d.ts.map +1 -1
- package/dist/metrics/usage_collector.js +5 -2
- package/dist/metrics/usage_collector.js.map +1 -1
- package/dist/metrics/utils.cjs +23 -7
- package/dist/metrics/utils.cjs.map +1 -1
- package/dist/metrics/utils.d.ts.map +1 -1
- package/dist/metrics/utils.js +23 -7
- package/dist/metrics/utils.js.map +1 -1
- package/dist/stream/deferred_stream.cjs +31 -10
- package/dist/stream/deferred_stream.cjs.map +1 -1
- package/dist/stream/deferred_stream.d.cts +6 -1
- package/dist/stream/deferred_stream.d.ts +6 -1
- package/dist/stream/deferred_stream.d.ts.map +1 -1
- package/dist/stream/deferred_stream.js +31 -10
- package/dist/stream/deferred_stream.js.map +1 -1
- package/dist/stream/deferred_stream.test.cjs +2 -2
- package/dist/stream/deferred_stream.test.cjs.map +1 -1
- package/dist/stream/deferred_stream.test.js +2 -2
- package/dist/stream/deferred_stream.test.js.map +1 -1
- package/dist/stream/index.cjs +3 -0
- package/dist/stream/index.cjs.map +1 -1
- package/dist/stream/index.d.cts +1 -0
- package/dist/stream/index.d.ts +1 -0
- package/dist/stream/index.d.ts.map +1 -1
- package/dist/stream/index.js +2 -0
- package/dist/stream/index.js.map +1 -1
- package/dist/stream/multi_input_stream.cjs +139 -0
- package/dist/stream/multi_input_stream.cjs.map +1 -0
- package/dist/stream/multi_input_stream.d.cts +55 -0
- package/dist/stream/multi_input_stream.d.ts +55 -0
- package/dist/stream/multi_input_stream.d.ts.map +1 -0
- package/dist/stream/multi_input_stream.js +115 -0
- package/dist/stream/multi_input_stream.js.map +1 -0
- package/dist/stream/multi_input_stream.test.cjs +344 -0
- package/dist/stream/multi_input_stream.test.cjs.map +1 -0
- package/dist/stream/multi_input_stream.test.js +343 -0
- package/dist/stream/multi_input_stream.test.js.map +1 -0
- package/dist/stream/stream_channel.cjs +39 -1
- package/dist/stream/stream_channel.cjs.map +1 -1
- package/dist/stream/stream_channel.d.cts +5 -2
- package/dist/stream/stream_channel.d.ts +5 -2
- package/dist/stream/stream_channel.d.ts.map +1 -1
- package/dist/stream/stream_channel.js +39 -1
- package/dist/stream/stream_channel.js.map +1 -1
- package/dist/stream/stream_channel.test.cjs +27 -0
- package/dist/stream/stream_channel.test.cjs.map +1 -1
- package/dist/stream/stream_channel.test.js +27 -0
- package/dist/stream/stream_channel.test.js.map +1 -1
- package/dist/stt/stream_adapter.cjs +24 -9
- package/dist/stt/stream_adapter.cjs.map +1 -1
- package/dist/stt/stream_adapter.d.cts +7 -3
- package/dist/stt/stream_adapter.d.ts +7 -3
- package/dist/stt/stream_adapter.d.ts.map +1 -1
- package/dist/stt/stream_adapter.js +24 -9
- package/dist/stt/stream_adapter.js.map +1 -1
- package/dist/stt/stt.cjs +86 -19
- package/dist/stt/stt.cjs.map +1 -1
- package/dist/stt/stt.d.cts +60 -5
- package/dist/stt/stt.d.ts +60 -5
- package/dist/stt/stt.d.ts.map +1 -1
- package/dist/stt/stt.js +88 -21
- package/dist/stt/stt.js.map +1 -1
- package/dist/telemetry/index.cjs +72 -0
- package/dist/telemetry/index.cjs.map +1 -0
- package/dist/telemetry/index.d.cts +7 -0
- package/dist/telemetry/index.d.ts +7 -0
- package/dist/telemetry/index.d.ts.map +1 -0
- package/dist/telemetry/index.js +37 -0
- package/dist/telemetry/index.js.map +1 -0
- package/dist/telemetry/logging.cjs +65 -0
- package/dist/telemetry/logging.cjs.map +1 -0
- package/dist/telemetry/logging.d.cts +21 -0
- package/dist/telemetry/logging.d.ts +21 -0
- package/dist/telemetry/logging.d.ts.map +1 -0
- package/dist/telemetry/logging.js +40 -0
- package/dist/telemetry/logging.js.map +1 -0
- package/dist/telemetry/otel_http_exporter.cjs +166 -0
- package/dist/telemetry/otel_http_exporter.cjs.map +1 -0
- package/dist/telemetry/otel_http_exporter.d.cts +63 -0
- package/dist/telemetry/otel_http_exporter.d.ts +63 -0
- package/dist/telemetry/otel_http_exporter.d.ts.map +1 -0
- package/dist/telemetry/otel_http_exporter.js +142 -0
- package/dist/telemetry/otel_http_exporter.js.map +1 -0
- package/dist/telemetry/pino_otel_transport.cjs +217 -0
- package/dist/telemetry/pino_otel_transport.cjs.map +1 -0
- package/dist/telemetry/pino_otel_transport.d.cts +58 -0
- package/dist/telemetry/pino_otel_transport.d.ts +58 -0
- package/dist/telemetry/pino_otel_transport.d.ts.map +1 -0
- package/dist/telemetry/pino_otel_transport.js +189 -0
- package/dist/telemetry/pino_otel_transport.js.map +1 -0
- package/dist/telemetry/trace_types.cjs +233 -0
- package/dist/telemetry/trace_types.cjs.map +1 -0
- package/dist/telemetry/trace_types.d.cts +74 -0
- package/dist/telemetry/trace_types.d.ts +74 -0
- package/dist/telemetry/trace_types.d.ts.map +1 -0
- package/dist/telemetry/trace_types.js +141 -0
- package/dist/telemetry/trace_types.js.map +1 -0
- package/dist/telemetry/traces.cjs +484 -0
- package/dist/telemetry/traces.cjs.map +1 -0
- package/dist/telemetry/traces.d.cts +116 -0
- package/dist/telemetry/traces.d.ts +116 -0
- package/dist/telemetry/traces.d.ts.map +1 -0
- package/dist/telemetry/traces.js +449 -0
- package/dist/telemetry/traces.js.map +1 -0
- package/dist/telemetry/utils.cjs +86 -0
- package/dist/telemetry/utils.cjs.map +1 -0
- package/dist/telemetry/utils.d.cts +5 -0
- package/dist/telemetry/utils.d.ts +5 -0
- package/dist/telemetry/utils.d.ts.map +1 -0
- package/dist/telemetry/utils.js +51 -0
- package/dist/telemetry/utils.js.map +1 -0
- package/dist/tokenize/basic/sentence.cjs +3 -3
- package/dist/tokenize/basic/sentence.cjs.map +1 -1
- package/dist/tokenize/basic/sentence.js +3 -3
- package/dist/tokenize/basic/sentence.js.map +1 -1
- package/dist/tokenize/tokenizer.test.cjs +3 -1
- package/dist/tokenize/tokenizer.test.cjs.map +1 -1
- package/dist/tokenize/tokenizer.test.js +3 -1
- package/dist/tokenize/tokenizer.test.js.map +1 -1
- package/dist/transcription.cjs.map +1 -1
- package/dist/transcription.d.cts +6 -0
- package/dist/transcription.d.ts +6 -0
- package/dist/transcription.d.ts.map +1 -1
- package/dist/transcription.js.map +1 -1
- package/dist/tts/fallback_adapter.cjs +466 -0
- package/dist/tts/fallback_adapter.cjs.map +1 -0
- package/dist/tts/fallback_adapter.d.cts +110 -0
- package/dist/tts/fallback_adapter.d.ts +110 -0
- package/dist/tts/fallback_adapter.d.ts.map +1 -0
- package/dist/tts/fallback_adapter.js +442 -0
- package/dist/tts/fallback_adapter.js.map +1 -0
- package/dist/tts/index.cjs +3 -0
- package/dist/tts/index.cjs.map +1 -1
- package/dist/tts/index.d.cts +1 -0
- package/dist/tts/index.d.ts +1 -0
- package/dist/tts/index.d.ts.map +1 -1
- package/dist/tts/index.js +2 -0
- package/dist/tts/index.js.map +1 -1
- package/dist/tts/stream_adapter.cjs +25 -8
- package/dist/tts/stream_adapter.cjs.map +1 -1
- package/dist/tts/stream_adapter.d.cts +6 -3
- package/dist/tts/stream_adapter.d.ts +6 -3
- package/dist/tts/stream_adapter.d.ts.map +1 -1
- package/dist/tts/stream_adapter.js +25 -8
- package/dist/tts/stream_adapter.js.map +1 -1
- package/dist/tts/tts.cjs +189 -57
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.cts +58 -6
- package/dist/tts/tts.d.ts +58 -6
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +191 -59
- package/dist/tts/tts.js.map +1 -1
- package/dist/types.cjs +24 -32
- package/dist/types.cjs.map +1 -1
- package/dist/types.d.cts +45 -10
- package/dist/types.d.ts +45 -10
- package/dist/types.d.ts.map +1 -1
- package/dist/types.js +20 -30
- package/dist/types.js.map +1 -1
- package/dist/utils.cjs +122 -26
- package/dist/utils.cjs.map +1 -1
- package/dist/utils.d.cts +41 -1
- package/dist/utils.d.ts +41 -1
- package/dist/utils.d.ts.map +1 -1
- package/dist/utils.js +117 -25
- package/dist/utils.js.map +1 -1
- package/dist/utils.test.cjs +73 -1
- package/dist/utils.test.cjs.map +1 -1
- package/dist/utils.test.js +74 -10
- package/dist/utils.test.js.map +1 -1
- package/dist/vad.cjs +35 -15
- package/dist/vad.cjs.map +1 -1
- package/dist/vad.d.cts +15 -5
- package/dist/vad.d.ts +15 -5
- package/dist/vad.d.ts.map +1 -1
- package/dist/vad.js +35 -15
- package/dist/vad.js.map +1 -1
- package/dist/version.cjs +1 -1
- package/dist/version.cjs.map +1 -1
- package/dist/version.d.cts +1 -1
- package/dist/version.d.ts +1 -1
- package/dist/version.d.ts.map +1 -1
- package/dist/version.js +1 -1
- package/dist/version.js.map +1 -1
- package/dist/voice/agent.cjs +258 -35
- package/dist/voice/agent.cjs.map +1 -1
- package/dist/voice/agent.d.cts +54 -13
- package/dist/voice/agent.d.ts +54 -13
- package/dist/voice/agent.d.ts.map +1 -1
- package/dist/voice/agent.js +254 -34
- package/dist/voice/agent.js.map +1 -1
- package/dist/voice/agent.test.cjs +314 -0
- package/dist/voice/agent.test.cjs.map +1 -1
- package/dist/voice/agent.test.js +316 -2
- package/dist/voice/agent.test.js.map +1 -1
- package/dist/voice/agent_activity.cjs +1116 -385
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.cts +72 -11
- package/dist/voice/agent_activity.d.ts +72 -11
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +1119 -383
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/agent_activity.test.cjs +135 -0
- package/dist/voice/agent_activity.test.cjs.map +1 -0
- package/dist/voice/agent_activity.test.js +134 -0
- package/dist/voice/agent_activity.test.js.map +1 -0
- package/dist/voice/agent_session.cjs +550 -90
- package/dist/voice/agent_session.cjs.map +1 -1
- package/dist/voice/agent_session.d.cts +185 -25
- package/dist/voice/agent_session.d.ts +185 -25
- package/dist/voice/agent_session.d.ts.map +1 -1
- package/dist/voice/agent_session.js +556 -91
- package/dist/voice/agent_session.js.map +1 -1
- package/dist/voice/audio_recognition.cjs +605 -46
- package/dist/voice/audio_recognition.cjs.map +1 -1
- package/dist/voice/audio_recognition.d.cts +96 -4
- package/dist/voice/audio_recognition.d.ts +96 -4
- package/dist/voice/audio_recognition.d.ts.map +1 -1
- package/dist/voice/audio_recognition.js +611 -47
