@livekit/agents 1.0.48 → 1.1.0-dev.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/constants.cjs +27 -0
- package/dist/constants.cjs.map +1 -1
- package/dist/constants.d.cts +9 -0
- package/dist/constants.d.ts +9 -0
- package/dist/constants.d.ts.map +1 -1
- package/dist/constants.js +18 -0
- package/dist/constants.js.map +1 -1
- package/dist/inference/api_protos.d.cts +71 -71
- package/dist/inference/api_protos.d.ts +71 -71
- 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 +147 -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 +121 -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 +181 -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 +147 -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 +329 -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 +295 -0
- package/dist/inference/interruption/ws_transport.js.map +1 -0
- package/dist/inference/llm.cjs +14 -10
- package/dist/inference/llm.cjs.map +1 -1
- package/dist/inference/llm.d.cts +2 -1
- package/dist/inference/llm.d.ts +2 -1
- package/dist/inference/llm.d.ts.map +1 -1
- package/dist/inference/llm.js +8 -10
- package/dist/inference/llm.js.map +1 -1
- package/dist/inference/stt.cjs +7 -2
- package/dist/inference/stt.cjs.map +1 -1
- package/dist/inference/stt.d.cts +2 -0
- package/dist/inference/stt.d.ts +2 -0
- package/dist/inference/stt.d.ts.map +1 -1
- package/dist/inference/stt.js +8 -3
- package/dist/inference/stt.js.map +1 -1
- package/dist/inference/tts.cjs +7 -2
- package/dist/inference/tts.cjs.map +1 -1
- package/dist/inference/tts.d.cts +2 -0
- package/dist/inference/tts.d.ts +2 -0
- package/dist/inference/tts.d.ts.map +1 -1
- package/dist/inference/tts.js +8 -3
- package/dist/inference/tts.js.map +1 -1
- package/dist/inference/utils.cjs +26 -7
- package/dist/inference/utils.cjs.map +1 -1
- package/dist/inference/utils.d.cts +13 -0
- package/dist/inference/utils.d.ts +13 -0
- 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/llm/chat_context.cjs +20 -2
- package/dist/llm/chat_context.cjs.map +1 -1
- package/dist/llm/chat_context.d.cts +19 -1
- package/dist/llm/chat_context.d.ts +19 -1
- package/dist/llm/chat_context.d.ts.map +1 -1
- package/dist/llm/chat_context.js +20 -2
- package/dist/llm/chat_context.js.map +1 -1
- package/dist/llm/index.cjs.map +1 -1
- package/dist/llm/index.d.cts +1 -1
- package/dist/llm/index.d.ts +1 -1
- package/dist/llm/index.d.ts.map +1 -1
- package/dist/llm/index.js.map +1 -1
- package/dist/llm/llm.cjs +16 -1
- package/dist/llm/llm.cjs.map +1 -1
- package/dist/llm/llm.d.cts +9 -0
- package/dist/llm/llm.d.ts +9 -0
- package/dist/llm/llm.d.ts.map +1 -1
- package/dist/llm/llm.js +16 -1
- package/dist/llm/llm.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 +1 -0
- package/dist/llm/realtime.d.ts +1 -0
- 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/metrics/base.cjs.map +1 -1
- package/dist/metrics/base.d.cts +45 -1
- package/dist/metrics/base.d.ts +45 -1
- 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 +3 -0
- package/dist/metrics/usage_collector.cjs.map +1 -1
- package/dist/metrics/usage_collector.d.cts +9 -0
- package/dist/metrics/usage_collector.d.ts +9 -0
- package/dist/metrics/usage_collector.d.ts.map +1 -1
- package/dist/metrics/usage_collector.js +3 -0
- package/dist/metrics/usage_collector.js.map +1 -1
- package/dist/metrics/utils.cjs +9 -0
- package/dist/metrics/utils.cjs.map +1 -1
- package/dist/metrics/utils.d.ts.map +1 -1
- package/dist/metrics/utils.js +9 -0
- package/dist/metrics/utils.js.map +1 -1
- package/dist/stream/multi_input_stream.test.cjs +4 -0
- package/dist/stream/multi_input_stream.test.cjs.map +1 -1
- package/dist/stream/multi_input_stream.test.js +5 -1
- package/dist/stream/multi_input_stream.test.js.map +1 -1
- package/dist/stream/stream_channel.cjs +31 -0
- package/dist/stream/stream_channel.cjs.map +1 -1
- package/dist/stream/stream_channel.d.cts +4 -2
- package/dist/stream/stream_channel.d.ts +4 -2
- package/dist/stream/stream_channel.d.ts.map +1 -1
- package/dist/stream/stream_channel.js +31 -0
- package/dist/stream/stream_channel.js.map +1 -1
- package/dist/stt/stt.cjs +34 -2
- package/dist/stt/stt.cjs.map +1 -1
- package/dist/stt/stt.d.cts +22 -0
- package/dist/stt/stt.d.ts +22 -0
- package/dist/stt/stt.d.ts.map +1 -1
- package/dist/stt/stt.js +34 -2
- package/dist/stt/stt.js.map +1 -1
- package/dist/telemetry/otel_http_exporter.cjs +24 -5
- package/dist/telemetry/otel_http_exporter.cjs.map +1 -1
- package/dist/telemetry/otel_http_exporter.d.cts +1 -0
- package/dist/telemetry/otel_http_exporter.d.ts +1 -0
- package/dist/telemetry/otel_http_exporter.d.ts.map +1 -1
- package/dist/telemetry/otel_http_exporter.js +24 -5
- package/dist/telemetry/otel_http_exporter.js.map +1 -1
- package/dist/telemetry/trace_types.cjs +5 -5
- package/dist/telemetry/trace_types.cjs.map +1 -1
- package/dist/telemetry/trace_types.d.cts +9 -5
- package/dist/telemetry/trace_types.d.ts +9 -5
- package/dist/telemetry/trace_types.d.ts.map +1 -1
- package/dist/telemetry/trace_types.js +5 -5
- package/dist/telemetry/trace_types.js.map +1 -1
- package/dist/telemetry/traces.cjs +47 -8
- package/dist/telemetry/traces.cjs.map +1 -1
- package/dist/telemetry/traces.d.ts.map +1 -1
- package/dist/telemetry/traces.js +47 -8
- package/dist/telemetry/traces.js.map +1 -1
- package/dist/tts/tts.cjs +64 -2
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.cts +34 -0
- package/dist/tts/tts.d.ts +34 -0
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +64 -2
- package/dist/tts/tts.js.map +1 -1
- package/dist/version.cjs +1 -1
- package/dist/version.js +1 -1
- package/dist/voice/agent.cjs +25 -4
- package/dist/voice/agent.cjs.map +1 -1
- package/dist/voice/agent.d.cts +10 -2
- package/dist/voice/agent.d.ts +10 -2
- package/dist/voice/agent.d.ts.map +1 -1
- package/dist/voice/agent.js +25 -4
- package/dist/voice/agent.js.map +1 -1
- package/dist/voice/agent_activity.cjs +261 -36
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.cts +20 -6
- package/dist/voice/agent_activity.d.ts +20 -6
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +262 -37
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/agent_session.cjs +105 -48
- package/dist/voice/agent_session.cjs.map +1 -1
- package/dist/voice/agent_session.d.cts +90 -20
- package/dist/voice/agent_session.d.ts +90 -20
- package/dist/voice/agent_session.d.ts.map +1 -1
- package/dist/voice/agent_session.js +105 -46
- package/dist/voice/agent_session.js.map +1 -1
- package/dist/voice/audio_recognition.cjs +287 -6
- package/dist/voice/audio_recognition.cjs.map +1 -1
- package/dist/voice/audio_recognition.d.cts +42 -3
- package/dist/voice/audio_recognition.d.ts +42 -3
- package/dist/voice/audio_recognition.d.ts.map +1 -1
- package/dist/voice/audio_recognition.js +289 -7
- package/dist/voice/audio_recognition.js.map +1 -1
- package/dist/voice/client_events.cjs +554 -0
- package/dist/voice/client_events.cjs.map +1 -0
- package/dist/voice/client_events.d.cts +195 -0
- package/dist/voice/client_events.d.ts +195 -0
- package/dist/voice/client_events.d.ts.map +1 -0
- package/dist/voice/client_events.js +548 -0
- package/dist/voice/client_events.js.map +1 -0
- package/dist/voice/events.cjs +1 -0
- package/dist/voice/events.cjs.map +1 -1
- package/dist/voice/events.d.cts +8 -5
- package/dist/voice/events.d.ts +8 -5
- package/dist/voice/events.d.ts.map +1 -1
- package/dist/voice/events.js +1 -0
- package/dist/voice/events.js.map +1 -1
- package/dist/voice/generation.cjs +43 -8
- package/dist/voice/generation.cjs.map +1 -1
- package/dist/voice/generation.d.cts +3 -3
- package/dist/voice/generation.d.ts +3 -3
- package/dist/voice/generation.d.ts.map +1 -1
- package/dist/voice/generation.js +43 -8
- package/dist/voice/generation.js.map +1 -1
- package/dist/voice/index.cjs +1 -0
- package/dist/voice/index.cjs.map +1 -1
- package/dist/voice/index.d.cts +1 -0
- package/dist/voice/index.d.ts +1 -0
- package/dist/voice/index.d.ts.map +1 -1