- package/dist/voice/audio_recognition.js.map +1 -1
- package/dist/voice/audio_recognition_span.test.cjs +295 -0
- package/dist/voice/audio_recognition_span.test.cjs.map +1 -0
- package/dist/voice/audio_recognition_span.test.js +299 -0
- package/dist/voice/audio_recognition_span.test.js.map +1 -0
- package/dist/voice/avatar/datastream_io.cjs +7 -1
- package/dist/voice/avatar/datastream_io.cjs.map +1 -1
- package/dist/voice/avatar/datastream_io.d.cts +1 -0
- package/dist/voice/avatar/datastream_io.d.ts +1 -0
- package/dist/voice/avatar/datastream_io.d.ts.map +1 -1
- package/dist/voice/avatar/datastream_io.js +7 -1
- package/dist/voice/avatar/datastream_io.js.map +1 -1
- package/dist/voice/background_audio.cjs +367 -0
- package/dist/voice/background_audio.cjs.map +1 -0
- package/dist/voice/background_audio.d.cts +123 -0
- package/dist/voice/background_audio.d.ts +123 -0
- package/dist/voice/background_audio.d.ts.map +1 -0
- package/dist/voice/background_audio.js +343 -0
- package/dist/voice/background_audio.js.map +1 -0
- package/dist/voice/events.cjs +3 -0
- package/dist/voice/events.cjs.map +1 -1
- package/dist/voice/events.d.cts +16 -9
- package/dist/voice/events.d.ts +16 -9
- package/dist/voice/events.d.ts.map +1 -1
- package/dist/voice/events.js +3 -0
- package/dist/voice/events.js.map +1 -1
- package/dist/voice/generation.cjs +205 -41
- package/dist/voice/generation.cjs.map +1 -1
- package/dist/voice/generation.d.cts +21 -5
- package/dist/voice/generation.d.ts +21 -5
- package/dist/voice/generation.d.ts.map +1 -1
- package/dist/voice/generation.js +215 -43
- package/dist/voice/generation.js.map +1 -1
- package/dist/voice/generation_tools.test.cjs +236 -0
- package/dist/voice/generation_tools.test.cjs.map +1 -0
- package/dist/voice/generation_tools.test.js +235 -0
- package/dist/voice/generation_tools.test.js.map +1 -0
- package/dist/voice/index.cjs +33 -2
- package/dist/voice/index.cjs.map +1 -1
- package/dist/voice/index.d.cts +8 -2
- package/dist/voice/index.d.ts +8 -2
- package/dist/voice/index.d.ts.map +1 -1
- package/dist/voice/index.js +19 -2
- package/dist/voice/index.js.map +1 -1
- package/dist/voice/interruption_detection.test.cjs +114 -0
- package/dist/voice/interruption_detection.test.cjs.map +1 -0
- package/dist/voice/interruption_detection.test.js +113 -0
- package/dist/voice/interruption_detection.test.js.map +1 -0
- package/dist/voice/io.cjs +66 -6
- package/dist/voice/io.cjs.map +1 -1
- package/dist/voice/io.d.cts +67 -7
- package/dist/voice/io.d.ts +67 -7
- package/dist/voice/io.d.ts.map +1 -1
- package/dist/voice/io.js +62 -5
- package/dist/voice/io.js.map +1 -1
- package/dist/voice/recorder_io/index.cjs +23 -0
- package/dist/voice/recorder_io/index.cjs.map +1 -0
- package/dist/voice/recorder_io/index.d.cts +2 -0
- package/dist/voice/recorder_io/index.d.ts +2 -0
- package/dist/voice/recorder_io/index.d.ts.map +1 -0
- package/dist/voice/recorder_io/index.js +2 -0
- package/dist/voice/recorder_io/index.js.map +1 -0
- package/dist/voice/recorder_io/recorder_io.cjs +607 -0
- package/dist/voice/recorder_io/recorder_io.cjs.map +1 -0
- package/dist/voice/recorder_io/recorder_io.d.cts +106 -0
- package/dist/voice/recorder_io/recorder_io.d.ts +106 -0
- package/dist/voice/recorder_io/recorder_io.d.ts.map +1 -0
- package/dist/voice/recorder_io/recorder_io.js +573 -0
- package/dist/voice/recorder_io/recorder_io.js.map +1 -0
- package/dist/voice/remote_session.cjs +922 -0
- package/dist/voice/remote_session.cjs.map +1 -0
- package/dist/voice/remote_session.d.cts +108 -0
- package/dist/voice/remote_session.d.ts +108 -0
- package/dist/voice/remote_session.d.ts.map +1 -0
- package/dist/voice/remote_session.js +887 -0
- package/dist/voice/remote_session.js.map +1 -0
- package/dist/voice/report.cjs +88 -0
- package/dist/voice/report.cjs.map +1 -0
- package/dist/voice/report.d.cts +49 -0
- package/dist/voice/report.d.ts +49 -0
- package/dist/voice/report.d.ts.map +1 -0
- package/dist/voice/report.js +63 -0
- package/dist/voice/report.js.map +1 -0
- package/dist/voice/report.test.cjs +121 -0
- package/dist/voice/report.test.cjs.map +1 -0
- package/dist/voice/report.test.js +120 -0
- package/dist/voice/report.test.js.map +1 -0
- package/dist/voice/room_io/_input.cjs +40 -7
- package/dist/voice/room_io/_input.cjs.map +1 -1
- package/dist/voice/room_io/_input.d.cts +5 -2
- package/dist/voice/room_io/_input.d.ts +5 -2
- package/dist/voice/room_io/_input.d.ts.map +1 -1
- package/dist/voice/room_io/_input.js +41 -8
- package/dist/voice/room_io/_input.js.map +1 -1
- package/dist/voice/room_io/_output.cjs +19 -11
- package/dist/voice/room_io/_output.cjs.map +1 -1
- package/dist/voice/room_io/_output.d.cts +7 -4
- package/dist/voice/room_io/_output.d.ts +7 -4
- package/dist/voice/room_io/_output.d.ts.map +1 -1
- package/dist/voice/room_io/_output.js +20 -12
- package/dist/voice/room_io/_output.js.map +1 -1
- package/dist/voice/room_io/room_io.cjs +33 -6
- package/dist/voice/room_io/room_io.cjs.map +1 -1
- package/dist/voice/room_io/room_io.d.cts +29 -9
- package/dist/voice/room_io/room_io.d.ts +29 -9
- package/dist/voice/room_io/room_io.d.ts.map +1 -1
- package/dist/voice/room_io/room_io.js +33 -7
- package/dist/voice/room_io/room_io.js.map +1 -1
- package/dist/voice/speech_handle.cjs +22 -4
- package/dist/voice/speech_handle.cjs.map +1 -1
- package/dist/voice/speech_handle.d.cts +17 -2
- package/dist/voice/speech_handle.d.ts +17 -2
- package/dist/voice/speech_handle.d.ts.map +1 -1
- package/dist/voice/speech_handle.js +21 -4
- package/dist/voice/speech_handle.js.map +1 -1
- package/dist/voice/testing/fake_llm.cjs +127 -0
- package/dist/voice/testing/fake_llm.cjs.map +1 -0
- package/dist/voice/testing/fake_llm.d.cts +30 -0
- package/dist/voice/testing/fake_llm.d.ts +30 -0
- package/dist/voice/testing/fake_llm.d.ts.map +1 -0
- package/dist/voice/testing/fake_llm.js +103 -0
- package/dist/voice/testing/fake_llm.js.map +1 -0
- package/dist/voice/testing/index.cjs +57 -0
- package/dist/voice/testing/index.cjs.map +1 -0
- package/dist/voice/testing/index.d.cts +21 -0
- package/dist/voice/testing/index.d.ts +21 -0
- package/dist/voice/testing/index.d.ts.map +1 -0
- package/dist/voice/testing/index.js +35 -0
- package/dist/voice/testing/index.js.map +1 -0
- package/dist/voice/testing/run_result.cjs +817 -0
- package/dist/voice/testing/run_result.cjs.map +1 -0
- package/dist/voice/testing/run_result.d.cts +385 -0
- package/dist/voice/testing/run_result.d.ts +385 -0
- package/dist/voice/testing/run_result.d.ts.map +1 -0
- package/dist/voice/testing/run_result.js +790 -0
- package/dist/voice/testing/run_result.js.map +1 -0
- package/dist/voice/testing/types.cjs +46 -0
- package/dist/voice/testing/types.cjs.map +1 -0
- package/dist/voice/testing/types.d.cts +83 -0
- package/dist/voice/testing/types.d.ts +83 -0
- package/dist/voice/testing/types.d.ts.map +1 -0
- package/dist/voice/testing/types.js +19 -0
- package/dist/voice/testing/types.js.map +1 -0
- package/dist/voice/transcription/synchronizer.cjs +139 -15
- package/dist/voice/transcription/synchronizer.cjs.map +1 -1
- package/dist/voice/transcription/synchronizer.d.cts +35 -4
- package/dist/voice/transcription/synchronizer.d.ts +35 -4
- package/dist/voice/transcription/synchronizer.d.ts.map +1 -1
- package/dist/voice/transcription/synchronizer.js +143 -16
- package/dist/voice/transcription/synchronizer.js.map +1 -1
- package/dist/voice/transcription/synchronizer.test.cjs +151 -0
- package/dist/voice/transcription/synchronizer.test.cjs.map +1 -0
- package/dist/voice/transcription/synchronizer.test.js +150 -0
- package/dist/voice/transcription/synchronizer.test.js.map +1 -0
- package/dist/voice/turn_config/endpointing.cjs +33 -0
- package/dist/voice/turn_config/endpointing.cjs.map +1 -0
- package/dist/voice/turn_config/endpointing.d.cts +30 -0
- package/dist/voice/turn_config/endpointing.d.ts +30 -0
- package/dist/voice/turn_config/endpointing.d.ts.map +1 -0
- package/dist/voice/turn_config/endpointing.js +9 -0
- package/dist/voice/turn_config/endpointing.js.map +1 -0
- package/dist/voice/turn_config/interruption.cjs +37 -0
- package/dist/voice/turn_config/interruption.cjs.map +1 -0
- package/dist/voice/turn_config/interruption.d.cts +53 -0
- package/dist/voice/turn_config/interruption.d.ts +53 -0
- package/dist/voice/turn_config/interruption.d.ts.map +1 -0
- package/dist/voice/turn_config/interruption.js +13 -0
- package/dist/voice/turn_config/interruption.js.map +1 -0
- package/dist/voice/turn_config/turn_handling.cjs +35 -0
- package/dist/voice/turn_config/turn_handling.cjs.map +1 -0
- package/dist/voice/turn_config/turn_handling.d.cts +36 -0
- package/dist/voice/turn_config/turn_handling.d.ts +36 -0
- package/dist/voice/turn_config/turn_handling.d.ts.map +1 -0
- package/dist/voice/turn_config/turn_handling.js +11 -0
- package/dist/voice/turn_config/turn_handling.js.map +1 -0
- package/dist/voice/turn_config/utils.cjs +157 -0
- package/dist/voice/turn_config/utils.cjs.map +1 -0
- package/dist/voice/turn_config/utils.d.cts +37 -0
- package/dist/voice/turn_config/utils.d.ts +37 -0
- package/dist/voice/turn_config/utils.d.ts.map +1 -0
- package/dist/voice/turn_config/utils.js +131 -0
- package/dist/voice/turn_config/utils.js.map +1 -0
- package/dist/voice/turn_config/utils.test.cjs +128 -0
- package/dist/voice/turn_config/utils.test.cjs.map +1 -0
- package/dist/voice/turn_config/utils.test.js +127 -0
- package/dist/voice/turn_config/utils.test.js.map +1 -0
- package/dist/voice/utils.cjs +47 -0
- package/dist/voice/utils.cjs.map +1 -0
- package/dist/voice/utils.d.cts +4 -0
- package/dist/voice/utils.d.ts +4 -0
- package/dist/voice/utils.d.ts.map +1 -0
- package/dist/voice/utils.js +23 -0
- package/dist/voice/utils.js.map +1 -0
- package/dist/worker.cjs +44 -52
- package/dist/worker.cjs.map +1 -1
- package/dist/worker.d.cts +18 -8
- package/dist/worker.d.ts +18 -8
- package/dist/worker.d.ts.map +1 -1
- package/dist/worker.js +43 -43
- package/dist/worker.js.map +1 -1
- package/package.json +32 -12
- package/resources/NOTICE +2 -0
- package/resources/keyboard-typing.ogg +0 -0
- package/resources/keyboard-typing2.ogg +0 -0
- package/resources/office-ambience.ogg +0 -0
- package/src/audio.ts +132 -1
- package/src/beta/index.ts +9 -0
- package/src/beta/workflows/index.ts +9 -0
- package/src/beta/workflows/task_group.ts +194 -0
- package/src/cli.ts +57 -66
- package/src/connection_pool.test.ts +346 -0
- package/src/connection_pool.ts +307 -0
- package/src/constants.ts +14 -0
- package/src/cpu.test.ts +239 -0
- package/src/cpu.ts +173 -0
- package/src/http_server.ts +18 -6
- package/src/index.ts +15 -13
- package/src/inference/api_protos.ts +85 -2
- package/src/inference/index.ts +32 -4
- package/src/inference/interruption/defaults.ts +51 -0
- package/src/inference/interruption/errors.ts +25 -0
- package/src/inference/interruption/http_transport.ts +206 -0
- package/src/inference/interruption/interruption_cache_entry.ts +50 -0
- package/src/inference/interruption/interruption_detector.ts +204 -0
- package/src/inference/interruption/interruption_stream.ts +467 -0
- package/src/inference/interruption/types.ts +84 -0
- package/src/inference/interruption/utils.test.ts +132 -0
- package/src/inference/interruption/utils.ts +137 -0
- package/src/inference/interruption/ws_transport.ts +406 -0
- package/src/inference/llm.ts +214 -163
- package/src/inference/stt.test.ts +253 -0
- package/src/inference/stt.ts +449 -208
- package/src/inference/tts.test.ts +267 -0
- package/src/inference/tts.ts +377 -115
- package/src/inference/utils.ts +30 -2
- package/src/ipc/inference_proc_executor.ts +11 -3
- package/src/ipc/inference_proc_lazy_main.ts +13 -1
- package/src/ipc/job_proc_executor.ts +11 -1
- package/src/ipc/job_proc_lazy_main.ts +86 -20
- package/src/ipc/supervised_proc.test.ts +153 -0
- package/src/ipc/supervised_proc.ts +39 -10
- package/src/job.ts +120 -1
- package/src/language.test.ts +62 -0
- package/src/language.ts +380 -0
- package/src/llm/__snapshots__/zod-utils.test.ts.snap +559 -0
- package/src/llm/chat_context.test.ts +655 -0
- package/src/llm/chat_context.ts +412 -2
- package/src/llm/fallback_adapter.test.ts +238 -0
- package/src/llm/fallback_adapter.ts +391 -0
- package/src/llm/index.ts +11 -0
- package/src/llm/llm.ts +77 -12
- package/src/llm/provider_format/google.test.ts +72 -1
- package/src/llm/provider_format/google.ts +10 -6
- package/src/llm/provider_format/index.ts +7 -2
- package/src/llm/provider_format/openai.test.ts +480 -2
- package/src/llm/provider_format/openai.ts +152 -21
- package/src/llm/provider_format/utils.ts +11 -5
- package/src/llm/realtime.ts +23 -2
- package/src/llm/remote_chat_context.ts +2 -2
- package/src/llm/tool_context.test.ts +210 -1
- package/src/llm/tool_context.ts +115 -17
- package/src/llm/utils.ts +24 -16
- package/src/llm/zod-utils.test.ts +577 -0
- package/src/llm/zod-utils.ts +153 -0
- package/src/log.ts +71 -19
- package/src/metrics/base.ts +78 -19
- package/src/metrics/index.ts +12 -0
- package/src/metrics/model_usage.test.ts +545 -0
- package/src/metrics/model_usage.ts +262 -0
- package/src/metrics/usage_collector.ts +14 -3
- package/src/metrics/utils.ts +27 -7
- package/src/stream/deferred_stream.test.ts +3 -3
- package/src/stream/deferred_stream.ts +43 -11
- package/src/stream/index.ts +1 -0
- package/src/stream/multi_input_stream.test.ts +545 -0
- package/src/stream/multi_input_stream.ts +172 -0
- package/src/stream/stream_channel.test.ts +37 -0
- package/src/stream/stream_channel.ts +43 -3
- package/src/stt/stream_adapter.ts +30 -9
- package/src/stt/stt.ts +131 -22
- package/src/telemetry/index.ts +28 -0
- package/src/telemetry/logging.ts +55 -0
- package/src/telemetry/otel_http_exporter.ts +218 -0
- package/src/telemetry/pino_otel_transport.ts +265 -0
- package/src/telemetry/trace_types.ts +109 -0
- package/src/telemetry/traces.ts +673 -0
- package/src/telemetry/utils.ts +61 -0
- package/src/tokenize/basic/sentence.ts +3 -3
- package/src/tokenize/tokenizer.test.ts +4 -0
- package/src/transcription.ts +6 -0
- package/src/tts/fallback_adapter.ts +579 -0
- package/src/tts/index.ts +1 -0
- package/src/tts/stream_adapter.ts +38 -8
- package/src/tts/tts.ts +245 -62
- package/src/types.ts +62 -33
- package/src/utils.test.ts +90 -10
- package/src/utils.ts +176 -31
- package/src/vad.ts +42 -18
- package/src/version.ts +1 -1
- package/src/voice/agent.test.ts +347 -2
- package/src/voice/agent.ts +346 -44
- package/src/voice/agent_activity.test.ts +194 -0
- package/src/voice/agent_activity.ts +1457 -388
- package/src/voice/agent_session.ts +817 -112
- package/src/voice/audio_recognition.ts +845 -70
- package/src/voice/audio_recognition_span.test.ts +341 -0
- package/src/voice/avatar/datastream_io.ts +9 -1
- package/src/voice/background_audio.ts +494 -0
- package/src/voice/events.ts +27 -7
- package/src/voice/generation.ts +310 -56
- package/src/voice/generation_tools.test.ts +268 -0
- package/src/voice/index.ts +17 -3
- package/src/voice/interruption_detection.test.ts +151 -0
- package/src/voice/io.ts +115 -12
- package/src/voice/recorder_io/index.ts +4 -0
- package/src/voice/recorder_io/recorder_io.ts +783 -0
- package/src/voice/remote_session.ts +1083 -0
- package/src/voice/report.test.ts +136 -0
- package/src/voice/report.ts +140 -0
- package/src/voice/room_io/_input.ts +45 -10
- package/src/voice/room_io/_output.ts +26 -14
- package/src/voice/room_io/room_io.ts +67 -22
- package/src/voice/speech_handle.ts +38 -6
- package/src/voice/testing/fake_llm.ts +138 -0
- package/src/voice/testing/index.ts +52 -0
- package/src/voice/testing/run_result.ts +995 -0
- package/src/voice/testing/types.ts +118 -0
- package/src/voice/transcription/synchronizer.test.ts +206 -0
- package/src/voice/transcription/synchronizer.ts +204 -19
- package/src/voice/turn_config/endpointing.ts +33 -0
- package/src/voice/turn_config/interruption.ts +56 -0
- package/src/voice/turn_config/turn_handling.ts +45 -0
- package/src/voice/turn_config/utils.test.ts +148 -0
- package/src/voice/turn_config/utils.ts +167 -0
- package/src/voice/utils.ts +29 -0
- package/src/worker.ts +92 -78
- package/src/llm/__snapshots__/utils.test.ts.snap +0 -65
|
@@ -92,5 +92,32 @@ describe("StreamChannel", () => {
|
|
|
92
92
|
const nextResult = await reader.read();
|
|
93
93
|
expect(nextResult.done).toBe(true);
|
|
94
94
|
});
|
|
95
|
+
it("should gracefully handle close while read is pending", async () => {
|
|
96
|
+
const channel = createStreamChannel();
|
|
97
|
+
const reader = channel.stream().getReader();
|
|
98
|
+
const readPromise = reader.read();
|
|
99
|
+
await channel.close();
|
|
100
|
+
const result = await readPromise;
|
|
101
|
+
expect(result.done).toBe(true);
|
|
102
|
+
expect(result.value).toBeUndefined();
|
|
103
|
+
});
|
|
104
|
+
it("should complete all pending reads when closed", async () => {
|
|
105
|
+
const channel = createStreamChannel();
|
|
106
|
+
const reader = channel.stream().getReader();
|
|
107
|
+
const read1 = reader.read();
|
|
108
|
+
const read2 = reader.read();
|
|
109
|
+
const read3 = reader.read();
|
|
110
|
+
await channel.write(42);
|
|
111
|
+
await channel.write(43);
|
|
112
|
+
await channel.close();
|
|
113
|
+
const result1 = await read1;
|
|
114
|
+
expect(result1.done).toBe(false);
|
|
115
|
+
expect(result1.value).toBe(42);
|
|
116
|
+
const result2 = await read2;
|
|
117
|
+
expect(result2.done).toBe(false);
|
|
118
|
+
expect(result2.value).toBe(43);
|
|
119
|
+
const result3 = await read3;
|
|
120
|
+
expect(result3.done).toBe(true);
|
|
121
|
+
});
|
|
95
122
|
});
|
|
96
123
|
//# sourceMappingURL=stream_channel.test.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/stream/stream_channel.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { describe, expect, it } from 'vitest';\nimport { createStreamChannel } from './stream_channel.js';\n\ndescribe('StreamChannel', () => {\n it('should write and read a single value', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n await channel.write('test value');\n await channel.close();\n\n const result = await reader.read();\n expect(result.done).toBe(false);\n expect(result.value).toBe('test value');\n\n const nextResult = await reader.read();\n expect(nextResult.done).toBe(true);\n });\n\n it('should write and read multiple values in sequence', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n const testValues = ['first', 'second', 'third'];\n\n for (const value of testValues) {\n await channel.write(value);\n }\n await channel.close();\n\n const results: string[] = [];\n let result = await reader.read();\n while (!result.done) {\n results.push(result.value);\n result = await reader.read();\n }\n\n expect(results).toEqual(testValues);\n });\n\n it('should handle arrays', async () => {\n const channel = createStreamChannel<number[]>();\n const reader = channel.stream().getReader();\n\n const testArray = [1, 2, 3, 4, 5];\n await channel.write(testArray);\n await channel.close();\n\n const result = await reader.read();\n expect(result.value).toEqual(testArray);\n expect(result.value).toBe(testArray);\n });\n\n it('should work with concurrent writing and reading', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n const testData = ['chunk1', 'chunk2', 'chunk3'];\n const results: string[] = [];\n\n const readPromise = (async () => {\n let result = await reader.read();\n while (!result.done) {\n results.push(result.value);\n result = await reader.read();\n }\n })();\n\n for (const chunk of testData) {\n await channel.write(chunk);\n }\n await