- package/dist/voice/index.js +1 -0
- package/dist/voice/index.js.map +1 -1
- package/dist/voice/report.cjs +20 -8
- package/dist/voice/report.cjs.map +1 -1
- package/dist/voice/report.d.cts +5 -0
- package/dist/voice/report.d.ts +5 -0
- package/dist/voice/report.d.ts.map +1 -1
- package/dist/voice/report.js +20 -8
- package/dist/voice/report.js.map +1 -1
- package/dist/voice/report.test.cjs +106 -0
- package/dist/voice/report.test.cjs.map +1 -0
- package/dist/voice/report.test.js +105 -0
- package/dist/voice/report.test.js.map +1 -0
- package/dist/voice/room_io/room_io.cjs +5 -39
- package/dist/voice/room_io/room_io.cjs.map +1 -1
- package/dist/voice/room_io/room_io.d.cts +4 -9
- package/dist/voice/room_io/room_io.d.ts +4 -9
- package/dist/voice/room_io/room_io.d.ts.map +1 -1
- package/dist/voice/room_io/room_io.js +5 -40
- package/dist/voice/room_io/room_io.js.map +1 -1
- 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 +97 -0
- package/dist/voice/turn_config/utils.cjs.map +1 -0
- package/dist/voice/turn_config/utils.d.cts +25 -0
- package/dist/voice/turn_config/utils.d.ts +25 -0
- package/dist/voice/turn_config/utils.d.ts.map +1 -0
- package/dist/voice/turn_config/utils.js +73 -0
- package/dist/voice/turn_config/utils.js.map +1 -0
- package/dist/voice/turn_config/utils.test.cjs +86 -0
- package/dist/voice/turn_config/utils.test.cjs.map +1 -0
- package/dist/voice/turn_config/utils.test.js +85 -0
- package/dist/voice/turn_config/utils.test.js.map +1 -0
- package/dist/voice/wire_format.cjs +798 -0
- package/dist/voice/wire_format.cjs.map +1 -0
- package/dist/voice/wire_format.d.cts +5503 -0
- package/dist/voice/wire_format.d.ts +5503 -0
- package/dist/voice/wire_format.d.ts.map +1 -0
- package/dist/voice/wire_format.js +728 -0
- package/dist/voice/wire_format.js.map +1 -0
- package/package.json +2 -1
- package/src/constants.ts +13 -0
- package/src/inference/interruption/defaults.ts +51 -0
- package/src/inference/interruption/errors.ts +25 -0
- package/src/inference/interruption/http_transport.ts +187 -0
- package/src/inference/interruption/interruption_cache_entry.ts +50 -0
- package/src/inference/interruption/interruption_detector.ts +188 -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 +402 -0
- package/src/inference/llm.ts +9 -12
- package/src/inference/stt.ts +10 -3
- package/src/inference/tts.ts +10 -3
- package/src/inference/utils.ts +29 -1
- package/src/llm/chat_context.ts +40 -2
- package/src/llm/index.ts +1 -0
- package/src/llm/llm.ts +16 -0
- package/src/llm/realtime.ts +4 -0
- package/src/metrics/base.ts +48 -1
- package/src/metrics/index.ts +11 -0
- package/src/metrics/model_usage.test.ts +545 -0
- package/src/metrics/model_usage.ts +262 -0
- package/src/metrics/usage_collector.ts +11 -0
- package/src/metrics/utils.ts +11 -0
- package/src/stream/multi_input_stream.test.ts +6 -1
- package/src/stream/stream_channel.ts +34 -2
- package/src/stt/stt.ts +38 -0
- package/src/telemetry/otel_http_exporter.ts +28 -5
- package/src/telemetry/trace_types.ts +11 -8
- package/src/telemetry/traces.ts +111 -54
- package/src/tts/tts.ts +69 -1
- package/src/voice/agent.ts +30 -3
- package/src/voice/agent_activity.ts +327 -28
- package/src/voice/agent_session.ts +207 -59
- package/src/voice/audio_recognition.ts +385 -9
- package/src/voice/client_events.ts +838 -0
- package/src/voice/events.ts +14 -4
- package/src/voice/generation.ts +52 -9
- package/src/voice/index.ts +1 -0
- package/src/voice/report.test.ts +117 -0
- package/src/voice/report.ts +29 -6
- package/src/voice/room_io/room_io.ts +7 -61
- 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 +100 -0
- package/src/voice/turn_config/utils.ts +103 -0
- package/src/voice/wire_format.ts +827 -0
package/dist/tts/tts.js
CHANGED
|
@@ -27,6 +27,28 @@ class TTS extends EventEmitter {
|
|
|
27
27
|
get numChannels() {
|
|
28
28
|
return this.#numChannels;
|
|
29
29
|
}
|
|
30
|
+
/**
|
|
31
|
+
* Get the model name/identifier for this TTS instance.