channel.close();\n\n await readPromise;\n expect(results).toEqual(testData);\n });\n\n it('should handle empty stream', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n await channel.close();\n\n const result = await reader.read();\n expect(result.done).toBe(true);\n });\n\n it('should handle non-awaited sequential writes', async () => {\n const channel = createStreamChannel<number>();\n const reader = channel.stream().getReader();\n\n const testNumbers = Array.from({ length: 100 }, (_, i) => i);\n\n for (const num of testNumbers) {\n channel.write(num);\n }\n channel.close();\n\n const results: number[] = [];\n let result = await reader.read();\n while (!result.done) {\n results.push(result.value);\n result = await reader.read();\n }\n\n expect(results).toEqual(testNumbers);\n });\n\n it('should handle double closing without error', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n await channel.write('test');\n\n await channel.close();\n // Close again - should not throw\n await expect(channel.close()).resolves.toBeUndefined();\n\n const result = await reader.read();\n expect(result.done).toBe(false);\n expect(result.value).toBe('test');\n\n const nextResult = await reader.read();\n expect(nextResult.done).toBe(true);\n });\n});\n"],"mappings":"AAGA,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAS,2BAA2B;AAEpC,SAAS,iBAAiB,MAAM;AAC9B,KAAG,wCAAwC,YAAY;AACrD,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,MAAM,YAAY;AAChC,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,IAAI,EAAE,KAAK,KAAK;AAC9B,WAAO,OAAO,KAAK,EAAE,KAAK,YAAY;AAEtC,UAAM,aAAa,MAAM,OAAO,KAAK;AACrC,WAAO,WAAW,IAAI,EAAE,KAAK,IAAI;AAAA,EACnC,CAAC;AAED,KAAG,qDAAqD,YAAY;AAClE,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,aAAa,CAAC,SAAS,UAAU,OAAO;AAE9C,eAAW,SAAS,YAAY;AAC9B,YAAM,QAAQ,MAAM,KAAK;AAAA,IAC3B;AACA,UAAM,QAAQ,MAAM;AAEpB,UAAM,UAAoB,CAAC;AAC3B,QAAI,SAAS,MAAM,OAAO,KAAK;AAC/B,WAAO,CAAC,OAAO,MAAM;AACnB,cAAQ,KAAK,OAAO,KAAK;AACzB,eAAS,MAAM,OAAO,KAAK;AAAA,IAC7B;AAEA,WAAO,OAAO,EAAE,QAAQ,UAAU;AAAA,EACpC,CAAC;AAED,KAAG,wBAAwB,YAAY;AACrC,UAAM,UAAU,oBAA8B;AAC9C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,YAAY,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC;AAChC,UAAM,QAAQ,MAAM,SAAS;AAC7B,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,KAAK,EAAE,QAAQ,SAAS;AACtC,WAAO,OAAO,KAAK,EAAE,KAAK,SAAS;AAAA,EACrC,CAAC;AAED,KAAG,mDAAmD,YAAY;AAChE,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,WAAW,CAAC,UAAU,UAAU,QAAQ;AAC9C,UAAM,UAAoB,CAAC;AAE3B,UAAM,eAAe,YAAY;AAC/B,UAAI,SAAS,MAAM,OAAO,KAAK;AAC/B,aAAO,CAAC,OAAO,MAAM;AACnB,gBAAQ,KAAK,OAAO,KAAK;AACzB,iBAAS,MAAM,OAAO,KAAK;AAAA,MAC7B;AAAA,IACF,GAAG;AAEH,eAAW,SAAS,UAAU;AAC5B,YAAM,QAAQ,MAAM,KAAK;AAAA,IAC3B;AACA,UAAM,QAAQ,MAAM;AAEpB,UAAM;AACN,WAAO,OAAO,EAAE,QAAQ,QAAQ;AAAA,EAClC,CAAC;AAED,KAAG,8BAA8B,YAAY;AAC3C,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,IAAI,EAAE,KAAK,IAAI;AAAA,EAC/B,CAAC;AAED,KAAG,+CAA+C,YAAY;AAC5D,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,cAAc,MAAM,KAAK,EAAE,QAAQ,IAAI,GAAG,CAAC,GAAG,MAAM,CAAC;AAE3D,eAAW,OAAO,aAAa;AAC7B,cAAQ,MAAM,GAAG;AAAA,IACnB;AACA,YAAQ,MAAM;AAEd,UAAM,UAAoB,CAAC;AAC3B,QAAI,SAAS,MAAM,OAAO,KAAK;AAC/B,WAAO,CAAC,OAAO,MAAM;AACnB,cAAQ,KAAK,OAAO,KAAK;AACzB,eAAS,MAAM,OAAO,KAAK;AAAA,IAC7B;AAEA,WAAO,OAAO,EAAE,QAAQ,WAAW;AAAA,EACrC,CAAC;AAED,KAAG,8CAA8C,YAAY;AAC3D,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,MAAM,MAAM;AAE1B,UAAM,QAAQ,MAAM;AAEpB,UAAM,OAAO,QAAQ,MAAM,CAAC,EAAE,SAAS,cAAc;AAErD,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,IAAI,EAAE,KAAK,KAAK;AAC9B,WAAO,OAAO,KAAK,EAAE,KAAK,MAAM;AAEhC,UAAM,aAAa,MAAM,OAAO,KAAK;AACrC,WAAO,WAAW,IAAI,EAAE,KAAK,IAAI;AAAA,EACnC,CAAC;AACH,CAAC;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/stream/stream_channel.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { describe, expect, it } from 'vitest';\nimport { createStreamChannel } from './stream_channel.js';\n\ndescribe('StreamChannel', () => {\n it('should write and read a single value', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n await channel.write('test value');\n await channel.close();\n\n const result = await reader.read();\n expect(result.done).toBe(false);\n expect(result.value).toBe('test value');\n\n const nextResult = await reader.read();\n expect(nextResult.done).toBe(true);\n });\n\n it('should write and read multiple values in sequence', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n const testValues = ['first', 'second', 'third'];\n\n for (const value of testValues) {\n await channel.write(value);\n }\n await channel.close();\n\n const results: string[] = [];\n let result = await reader.read();\n while (!result.done) {\n results.push(result.value);\n result = await reader.read();\n }\n\n expect(results).toEqual(testValues);\n });\n\n it('should handle arrays', async () => {\n const channel = createStreamChannel<number[]>();\n const reader = channel.stream().getReader();\n\n const testArray = [1, 2, 3, 4, 5];\n await channel.write(testArray);\n await channel.close();\n\n const result = await reader.read();\n expect(result.value).toEqual(testArray);\n expect(result.value).toBe(testArray);\n });\n\n it('should work with concurrent writing and reading', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n const testData = ['chunk1', 'chunk2', 'chunk3'];\n const results: string[] = [];\n\n const readPromise = (async () => {\n let result = await reader.read();\n while (!result.done) {\n results.push(result.value);\n result = await reader.read();\n }\n })();\n\n for (const chunk of testData) {\n await channel.write(chunk);\n }\n await channel.close();\n\n await readPromise;\n expect(results).toEqual(testData);\n });\n\n it('should handle empty stream', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n await channel.close();\n\n const result = await reader.read();\n expect(result.done).toBe(true);\n });\n\n it('should handle non-awaited sequential writes', async () => {\n const channel = createStreamChannel<number>();\n const reader = channel.stream().getReader();\n\n const testNumbers = Array.from({ length: 100 }, (_, i) => i);\n\n for (const num of testNumbers) {\n channel.write(num);\n }\n channel.close();\n\n const results: number[] = [];\n let result = await reader.read();\n while (!result.done) {\n results.push(result.value);\n result = await reader.read();\n }\n\n expect(results).toEqual(testNumbers);\n });\n\n it('should handle double closing without error', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n await channel.write('test');\n\n await channel.close();\n // Close again - should not throw\n await expect(channel.close()).resolves.toBeUndefined();\n\n const result = await reader.read();\n expect(result.done).toBe(false);\n expect(result.value).toBe('test');\n\n const nextResult = await reader.read();\n expect(nextResult.done).toBe(true);\n });\n\n it('should gracefully handle close while read is pending', async () => {\n const channel = createStreamChannel<string>();\n const reader = channel.stream().getReader();\n\n const readPromise = reader.read();\n\n await channel.close();\n\n const result = await readPromise;\n expect(result.done).toBe(true);\n expect(result.value).toBeUndefined();\n });\n\n it('should complete all pending reads when closed', async () => {\n const channel = createStreamChannel<number>();\n const reader = channel.stream().getReader();\n\n const read1 = reader.read();\n const read2 = reader.read();\n const read3 = reader.read();\n\n await channel.write(42);\n await channel.write(43);\n await channel.close();\n\n const result1 = await read1;\n expect(result1.done).toBe(false);\n expect(result1.value).toBe(42);\n\n const result2 = await read2;\n expect(result2.done).toBe(false);\n expect(result2.value).toBe(43);\n\n const result3 = await read3;\n expect(result3.done).toBe(true);\n });\n});\n"],"mappings":"AAGA,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAS,2BAA2B;AAEpC,SAAS,iBAAiB,MAAM;AAC9B,KAAG,wCAAwC,YAAY;AACrD,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,MAAM,YAAY;AAChC,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,IAAI,EAAE,KAAK,KAAK;AAC9B,WAAO,OAAO,KAAK,EAAE,KAAK,YAAY;AAEtC,UAAM,aAAa,MAAM,OAAO,KAAK;AACrC,WAAO,WAAW,IAAI,EAAE,KAAK,IAAI;AAAA,EACnC,CAAC;AAED,KAAG,qDAAqD,YAAY;AAClE,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,aAAa,CAAC,SAAS,UAAU,OAAO;AAE9C,eAAW,SAAS,YAAY;AAC9B,YAAM,QAAQ,MAAM,KAAK;AAAA,IAC3B;AACA,UAAM,QAAQ,MAAM;AAEpB,UAAM,UAAoB,CAAC;AAC3B,QAAI,SAAS,MAAM,OAAO,KAAK;AAC/B,WAAO,CAAC,OAAO,MAAM;AACnB,cAAQ,KAAK,OAAO,KAAK;AACzB,eAAS,MAAM,OAAO,KAAK;AAAA,IAC7B;AAEA,WAAO,OAAO,EAAE,QAAQ,UAAU;AAAA,EACpC,CAAC;AAED,KAAG,wBAAwB,YAAY;AACrC,UAAM,UAAU,oBAA8B;AAC9C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,YAAY,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC;AAChC,UAAM,QAAQ,MAAM,SAAS;AAC7B,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,KAAK,EAAE,QAAQ,SAAS;AACtC,WAAO,OAAO,KAAK,EAAE,KAAK,SAAS;AAAA,EACrC,CAAC;AAED,KAAG,mDAAmD,YAAY;AAChE,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,WAAW,CAAC,UAAU,UAAU,QAAQ;AAC9C,UAAM,UAAoB,CAAC;AAE3B,UAAM,eAAe,YAAY;AAC/B,UAAI,SAAS,MAAM,OAAO,KAAK;AAC/B,aAAO,CAAC,OAAO,MAAM;AACnB,gBAAQ,KAAK,OAAO,KAAK;AACzB,iBAAS,MAAM,OAAO,KAAK;AAAA,MAC7B;AAAA,IACF,GAAG;AAEH,eAAW,SAAS,UAAU;AAC5B,YAAM,QAAQ,MAAM,KAAK;AAAA,IAC3B;AACA,UAAM,QAAQ,MAAM;AAEpB,UAAM;AACN,WAAO,OAAO,EAAE,QAAQ,QAAQ;AAAA,EAClC,CAAC;AAED,KAAG,8BAA8B,YAAY;AAC3C,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,IAAI,EAAE,KAAK,IAAI;AAAA,EAC/B,CAAC;AAED,KAAG,+CAA+C,YAAY;AAC5D,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,cAAc,MAAM,KAAK,EAAE,QAAQ,IAAI,GAAG,CAAC,GAAG,MAAM,CAAC;AAE3D,eAAW,OAAO,aAAa;AAC7B,cAAQ,MAAM,GAAG;AAAA,IACnB;AACA,YAAQ,MAAM;AAEd,UAAM,UAAoB,CAAC;AAC3B,QAAI,SAAS,MAAM,OAAO,KAAK;AAC/B,WAAO,CAAC,OAAO,MAAM;AACnB,cAAQ,KAAK,OAAO,KAAK;AACzB,eAAS,MAAM,OAAO,KAAK;AAAA,IAC7B;AAEA,WAAO,OAAO,EAAE,QAAQ,WAAW;AAAA,EACrC,CAAC;AAED,KAAG,8CAA8C,YAAY;AAC3D,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,MAAM,MAAM;AAE1B,UAAM,QAAQ,MAAM;AAEpB,UAAM,OAAO,QAAQ,MAAM,CAAC,EAAE,SAAS,cAAc;AAErD,UAAM,SAAS,MAAM,OAAO,KAAK;AACjC,WAAO,OAAO,IAAI,EAAE,KAAK,KAAK;AAC9B,WAAO,OAAO,KAAK,EAAE,KAAK,MAAM;AAEhC,UAAM,aAAa,MAAM,OAAO,KAAK;AACrC,WAAO,WAAW,IAAI,EAAE,KAAK,IAAI;AAAA,EACnC,CAAC;AAED,KAAG,wDAAwD,YAAY;AACrE,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,cAAc,OAAO,KAAK;AAEhC,UAAM,QAAQ,MAAM;AAEpB,UAAM,SAAS,MAAM;AACrB,WAAO,OAAO,IAAI,EAAE,KAAK,IAAI;AAC7B,WAAO,OAAO,KAAK,EAAE,cAAc;AAAA,EACrC,CAAC;AAED,KAAG,iDAAiD,YAAY;AAC9D,UAAM,UAAU,oBAA4B;AAC5C,UAAM,SAAS,QAAQ,OAAO,EAAE,UAAU;AAE1C,UAAM,QAAQ,OAAO,KAAK;AAC1B,UAAM,QAAQ,OAAO,KAAK;AAC1B,UAAM,QAAQ,OAAO,KAAK;AAE1B,UAAM,QAAQ,MAAM,EAAE;AACtB,UAAM,QAAQ,MAAM,EAAE;AACtB,UAAM,QAAQ,MAAM;AAEpB,UAAM,UAAU,MAAM;AACtB,WAAO,QAAQ,IAAI,EAAE,KAAK,KAAK;AAC/B,WAAO,QAAQ,KAAK,EAAE,KAAK,EAAE;AAE7B,UAAM,UAAU,MAAM;AACtB,WAAO,QAAQ,IAAI,EAAE,KAAK,KAAK;AAC/B,WAAO,QAAQ,KAAK,EAAE,KAAK,EAAE;AAE7B,UAAM,UAAU,MAAM;AACtB,WAAO,QAAQ,IAAI,EAAE,KAAK,IAAI;AAAA,EAChC,CAAC;AACH,CAAC;","names":[]}
|
|
@@ -23,6 +23,7 @@ __export(stream_adapter_exports, {
|
|
|
23
23
|
});
|
|
24
24
|
module.exports = __toCommonJS(stream_adapter_exports);
|
|
25
25
|
var import_log = require("../log.cjs");
|
|
26
|
+
var import_utils = require("../utils.cjs");
|
|
26
27
|
var import_vad = require("../vad.cjs");
|
|
27
28
|
var import_stt = require("./stt.cjs");
|
|
28
29
|
class StreamAdapter extends import_stt.STT {
|
|
@@ -37,24 +38,31 @@ class StreamAdapter extends import_stt.STT {
|
|
|
37
38
|
this.#stt.on("metrics_collected", (metrics) => {
|
|
38
39
|
this.emit("metrics_collected", metrics);
|
|
39
40
|
});
|
|
41
|
+
this.#stt.on("error", (error) => {
|
|
42
|
+
this.emit("error", error);
|
|
43
|
+
});
|
|
40
44
|
}
|
|
41
|
-
_recognize(frame) {
|
|
42
|
-
return this.#stt.recognize(frame);
|
|
45
|
+
_recognize(frame, abortSignal) {
|
|
46
|
+
return this.#stt.recognize(frame, abortSignal);
|
|
43
47
|
}
|
|
44
|
-
stream() {
|
|
45
|
-
return new StreamAdapterWrapper(this.#stt, this.#vad);
|
|
48
|
+
stream(options) {
|
|
49
|
+
return new StreamAdapterWrapper(this.#stt, this.#vad, options == null ? void 0 : options.connOptions);
|
|
46
50
|
}
|
|
47
51
|
}
|
|
48
52
|
class StreamAdapterWrapper extends import_stt.SpeechStream {
|
|
49
53
|
#stt;
|
|
50
54
|
#vadStream;
|
|
51
55
|
label;
|
|
52
|
-
constructor(stt, vad) {
|
|
53
|
-
super(stt);
|
|
56
|
+
constructor(stt, vad, connOptions) {
|
|
57
|
+
super(stt, void 0, connOptions);
|
|
54
58
|
this.#stt = stt;
|
|
55
59
|
this.#vadStream = vad.stream();
|
|
56
60
|
this.label = `stt.StreamAdapterWrapper<${this.#stt.label}>`;
|
|
57
61
|
}
|
|
62
|
+
close() {
|
|
63
|
+
super.close();
|
|
64
|
+
this.#vadStream.close();
|
|
65
|
+
}
|
|
58
66
|
async monitorMetrics() {
|
|
59
67
|
return;
|
|
60
68
|
}
|
|
@@ -67,7 +75,14 @@ class StreamAdapterWrapper extends import_stt.SpeechStream {
|
|
|
67
75
|
this.#vadStream.pushFrame(input);
|
|
68
76
|
}
|
|
69
77
|
}
|
|
70
|
-
|
|
78
|
+
try {
|
|
79
|
+
this.#vadStream.endInput();
|
|
80
|
+
} catch (e) {
|
|
81
|
+
if ((0, import_utils.isStreamClosedError)(e)) {
|
|
82
|
+
return;
|
|
83
|
+
}
|
|
84
|
+
throw e;
|
|
85
|
+
}
|
|
71
86
|
};
|
|
72
87
|
const recognize = async () => {
|
|
73
88
|
for await (const ev of this.#vadStream) {
|
|
@@ -78,7 +93,7 @@ class StreamAdapterWrapper extends import_stt.SpeechStream {
|
|
|
78
93
|
case import_vad.VADEventType.END_OF_SPEECH:
|
|
79
94
|
this.output.put({ type: import_stt.SpeechEventType.END_OF_SPEECH });
|
|
80
95
|
try {
|
|
81
|
-
const event = await this.#stt.recognize(ev.frames);
|
|
96
|
+
const event = await this.#stt.recognize(ev.frames, this.abortSignal);
|
|
82
97
|
if (!event.alternatives[0].text) {
|
|
83
98
|
continue;
|
|
84
99
|
}
|
|
@@ -97,7 +112,7 @@ class StreamAdapterWrapper extends import_stt.SpeechStream {
|
|
|
97
112
|
}
|
|
98
113
|
}
|
|
99
114
|
};
|
|
100
|
-
Promise.all([forwardInput(), recognize()]);
|
|
115
|
+
await Promise.all([forwardInput(), recognize()]);
|
|
101
116
|
}
|
|
102
117
|
}
|
|
103
118
|
// Annotate the CommonJS export names for ESM import in node:
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/stt/stream_adapter.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport type { VAD, VADStream } from '../vad.js';\nimport { VADEventType } from '../vad.js';\nimport type { SpeechEvent } from './stt.js';\nimport { STT, SpeechEventType, SpeechStream } from './stt.js';\n\nexport class StreamAdapter extends STT {\n #stt: STT;\n #vad: VAD;\n label: string;\n\n constructor(stt: STT, vad: VAD) {\n super({ streaming: true, interimResults: false });\n this.#stt = stt;\n this.#vad = vad;\n this.label = `stt.StreamAdapter<${this.#stt.label}>`;\n\n this.#stt.on('metrics_collected', (metrics) => {\n this.emit('metrics_collected', metrics);\n });\n }\n\n _recognize(frame: AudioFrame): Promise<SpeechEvent> {\n return this.#stt.recognize(frame);\n }\n\n stream(): StreamAdapterWrapper {\n return new StreamAdapterWrapper(this.#stt, this.#vad);\n }\n}\n\nexport class StreamAdapterWrapper extends SpeechStream {\n #stt: STT;\n #vadStream: VADStream;\n label: string;\n\n constructor(stt: STT, vad: VAD) {\n super(stt);\n this.#stt = stt;\n this.#vadStream = vad.stream();\n this.label = `stt.StreamAdapterWrapper<${this.#stt.label}>`;\n }\n\n async monitorMetrics() {\n return; // do nothing\n }\n\n protected async run() {\n const forwardInput = async () => {\n for await (const input of this.input) {\n if (input === SpeechStream.FLUSH_SENTINEL) {\n this.#vadStream.flush();\n } else {\n this.#vadStream.pushFrame(input);\n }\n }\n this.#vadStream.endInput();\n };\n\n const recognize = async () => {\n for await (const ev of this.#vadStream) {\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.output.put({ type: SpeechEventType.START_OF_SPEECH });\n break;\n case VADEventType.END_OF_SPEECH:\n this.output.put({ type: SpeechEventType.END_OF_SPEECH });\n\n try {\n const event = await this.#stt.recognize(ev.frames);\n if (!event.alternatives![0].text) {\n continue;\n }\n\n this.output.put(event);\n break;\n } catch (error) {\n let logger = log();\n if (error instanceof Error) {\n logger = logger.child({ error: error.message });\n } else {\n logger = logger.child({ error });\n }\n logger.error(`${this.label}: provider recognize task failed`);\n continue;\n }\n }\n }\n };\n\n Promise.all([forwardInput(), recognize()]);\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,iBAAoB;AAEpB,iBAA6B;AAE7B,iBAAmD;AAE5C,MAAM,sBAAsB,eAAI;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,KAAU;AAC9B,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,CAAC;AAChD,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,QAAQ,qBAAqB,KAAK,KAAK,KAAK;AAEjD,SAAK,KAAK,GAAG,qBAAqB,CAAC,YAAY;AAC7C,WAAK,KAAK,qBAAqB,OAAO;AAAA,IACxC,CAAC;AAAA,EACH;AAAA,EAEA,WAAW,
|
|
1
|
+
{"version":3,"sources":["../../src/stt/stream_adapter.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport type { APIConnectOptions } from '../types.js';\nimport { isStreamClosedError } from '../utils.js';\nimport type { VAD, VADStream } from '../vad.js';\nimport { VADEventType } from '../vad.js';\nimport type { SpeechEvent } from './stt.js';\nimport { STT, SpeechEventType, SpeechStream } from './stt.js';\n\nexport class StreamAdapter extends STT {\n #stt: STT;\n #vad: VAD;\n label: string;\n\n constructor(stt: STT, vad: VAD) {\n super({ streaming: true, interimResults: false });\n this.#stt = stt;\n this.#vad = vad;\n this.label = `stt.StreamAdapter<${this.#stt.label}>`;\n\n this.#stt.on('metrics_collected', (metrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n this.#stt.on('error', (error) => {\n this.emit('error', error);\n });\n }\n\n _recognize(frame: AudioFrame, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n return this.#stt.recognize(frame, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): StreamAdapterWrapper {\n return new StreamAdapterWrapper(this.#stt, this.#vad, options?.connOptions);\n }\n}\n\nexport class StreamAdapterWrapper extends SpeechStream {\n #stt: STT;\n #vadStream: VADStream;\n label: string;\n\n constructor(stt: STT, vad: VAD, connOptions?: APIConnectOptions) {\n super(stt, undefined, connOptions);\n this.#stt = stt;\n this.#vadStream = vad.stream();\n this.label = `stt.StreamAdapterWrapper<${this.#stt.label}>`;\n }\n\n close() {\n super.close();\n this.#vadStream.close();\n }\n\n async monitorMetrics() {\n return; // do nothing\n }\n\n protected async run() {\n const forwardInput = async () => {\n for await (const input of this.input) {\n if (input === SpeechStream.FLUSH_SENTINEL) {\n this.#vadStream.flush();\n } else {\n this.#vadStream.pushFrame(input);\n }\n }\n\n // Guard against calling endInput() on already-closed stream\n // This happens during handover when close() is called while forwardInput is running\n try {\n this.#vadStream.endInput();\n } catch (e) {\n if (isStreamClosedError(e)) {\n return;\n }\n throw e;\n }\n };\n\n const recognize = async () => {\n for await (const ev of this.#vadStream) {\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.output.put({ type: SpeechEventType.START_OF_SPEECH });\n break;\n case VADEventType.END_OF_SPEECH:\n this.output.put({ type: SpeechEventType.END_OF_SPEECH });\n\n try {\n const event = await this.