|
|
32
|
+
*
|
|
33
|
+
* @returns The model name if available, "unknown" otherwise.
|
|
34
|
+
*
|
|
35
|
+
* @remarks
|
|
36
|
+
* Plugins should override this property to provide their model information.
|
|
37
|
+
*/
|
|
38
|
+
get model() {
|
|
39
|
+
return "unknown";
|
|
40
|
+
}
|
|
41
|
+
/**
|
|
42
|
+
* Get the provider name for this TTS instance.
|
|
43
|
+
*
|
|
44
|
+
* @returns The provider name if available, "unknown" otherwise.
|
|
45
|
+
*
|
|
46
|
+
* @remarks
|
|
47
|
+
* Plugins should override this property to provide their provider information.
|
|
48
|
+
*/
|
|
49
|
+
get provider() {
|
|
50
|
+
return "unknown";
|
|
51
|
+
}
|
|
30
52
|
async close() {
|
|
31
53
|
return;
|
|
32
54
|
}
|
|
@@ -47,6 +69,8 @@ class SynthesizeStream {
|
|
|
47
69
|
#metricsText = "";
|
|
48
70
|
#monitorMetricsTask;
|
|
49
71
|
#ttsRequestSpan;
|
|
72
|
+
#inputTokens = 0;
|
|
73
|
+
#outputTokens = 0;
|
|
50
74
|
constructor(tts, connOptions = DEFAULT_API_CONNECT_OPTIONS) {
|
|
51
75
|
this.#tts = tts;
|
|
52
76
|
this.connOptions = connOptions;
|
|
@@ -151,6 +175,17 @@ class SynthesizeStream {
|
|
|
151
175
|
}
|
|
152
176
|
}
|
|
153
177
|
}
|
|
178
|
+
/**
|
|
179
|
+
* Set token usage for token-based TTS billing (e.g., OpenAI TTS).
|
|
180
|
+
* Plugins should call this method to report token usage.
|
|
181
|
+
*/
|
|
182
|
+
setTokenUsage({
|
|
183
|
+
inputTokens = 0,
|
|
184
|
+
outputTokens = 0
|
|
185
|
+
} = {}) {
|
|
186
|
+
this.#inputTokens = inputTokens;
|
|
187
|
+
this.#outputTokens = outputTokens;
|
|
188
|
+
}
|
|
154
189
|
async monitorMetrics() {
|
|
155
190
|
const startTime = process.hrtime.bigint();
|
|
156
191
|
let audioDurationMs = 0;
|
|
@@ -171,12 +206,20 @@ class SynthesizeStream {
|
|
|
171
206
|
audioDurationMs: roundedAudioDurationMs,
|
|
172
207
|
cancelled: this.abortController.signal.aborted,
|
|
173
208
|
label: this.#tts.label,
|
|
174
|
-
|
|
209
|
+
inputTokens: this.#inputTokens,
|
|
210
|
+
outputTokens: this.#outputTokens,
|
|
211
|
+
streamed: true,
|
|
212
|
+
metadata: {
|
|
213
|
+
modelProvider: this.#tts.provider,
|
|
214
|
+
modelName: this.#tts.model
|
|
215
|
+
}
|
|
175
216
|
};
|
|
176
217
|
if (this.#ttsRequestSpan) {
|
|
177
218
|
this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));
|
|
178
219
|
}
|
|
179
220
|
this.#tts.emit("metrics_collected", metrics);
|
|
221
|
+
this.#inputTokens = 0;
|
|
222
|
+
this.#outputTokens = 0;
|
|
180
223
|
}
|
|
181
224
|
};
|
|
182
225
|
for await (const audio of this.queue) {
|
|
@@ -260,6 +303,8 @@ class ChunkedStream {
|
|
|
260
303
|
#ttsRequestSpan;
|
|
261
304
|
_connOptions;
|
|
262
305
|
logger = log();
|
|
306
|
+
#inputTokens = 0;
|
|
307
|
+
#outputTokens = 0;
|
|
263
308
|
abortController = new AbortController();
|
|
264
309
|
constructor(text, tts, connOptions = DEFAULT_API_CONNECT_OPTIONS, abortSignal) {
|
|
265
310
|
this.#text = text;
|
|
@@ -340,6 +385,17 @@ class ChunkedStream {
|
|
|
340
385
|
get abortSignal() {
|
|
341
386
|
return this.abortController.signal;
|
|
342
387
|
}
|
|
388
|
+
/**
|
|
389
|
+
* Set token usage for token-based TTS billing (e.g., OpenAI TTS).