#stt.recognize(ev.frames, this.abortSignal);\n if (!event.alternatives![0].text) {\n continue;\n }\n\n this.output.put(event);\n break;\n } catch (error) {\n let logger = log();\n if (error instanceof Error) {\n logger = logger.child({ error: error.message });\n } else {\n logger = logger.child({ error });\n }\n logger.error(`${this.label}: provider recognize task failed`);\n continue;\n }\n }\n }\n };\n\n await Promise.all([forwardInput(), recognize()]);\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,iBAAoB;AAEpB,mBAAoC;AAEpC,iBAA6B;AAE7B,iBAAmD;AAE5C,MAAM,sBAAsB,eAAI;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,KAAU;AAC9B,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,CAAC;AAChD,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,QAAQ,qBAAqB,KAAK,KAAK,KAAK;AAEjD,SAAK,KAAK,GAAG,qBAAqB,CAAC,YAAY;AAC7C,WAAK,KAAK,qBAAqB,OAAO;AAAA,IACxC,CAAC;AAED,SAAK,KAAK,GAAG,SAAS,CAAC,UAAU;AAC/B,WAAK,KAAK,SAAS,KAAK;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA,EAEA,WAAW,OAAmB,aAAiD;AAC7E,WAAO,KAAK,KAAK,UAAU,OAAO,WAAW;AAAA,EAC/C;AAAA,EAEA,OAAO,SAAqE;AAC1E,WAAO,IAAI,qBAAqB,KAAK,MAAM,KAAK,MAAM,mCAAS,WAAW;AAAA,EAC5E;AACF;AAEO,MAAM,6BAA6B,wBAAa;AAAA,EACrD;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,KAAU,aAAiC;AAC/D,UAAM,KAAK,QAAW,WAAW;AACjC,SAAK,OAAO;AACZ,SAAK,aAAa,IAAI,OAAO;AAC7B,SAAK,QAAQ,4BAA4B,KAAK,KAAK,KAAK;AAAA,EAC1D;AAAA,EAEA,QAAQ;AACN,UAAM,MAAM;AACZ,SAAK,WAAW,MAAM;AAAA,EACxB;AAAA,EAEA,MAAM,iBAAiB;AACrB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,eAAe,YAAY;AAC/B,uBAAiB,SAAS,KAAK,OAAO;AACpC,YAAI,UAAU,wBAAa,gBAAgB;AACzC,eAAK,WAAW,MAAM;AAAA,QACxB,OAAO;AACL,eAAK,WAAW,UAAU,KAAK;AAAA,QACjC;AAAA,MACF;AAIA,UAAI;AACF,aAAK,WAAW,SAAS;AAAA,MAC3B,SAAS,GAAG;AACV,gBAAI,kCAAoB,CAAC,GAAG;AAC1B;AAAA,QACF;AACA,cAAM;AAAA,MACR;AAAA,IACF;AAEA,UAAM,YAAY,YAAY;AAC5B,uBAAiB,MAAM,KAAK,YAAY;AACtC,gBAAQ,GAAG,MAAM;AAAA,UACf,KAAK,wBAAa;AAChB,iBAAK,OAAO,IAAI,EAAE,MAAM,2BAAgB,gBAAgB,CAAC;AACzD;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,OAAO,IAAI,EAAE,MAAM,2BAAgB,cAAc,CAAC;AAEvD,gBAAI;AACF,oBAAM,QAAQ,MAAM,KAAK,KAAK,UAAU,GAAG,QAAQ,KAAK,WAAW;AACnE,kBAAI,CAAC,MAAM,aAAc,CAAC,EAAE,MAAM;AAChC;AAAA,cACF;AAEA,mBAAK,OAAO,IAAI,KAAK;AACrB;AAAA,YACF,SAAS,OAAO;AACd,kBAAI,aAAS,gBAAI;AACjB,kBAAI,iBAAiB,OAAO;AAC1B,yBAAS,OAAO,MAAM,EAAE,OAAO,MAAM,QAAQ,CAAC;AAAA,cAChD,OAAO;AACL,yBAAS,OAAO,MAAM,EAAE,MAAM,CAAC;AAAA,cACjC;AACA,qBAAO,MAAM,GAAG,KAAK,KAAK,kCAAkC;AAC5D;AAAA,YACF;AAAA,QACJ;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,IAAI,CAAC,aAAa,GAAG,UAAU,CAAC,CAAC;AAAA,EACjD;AACF;","names":[]}
|
|
@@ -1,4 +1,5 @@
|
|
|
1
1
|
import type { AudioFrame } from '@livekit/rtc-node';
|
|
2
|
+
import type { APIConnectOptions } from '../types.js';
|
|
2
3
|
import type { VAD } from '../vad.js';
|
|
3
4
|
import type { SpeechEvent } from './stt.js';
|
|
4
5
|
import { STT, SpeechStream } from './stt.js';
|
|
@@ -6,13 +7,16 @@ export declare class StreamAdapter extends STT {
|
|
|
6
7
|
#private;
|
|
7
8
|
label: string;
|
|
8
9
|
constructor(stt: STT, vad: VAD);
|
|
9
|
-
_recognize(frame: AudioFrame): Promise<SpeechEvent>;
|
|
10
|
-
stream(
|
|
10
|
+
_recognize(frame: AudioFrame, abortSignal?: AbortSignal): Promise<SpeechEvent>;
|
|
11
|
+
stream(options?: {
|
|
12
|
+
connOptions?: APIConnectOptions;
|
|
13
|
+
}): StreamAdapterWrapper;
|
|
11
14
|
}
|
|
12
15
|
export declare class StreamAdapterWrapper extends SpeechStream {
|
|
13
16
|
#private;
|
|
14
17
|
label: string;
|
|
15
|
-
constructor(stt: STT, vad: VAD);
|
|
18
|
+
constructor(stt: STT, vad: VAD, connOptions?: APIConnectOptions);
|
|
19
|
+
close(): void;
|
|
16
20
|
monitorMetrics(): Promise<void>;
|
|
17
21
|
protected run(): Promise<void>;
|
|
18
22
|
}
|
|
@@ -1,4 +1,5 @@
|
|
|
1
1
|
import type { AudioFrame } from '@livekit/rtc-node';
|
|
2
|
+
import type { APIConnectOptions } from '../types.js';
|
|
2
3
|
import type { VAD } from '../vad.js';
|
|
3
4
|
import type { SpeechEvent } from './stt.js';
|
|
4
5
|
import { STT, SpeechStream } from './stt.js';
|
|
@@ -6,13 +7,16 @@ export declare class StreamAdapter extends STT {
|
|
|
6
7
|
#private;
|
|
7
8
|
label: string;
|
|
8
9
|
constructor(stt: STT, vad: VAD);
|
|
9
|
-
_recognize(frame: AudioFrame): Promise<SpeechEvent>;
|
|
10
|
-
stream(
|
|
10
|
+
_recognize(frame: AudioFrame, abortSignal?: AbortSignal): Promise<SpeechEvent>;
|
|
11
|
+
stream(options?: {
|
|
12
|
+
connOptions?: APIConnectOptions;
|
|
13
|
+
}): StreamAdapterWrapper;
|
|
11
14
|
}
|
|
12
15
|
export declare class StreamAdapterWrapper extends SpeechStream {
|
|
13
16
|
#private;
|
|
14
17
|
label: string;
|
|
15
|
-
constructor(stt: STT, vad: VAD);
|
|
18
|
+
constructor(stt: STT, vad: VAD, connOptions?: APIConnectOptions);
|
|
19
|
+
close(): void;
|
|
16
20
|
monitorMetrics(): Promise<void>;
|
|
17
21
|
protected run(): Promise<void>;
|
|
18
22
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"stream_adapter.d.ts","sourceRoot":"","sources":["../../src/stt/stream_adapter.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAEpD,OAAO,KAAK,EAAE,GAAG,EAAa,MAAM,WAAW,CAAC;AAEhD,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAC5C,OAAO,EAAE,GAAG,EAAmB,YAAY,EAAE,MAAM,UAAU,CAAC;AAE9D,qBAAa,aAAc,SAAQ,GAAG;;IAGpC,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG;
|
|
1
|
+
{"version":3,"file":"stream_adapter.d.ts","sourceRoot":"","sources":["../../src/stt/stream_adapter.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAEpD,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAErD,OAAO,KAAK,EAAE,GAAG,EAAa,MAAM,WAAW,CAAC;AAEhD,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAC5C,OAAO,EAAE,GAAG,EAAmB,YAAY,EAAE,MAAM,UAAU,CAAC;AAE9D,qBAAa,aAAc,SAAQ,GAAG;;IAGpC,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG;IAe9B,UAAU,CAAC,KAAK,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAI9E,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,oBAAoB;CAG5E;AAED,qBAAa,oBAAqB,SAAQ,YAAY;;IAGpD,KAAK,EAAE,MAAM,CAAC;gBAEF,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,WAAW,CAAC,EAAE,iBAAiB;IAO/D,KAAK;IAKC,cAAc;cAIJ,GAAG;CAuDpB"}
|
|
@@ -1,4 +1,5 @@
|
|
|
1
1
|
import { log } from "../log.js";
|
|
2
|
+
import { isStreamClosedError } from "../utils.js";
|
|
2
3
|
import { VADEventType } from "../vad.js";
|
|
3
4
|
import { STT, SpeechEventType, SpeechStream } from "./stt.js";
|
|
4
5
|
class StreamAdapter extends STT {
|
|
@@ -13,24 +14,31 @@ class StreamAdapter extends STT {
|
|
|
13
14
|
this.#stt.on("metrics_collected", (metrics) => {
|
|
14
15
|
this.emit("metrics_collected", metrics);
|
|
15
16
|
});
|
|
17
|
+
this.#stt.on("error", (error) => {
|
|
18
|
+
this.emit("error", error);
|
|
19
|
+
});
|
|
16
20
|
}
|
|
17
|
-
_recognize(frame) {
|
|
18
|
-
return this.#stt.recognize(frame);
|
|
21
|
+
_recognize(frame, abortSignal) {
|
|
22
|
+
return this.#stt.recognize(frame, abortSignal);
|
|
19
23
|
}
|
|
20
|
-
stream() {
|
|
21
|
-
return new StreamAdapterWrapper(this.#stt, this.#vad);
|
|
24
|
+
stream(options) {
|
|
25
|
+
return new StreamAdapterWrapper(this.#stt, this.#vad, options == null ? void 0 : options.connOptions);
|
|
22
26
|
}
|
|
23
27
|
}
|
|
24
28
|
class StreamAdapterWrapper extends SpeechStream {
|
|
25
29
|
#stt;
|
|
26
30
|
#vadStream;
|
|
27
31
|
label;
|
|
28
|
-
constructor(stt, vad) {
|
|
29
|
-
super(stt);
|
|
32
|
+
constructor(stt, vad, connOptions) {
|
|
33
|
+
super(stt, void 0, connOptions);
|
|
30
34
|
this.#stt = stt;
|
|
31
35
|
this.#vadStream = vad.stream();
|
|
32
36
|
this.label = `stt.StreamAdapterWrapper<${this.#stt.label}>`;
|
|
33
37
|
}
|
|
38
|
+
close() {
|
|
39
|
+
super.close();
|
|
40
|
+
this.#vadStream.close();
|
|
41
|
+
}
|
|
34
42
|
async monitorMetrics() {
|
|
35
43
|
return;
|
|
36
44
|
}
|
|
@@ -43,7 +51,14 @@ class StreamAdapterWrapper extends SpeechStream {
|
|
|
43
51
|
this.#vadStream.pushFrame(input);
|
|
44
52
|
}
|
|
45
53
|
}
|
|
46
|
-
|
|
54
|
+
try {
|
|
55
|
+
this.#vadStream.endInput();
|
|
56
|
+
} catch (e) {
|
|
57
|
+
if (isStreamClosedError(e)) {
|
|
58
|
+
return;
|
|
59
|
+
}
|
|
60
|
+
throw e;
|
|
61
|
+
}
|
|
47
62
|
};
|
|
48
63
|
const recognize = async () => {
|
|
49
64
|
for await (const ev of this.#vadStream) {
|
|
@@ -54,7 +69,7 @@ class StreamAdapterWrapper extends SpeechStream {
|
|
|
54
69
|
case VADEventType.END_OF_SPEECH:
|
|
55
70
|
this.output.put({ type: SpeechEventType.END_OF_SPEECH });
|
|
56
71
|
try {
|
|
57
|
-
const event = await this.#stt.recognize(ev.frames);
|
|
72
|
+
const event = await this.#stt.recognize(ev.frames, this.abortSignal);
|
|
58
73
|
if (!event.alternatives[0].text) {
|
|
59
74
|
continue;
|
|
60
75
|
}
|
|
@@ -73,7 +88,7 @@ class StreamAdapterWrapper extends SpeechStream {
|
|
|
73
88
|
}
|
|
74
89
|
}
|
|
75
90
|
};
|
|
76
|
-
Promise.all([forwardInput(), recognize()]);
|
|
91
|
+
await Promise.all([forwardInput(), recognize()]);
|
|
77
92
|
}
|
|
78
93
|
}
|
|
79
94
|
export {
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/stt/stream_adapter.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport type { VAD, VADStream } from '../vad.js';\nimport { VADEventType } from '../vad.js';\nimport type { SpeechEvent } from './stt.js';\nimport { STT, SpeechEventType, SpeechStream } from './stt.js';\n\nexport class StreamAdapter extends STT {\n #stt: STT;\n #vad: VAD;\n label: string;\n\n constructor(stt: STT, vad: VAD) {\n super({ streaming: true, interimResults: false });\n this.#stt = stt;\n this.#vad = vad;\n this.label = `stt.StreamAdapter<${this.#stt.label}>`;\n\n this.#stt.on('metrics_collected', (metrics) => {\n this.emit('metrics_collected', metrics);\n });\n }\n\n _recognize(frame: AudioFrame): Promise<SpeechEvent> {\n return this.#stt.recognize(frame);\n }\n\n stream(): StreamAdapterWrapper {\n return new StreamAdapterWrapper(this.#stt, this.#vad);\n }\n}\n\nexport class StreamAdapterWrapper extends SpeechStream {\n #stt: STT;\n #vadStream: VADStream;\n label: string;\n\n constructor(stt: STT, vad: VAD) {\n super(stt);\n this.#stt = stt;\n this.#vadStream = vad.stream();\n this.label = `stt.StreamAdapterWrapper<${this.#stt.label}>`;\n }\n\n async monitorMetrics() {\n return; // do nothing\n }\n\n protected async run() {\n const forwardInput = async () => {\n for await (const input of this.input) {\n if (input === SpeechStream.FLUSH_SENTINEL) {\n this.#vadStream.flush();\n } else {\n this.#vadStream.pushFrame(input);\n }\n }\n this.#vadStream.endInput();\n };\n\n const recognize = async () => {\n for await (const ev of this.#vadStream) {\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.output.put({ type: SpeechEventType.START_OF_SPEECH });\n break;\n case VADEventType.END_OF_SPEECH:\n this.output.put({ type: SpeechEventType.END_OF_SPEECH });\n\n try {\n const event = await this.#stt.recognize(ev.frames);\n if (!event.alternatives![0].text) {\n continue;\n }\n\n this.output.put(event);\n break;\n } catch (error) {\n let logger = log();\n if (error instanceof Error) {\n logger = logger.child({ error: error.message });\n } else {\n logger = logger.child({ error });\n }\n logger.error(`${this.label}: provider recognize task failed`);\n continue;\n }\n }\n }\n };\n\n Promise.all([forwardInput(), recognize()]);\n }\n}\n"],"mappings":"AAIA,SAAS,WAAW;AAEpB,SAAS,oBAAoB;AAE7B,SAAS,KAAK,iBAAiB,oBAAoB;AAE5C,MAAM,sBAAsB,IAAI;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,KAAU;AAC9B,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,CAAC;AAChD,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,QAAQ,qBAAqB,KAAK,KAAK,KAAK;AAEjD,SAAK,KAAK,GAAG,qBAAqB,CAAC,YAAY;AAC7C,WAAK,KAAK,qBAAqB,OAAO;AAAA,IACxC,CAAC;AAAA,EACH;AAAA,EAEA,WAAW,
|
|
1
|
+
{"version":3,"sources":["../../src/stt/stream_adapter.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport type { APIConnectOptions } from '../types.js';\nimport { isStreamClosedError } from '../utils.js';\nimport type { VAD, VADStream } from '../vad.js';\nimport { VADEventType } from '../vad.js';\nimport type { SpeechEvent } from './stt.js';\nimport { STT, SpeechEventType, SpeechStream } from './stt.js';\n\nexport class StreamAdapter extends STT {\n #stt: STT;\n #vad: VAD;\n label: string;\n\n constructor(stt: STT, vad: VAD) {\n super({ streaming: true, interimResults: false });\n this.#stt = stt;\n this.#vad = vad;\n this.label = `stt.StreamAdapter<${this.#stt.label}>`;\n\n this.#stt.on('metrics_collected', (metrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n this.#stt.on('error', (error) => {\n this.emit('error', error);\n });\n }\n\n _recognize(frame: AudioFrame, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n return this.#stt.recognize(frame, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): StreamAdapterWrapper {\n return new StreamAdapterWrapper(this.#stt, this.#vad, options?.connOptions);\n }\n}\n\nexport class StreamAdapterWrapper extends SpeechStream {\n #stt: STT;\n #vadStream: VADStream;\n label: string;\n\n constructor(stt: STT, vad: VAD, connOptions?: APIConnectOptions) {\n super(stt, undefined, connOptions);\n this.#stt = stt;\n this.#vadStream = vad.stream();\n this.label = `stt.StreamAdapterWrapper<${this.#stt.label}>`;\n }\n\n close() {\n super.close();\n this.#vadStream.close();\n }\n\n async monitorMetrics() {\n return; // do nothing\n }\n\n protected async run() {\n const forwardInput = async () => {\n for await (const input of this.input) {\n if (input === SpeechStream.FLUSH_SENTINEL) {\n this.#vadStream.flush();\n } else {\n this.#vadStream.pushFrame(input);\n }\n }\n\n // Guard against calling endInput() on already-closed stream\n // This happens during handover when close() is called while forwardInput is running\n try {\n this.#vadStream.endInput();\n } catch (e) {\n if (isStreamClosedError(e)) {\n return;\n }\n throw e;\n }\n };\n\n const recognize = async () => {\n for await (const ev of this.#vadStream) {\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.output.put({ type: SpeechEventType.START_OF_SPEECH });\n break;\n case VADEventType.END_OF_SPEECH:\n this.output.put({ type: SpeechEventType.END_OF_SPEECH });\n\n try {\n const event = await this.#stt.recognize(ev.frames, this.abortSignal);\n if (!event.alternatives![0].text) {\n continue;\n }\n\n this.output.put(event);\n break;\n } catch (error) {\n let logger = log();\n if (error instanceof Error) {\n logger = logger.child({ error: error.message });\n } else {\n logger = logger.child({ error });\n }\n logger.error(`${this.label}: provider recognize task failed`);\n continue;\n }\n }\n }\n };\n\n await Promise.all([forwardInput(), recognize()]);\n }\n}\n"],"mappings":"AAIA,SAAS,WAAW;AAEpB,SAAS,2BAA2B;AAEpC,SAAS,oBAAoB;AAE7B,SAAS,KAAK,iBAAiB,oBAAoB;AAE5C,MAAM,sBAAsB,IAAI;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,KAAU;AAC9B,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,CAAC;AAChD,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,QAAQ,qBAAqB,KAAK,KAAK,KAAK;AAEjD,SAAK,KAAK,GAAG,qBAAqB,CAAC,YAAY;AAC7C,WAAK,KAAK,qBAAqB,OAAO;AAAA,IACxC,CAAC;AAED,SAAK,KAAK,GAAG,SAAS,CAAC,UAAU;AAC/B,WAAK,KAAK,SAAS,KAAK;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA,EAEA,WAAW,OAAmB,aAAiD;AAC7E,WAAO,KAAK,KAAK,UAAU,OAAO,WAAW;AAAA,EAC/C;AAAA,EAEA,OAAO,SAAqE;AAC1E,WAAO,IAAI,qBAAqB,KAAK,MAAM,KAAK,MAAM,mCAAS,WAAW;AAAA,EAC5E;AACF;AAEO,MAAM,6BAA6B,aAAa;AAAA,EACrD;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,KAAU,aAAiC;AAC/D,UAAM,KAAK,QAAW,WAAW;AACjC,SAAK,OAAO;AACZ,SAAK,aAAa,IAAI,OAAO;AAC7B,SAAK,QAAQ,4BAA4B,KAAK,KAAK,KAAK;AAAA,EAC1D;AAAA,EAEA,QAAQ;AACN,UAAM,MAAM;AACZ,SAAK,WAAW,MAAM;AAAA,EACxB;AAAA,EAEA,MAAM,iBAAiB;AACrB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,eAAe,YAAY;AAC/B,uBAAiB,SAAS,KAAK,OAAO;AACpC,YAAI,UAAU,aAAa,gBAAgB;AACzC,eAAK,WAAW,MAAM;AAAA,QACxB,OAAO;AACL,eAAK,WAAW,UAAU,KAAK;AAAA,QACjC;AAAA,MACF;AAIA,UAAI;AACF,aAAK,WAAW,SAAS;AAAA,MAC3B,SAAS,GAAG;AACV,YAAI,oBAAoB,CAAC,GAAG;AAC1B;AAAA,QACF;AACA,cAAM;AAAA,MACR;AAAA,IACF;AAEA,UAAM,YAAY,YAAY;AAC5B,uBAAiB,MAAM,KAAK,YAAY;AACtC,gBAAQ,GAAG,MAAM;AAAA,UACf,KAAK,aAAa;AAChB,iBAAK,OAAO,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AACzD;AAAA,UACF,KAAK,aAAa;AAChB,iBAAK,OAAO,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAEvD,gBAAI;AACF,oBAAM,QAAQ,MAAM,KAAK,KAAK,UAAU,GAAG,QAAQ,KAAK,WAAW;AACnE,kBAAI,CAAC,MAAM,aAAc,CAAC,EAAE,MAAM;AAChC;AAAA,cACF;AAEA,mBAAK,OAAO,IAAI,KAAK;AACrB;AAAA,YACF,SAAS,OAAO;AACd,kBAAI,SAAS,IAAI;AACjB,kBAAI,iBAAiB,OAAO;AAC1B,yBAAS,OAAO,MAAM,EAAE,OAAO,MAAM,QAAQ,CAAC;AAAA,cAChD,OAAO;AACL,yBAAS,OAAO,MAAM,EAAE,MAAM,CAAC;AAAA,cACjC;AACA,qBAAO,MAAM,GAAG,KAAK,KAAK,kCAAkC;AAC5D;AAAA,YACF;AAAA,QACJ;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,IAAI,CAAC,aAAa,GAAG,UAAU,CAAC,CAAC;AAAA,EACjD;AACF;","names":[]}
|
package/dist/stt/stt.cjs
CHANGED
|
@@ -37,6 +37,7 @@ var SpeechEventType = /* @__PURE__ */ ((SpeechEventType2) => {
|
|
|
37
37
|
SpeechEventType2[SpeechEventType2["FINAL_TRANSCRIPT"] = 2] = "FINAL_TRANSCRIPT";
|
|
38
38
|
SpeechEventType2[SpeechEventType2["END_OF_SPEECH"] = 3] = "END_OF_SPEECH";
|
|
39
39
|
SpeechEventType2[SpeechEventType2["RECOGNITION_USAGE"] = 4] = "RECOGNITION_USAGE";
|
|
40
|
+
SpeechEventType2[SpeechEventType2["PREFLIGHT_TRANSCRIPT"] = 5] = "PREFLIGHT_TRANSCRIPT";
|
|
40
41
|
return SpeechEventType2;
|
|
41
42
|
})(SpeechEventType || {});
|
|
42
43
|
class STT extends import_node_events.EventEmitter {
|
|
@@ -49,22 +50,51 @@ class STT extends import_node_events.EventEmitter {
|
|
|
49
50
|
get capabilities() {
|
|
50
51
|
return this.#capabilities;
|
|
51
52
|
}
|
|
53
|
+
/**
|
|
54
|
+
* Get the model name/identifier for this STT instance.
|
|
55
|
+
*
|
|
56
|
+
* @returns The model name if available, "unknown" otherwise.
|
|
57
|
+
*
|
|
58
|
+
* @remarks
|
|
59
|
+
* Plugins should override this property to provide their model information.
|
|
60
|
+
*/
|
|
61
|
+
get model() {
|
|
62
|
+
return "unknown";
|
|
63
|
+
}
|
|
64
|
+
/**
|
|
65
|
+
* Get the provider name for this STT instance.
|
|
66
|
+
*
|
|
67
|
+
* @returns The provider name if available, "unknown" otherwise.
|
|
68
|
+
*
|
|
69
|
+
* @remarks
|
|
70
|
+
* Plugins should override this property to provide their provider information.
|
|
71
|
+
*/
|
|
72
|
+
get provider() {
|
|
73
|
+
return "unknown";
|
|
74
|
+
}
|
|
52
75
|
/** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
|
|
53
|
-
async recognize(frame) {
|
|
76
|
+
async recognize(frame, abortSignal) {
|
|
54
77
|
const startTime = process.hrtime.bigint();
|
|
55
|
-
const event = await this._recognize(frame);
|
|
56
|
-
const
|
|
78
|
+
const event = await this._recognize(frame, abortSignal);
|
|
79
|
+
const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1e6));
|
|
57
80
|
this.emit("metrics_collected", {
|
|
58
81
|
type: "stt_metrics",
|
|
59
82
|
requestId: event.requestId ?? "",
|
|
60
83
|
timestamp: Date.now(),
|
|
61
|
-
|
|
84
|
+
durationMs,
|
|
62
85
|
label: this.label,
|
|
63
|
-
|
|
64
|
-
streamed: false
|
|
86
|
+
audioDurationMs: Math.round((0, import_audio.calculateAudioDurationSeconds)(frame) * 1e3),
|
|
87
|
+
streamed: false,
|
|
88
|
+
metadata: {
|
|
89
|
+
modelProvider: this.provider,
|
|
90
|
+
modelName: this.model
|
|
91
|
+
}
|
|
65
92
|
});
|
|
66
93
|
return event;
|
|
67
94
|
}
|
|
95
|
+
async close() {
|
|
96
|
+
return;
|
|
97
|
+
}
|
|
68
98
|
}
|
|
69
99
|
class SpeechStream {
|
|
70
100
|
static FLUSH_SENTINEL = Symbol("FLUSH_SENTINEL");
|
|
@@ -78,6 +108,8 @@ class SpeechStream {
|
|
|
78
108
|
deferredInputStream;
|
|
79
109
|
logger = (0, import_log.log)();
|
|
80
110
|
_connOptions;
|
|
111
|
+
_startTimeOffset = 0;
|
|
112
|
+
abortController = new AbortController();
|
|
81
113
|
constructor(stt, sampleRate, connectionOptions = import_types.DEFAULT_API_CONNECT_OPTIONS) {
|
|
82
114
|
this.#stt = stt;
|
|
83
115
|
this._connOptions = connectionOptions;
|
|
@@ -85,7 +117,7 @@ class SpeechStream {
|
|
|
85
117
|
this.neededSampleRate = sampleRate;
|
|
86
118
|
this.monitorMetrics();
|
|
87
119
|
this.pumpInput();
|
|
88
|
-
(0, import_utils.startSoon)(() => this.mainTask().
|
|
120
|
+
(0, import_utils.startSoon)(() => this.mainTask().finally(() => this.queue.close()));
|
|
89
121
|
}
|
|
90
122
|
async mainTask() {
|
|
91
123
|
for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {
|
|
@@ -93,7 +125,7 @@ class SpeechStream {
|
|
|
93
125
|
return await this.run();
|
|
94
126
|
} catch (error) {
|
|
95
127
|
if (error instanceof import_exceptions.APIError) {
|
|
96
|
-
const retryInterval = this._connOptions
|
|
128
|
+
const retryInterval = (0, import_types.intervalForRetry)(this._connOptions, i);
|
|
97
129
|
if (this._connOptions.maxRetry === 0 || !error.retryable) {
|
|
98
130
|
this.emitError({ error, recoverable: false });
|
|
99
131
|
throw error;
|
|
@@ -104,10 +136,9 @@ class SpeechStream {
|
|
|
104
136
|
options: { retryable: false }
|
|
105
137
|
});
|
|
106
138
|
} else {
|
|
107
|
-
this.emitError({ error, recoverable: true });
|
|
108
139
|
this.logger.warn(
|
|
109
|
-
{
|
|
110
|
-
`failed to recognize speech, retrying in ${retryInterval}
|
|
140
|
+
{ stt: this.#stt.label, attempt: i + 1, error },
|
|
141
|
+
`failed to recognize speech, retrying in ${retryInterval}ms`
|
|
111
142
|
);
|
|
112
143
|
}
|
|
113
144
|
if (retryInterval > 0) {
|
|
@@ -146,20 +177,51 @@ class SpeechStream {
|
|
|
146
177
|
}
|
|
147
178
|
async monitorMetrics() {
|
|
148
179
|
for await (const event of this.queue) {
|
|
149
|
-
this.output.
|
|
180
|
+
if (!this.output.closed) {
|
|
181
|
+
try {
|
|
182
|
+
this.output.put(event);
|
|
183
|
+
} catch (e) {
|
|
184
|
+
if (e instanceof Error && e.message.includes("Queue is closed")) {
|
|
185
|
+
this.logger.warn(
|
|
186
|
+
{ err: e },
|
|
187
|
+
"Queue closed during transcript processing (expected during disconnect)"
|
|
188
|
+
);
|
|
189
|
+
}
|
|
190
|
+
}
|
|
191
|
+
}
|
|
150
192
|
if (event.type !== 4 /* RECOGNITION_USAGE */) continue;
|
|
151
193
|
const metrics = {
|
|
152
194
|
type: "stt_metrics",
|
|
153
195
|
timestamp: Date.now(),
|
|
154
196
|
requestId: event.requestId,
|
|
155
|
-
|
|
197
|
+
durationMs: 0,
|
|
156
198
|
label: this.#stt.label,
|
|
157
|
-
|
|
158
|
-
|
|
199
|
+
audioDurationMs: Math.round(event.recognitionUsage.audioDuration * 1e3),
|
|
200
|
+
inputTokens: event.recognitionUsage.inputTokens ?? 0,
|
|
201
|
+
outputTokens: event.recognitionUsage.outputTokens ?? 0,
|
|
202
|
+
streamed: true,
|
|
203
|
+
metadata: {
|
|
204
|
+
modelProvider: this.#stt.provider,
|
|
205
|
+
modelName: this.#stt.model
|
|
206
|
+
}
|
|
159
207
|
};
|
|
160
208
|
this.#stt.emit("metrics_collected", metrics);
|
|
161
209
|
}
|
|
162
|
-
this.output.
|
|
210
|
+
if (!this.output.closed) {
|
|
211
|
+
this.output.close();
|
|
212
|
+
}
|
|
213
|
+
}
|
|
214
|
+
get abortSignal() {
|
|
215
|
+
return this.abortController.signal;
|
|
216
|
+
}
|
|
217
|
+
get startTimeOffset() {
|
|
218
|
+
return this._startTimeOffset;
|
|
219
|
+
}
|
|
220
|
+
set startTimeOffset(value) {
|
|
221
|
+
if (value < 0) {
|
|
222
|
+
throw new Error("startTimeOffset must be non-negative");
|
|
223
|
+
}
|
|
224
|
+
this._startTimeOffset = value;
|
|
163
225
|
}
|
|
164
226
|
updateInputStream(audioStream) {
|
|
165
227
|
this.deferredInputStream.setSource(audioStream);
|
|
@@ -180,6 +242,10 @@ class SpeechStream {
|
|
|
180
242
|
this.resampler = new import_rtc_node.AudioResampler(frame.sampleRate, this.neededSampleRate);
|
|
181
243
|
}
|
|
182
244
|
}
|
|
245
|
+
if (frame.samplesPerChannel === 0) {
|
|
246
|
+
this.input.put(frame);
|
|
247
|
+
return;
|
|
248
|
+
}
|
|
183
249
|
if (this.resampler) {
|
|
184
250
|
const frames = this.resampler.push(frame);
|
|
185
251
|
for (const frame2 of frames) {
|
|
@@ -214,9 +280,10 @@ class SpeechStream {
|
|
|
214
280
|
}
|
|
215
281
|
/** Close both the input and output of the STT stream */
|
|
216
282
|
close() {
|
|
217
|
-
this.input.close();
|
|
218
|
-
this.queue.close();
|
|
219
|
-
this.output.close();
|
|
283
|
+
if (!this.input.closed) this.input.close();
|
|
284
|
+
if (!this.queue.closed) this.queue.close();
|
|
285
|
+
if (!this.output.closed) this.output.close();
|
|
286
|
+
if (!this.abortController.signal.aborted) this.abortController.abort();
|
|
220
287
|
this.closed = true;
|
|
221
288
|
}
|
|
222
289
|
[Symbol.asyncIterator]() {
|
package/dist/stt/stt.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDuration } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\n const duration = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n duration,\n label: this.label,\n audioDuration: calculateAudioDuration(frame),\n streamed: false,\n });\n return event;\n }\n protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): SpeechStream;\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n this.emitError({ error, recoverable: true });\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n this.output.put(event);\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n duration: 0,\n label: this.#stt.label,\n audioDuration: event.recognitionUsage!.audioDuration,\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n this.output.close();\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n this.input.close();\n this.queue.close();\n this.output.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAEhD,yBAA6B;AAE7B,wBAA6C;AAC7C,mBAAuC;AACvC,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAoE;AAEpE,mBAA8D;AAGvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAtBU,SAAAA;AAAA,GAAA;AA6EL,MAAe,YAAa,gCAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,UAAM,WAAW,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AAC/E,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,mBAAe,qCAAuB,KAAK;AAAA,MAC3C,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAQF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,gCAAoE;AAAA,EAChF,SAAS,IAAI,gCAAgC;AAAA,EAC7C,QAAQ,IAAI,gCAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EACb;AAAA,EAER,YACE,KACA,YACA,oBAAuC,0CACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,gCAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AACL,iBAAK,UAAU,EAAE,OAAO,aAAa,KAAK,CAAC;AAC3C,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,UAAU;AAAA,QACV,OAAO,KAAK,KAAK;AAAA,QACjB,eAAe,MAAM,iBAAkB;AAAA,QACvC,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,SAAK,OAAO,MAAM;AAAA,EACpB;AAAA,EAIA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,+BAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
|
|
1
|
+
{"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport type { LanguageCode } from '../language.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\nimport type { TimedString } from '../voice/index.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n /** Language code of the speech. */\n language: LanguageCode;\n /** Transcribed text. */\n text: string;\n /** Start time of the speech segment in seconds. */\n startTime: number;\n /** End time of the speech segment in seconds. */\n endTime: number;\n /** Confidence score of the transcription (0-1). */\n confidence: number;\n /** Word-level timing information. */\n words?: TimedString[];\n}\n\nexport interface RecognitionUsage {\n /** Duration of the audio that was recognized in seconds. */\n audioDuration: number;\n /** Input audio tokens (for token-based STT billing). */\n inputTokens?: number;\n /** Output text tokens (for token-based STT billing). */\n outputTokens?: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n /**\n * Whether this STT supports aligned transcripts with word/chunk timestamps.\n * - 'word': Provider returns word-level timestamps\n * - 'chunk': Provider returns chunk-level timestamps (e.g., sentence/phrase boundaries)\n * - false: Provider does not support aligned transcripts\n */\n alignedTranscript?: 'word' | 'chunk' | false;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /**\n * Get the model name/identifier for this STT instance.\n *\n * @returns The model name if available, \"unknown\" otherwise.\n *\n * @remarks\n * Plugins should override this property to provide their model information.\n */\n get model(): string {\n return 'unknown';\n }\n\n /**\n * Get the provider name for this STT instance.\n *\n * @returns The provider name if available, \"unknown\" otherwise.\n *\n * @remarks\n * Plugins should override this property to provide their provider information.\n */\n get provider(): string {\n return 'unknown';\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame, abortSignal);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n metadata: {\n modelProvider: this.provider,\n modelName: this.model,\n },\n });\n return event;\n }\n\n protected abstract _recognize(\n frame: AudioBuffer,\n abortSignal?: AbortSignal,\n ): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n private _startTimeOffset: number = 0;\n\n protected abortController = new AbortController();\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { stt: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}ms`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n inputTokens: event.recognitionUsage!.inputTokens ?? 0,\n outputTokens: event.recognitionUsage!.outputTokens ?? 0,\n streamed: true,\n metadata: {\n modelProvider: this.#stt.provider,\n modelName: this.#stt.model,\n },\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n protected get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n get startTimeOffset(): number {\n return this._startTimeOffset;\n }\n\n set startTimeOffset(value: number) {\n if (value < 0) {\n throw new Error('startTimeOffset must be non-negative');\n }\n this._startTimeOffset = value;\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (frame.samplesPerChannel === 0) {\n this.input.put(frame);\n return;\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAEhD,yBAA6B;AAE7B,wBAA6C;AAC7C,mBAA8C;AAE9C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAsF;AAEtF,mBAA8D;AAIvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AAsGL,MAAe,YAAa,gCAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA;AAAA,EAGA,MAAM,UAAU,OAAoB,aAAiD;AACnF,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,OAAO,WAAW;AACtD,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,UAAM,4CAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,MACV,UAAU;AAAA,QACR,eAAe,KAAK;AAAA,QACpB,WAAW,KAAK;AAAA,MAClB;AAAA,IACF,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAeA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,gCAAoE;AAAA,EAChF,SAAS,IAAI,gCAAgC;AAAA,EAC7C,QAAQ,IAAI,gCAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EACb;AAAA,EACA,mBAA2B;AAAA,EAEzB,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,KACA,YACA,oBAAuC,0CACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,gCAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,oBAAgB,+BAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,aAAa,MAAM,iBAAkB,eAAe;AAAA,QACpD,cAAc,MAAM,iBAAkB,gBAAgB;AAAA,QACtD,UAAU;AAAA,QACV,UAAU;AAAA,UACR,eAAe,KAAK,KAAK;AAAA,UACzB,WAAW,KAAK,KAAK;AAAA,QACvB;AAAA,MACF;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,IAAc,cAA2B;AACvC,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,IAAI,kBAA0B;AAC5B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,gBAAgB,OAAe;AACjC,QAAI,QAAQ,GAAG;AACb,YAAM,IAAI,MAAM,sCAAsC;AAAA,IACxD;AACA,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,+BAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,MAAM,sBAAsB,GAAG;AACjC,WAAK,MAAM,IAAI,KAAK;AACpB;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
|