|
|
390
|
+
* Plugins should call this method to report token usage.
|
|
391
|
+
*/
|
|
392
|
+
setTokenUsage({
|
|
393
|
+
inputTokens = 0,
|
|
394
|
+
outputTokens = 0
|
|
395
|
+
} = {}) {
|
|
396
|
+
this.#inputTokens = inputTokens;
|
|
397
|
+
this.#outputTokens = outputTokens;
|
|
398
|
+
}
|
|
343
399
|
async monitorMetrics() {
|
|
344
400
|
const startTime = process.hrtime.bigint();
|
|
345
401
|
let audioDurationMs = 0;
|
|
@@ -366,7 +422,13 @@ class ChunkedStream {
|
|
|
366
422
|
cancelled: false,
|
|
367
423
|
// TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here
|
|
368
424
|
label: this.#tts.label,
|
|
369
|
-
|
|
425
|
+
inputTokens: this.#inputTokens,
|
|
426
|
+
outputTokens: this.#outputTokens,
|
|
427
|
+
streamed: false,
|
|
428
|
+
metadata: {
|
|
429
|
+
modelProvider: this.#tts.provider,
|
|
430
|
+
modelName: this.#tts.model
|
|
431
|
+
}
|
|
370
432
|
};
|
|
371
433
|
if (this.#ttsRequestSpan) {
|
|
372
434
|
this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));
|
package/dist/tts/tts.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/tts/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Span } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { recordException, traceTypes, tracer } from '../telemetry/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\nimport type { TimedString } from '../voice/io.js';\n\n/**\n * SynthesizedAudio is a packet of speech synthesis as returned by the TTS.\n */\nexport interface SynthesizedAudio {\n /** Request ID (one segment could be made up of multiple requests) */\n requestId: string;\n /** Segment ID, each segment is separated by a flush */\n segmentId: string;\n /** Synthesized audio frame */\n frame: AudioFrame;\n /** Current segment of the synthesized audio */\n deltaText?: string;\n /** Whether this is the last frame of the segment (streaming only) */\n final: boolean;\n /**\n * Timed transcripts associated with this audio packet (word-level timestamps).\n */\n timedTranscripts?: TimedString[];\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n streaming: boolean;\n // Whether this TTS supports aligned transcripts (word-level timestamps).\n alignedTranscript?: boolean;\n}\n\nexport interface TTSError {\n type: 'tts_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n ['metrics_collected']: (metrics: TTSMetrics) => void;\n ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech 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 TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n #capabilities: TTSCapabilities;\n #sampleRate: number;\n #numChannels: number;\n abstract label: string;\n\n constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n super();\n this.#capabilities = capabilities;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n }\n\n /** Returns this TTS's capabilities */\n get capabilities(): TTSCapabilities {\n return this.#capabilities;\n }\n\n /** Returns the sample rate of audio frames returned by this TTS */\n get sampleRate(): number {\n return this.#sampleRate;\n }\n\n /** Returns the channel count of audio frames returned by this TTS */\n get numChannels(): number {\n return this.#numChannels;\n }\n\n /**\n * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n */\n abstract synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream;\n\n /**\n * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\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 SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n protected queue = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected output = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected closed = false;\n protected connOptions: APIConnectOptions;\n protected abortController = new AbortController();\n\n private deferredInputStream: DeferredReadableStream<\n string | typeof SynthesizeStream.FLUSH_SENTINEL\n >;\n private logger = log();\n\n abstract label: string;\n\n #tts: TTS;\n #metricsPendingTexts: string[] = [];\n #metricsText = '';\n #monitorMetricsTask?: Promise<void>;\n #ttsRequestSpan?: Span;\n\n constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n this.#tts = tts;\n this.connOptions = connOptions;\n this.deferredInputStream = new DeferredReadableStream();\n this.pumpInput();\n\n this.abortController.signal.addEventListener('abort', () => {\n this.deferredInputStream.detachSource();\n // TODO (AJS-36) clean this up when we refactor with streams\n if (!this.input.closed) this.input.close();\n if (!this.output.closed) this.output.close();\n this.closed = true;\n });\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 _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: true,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this.connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\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 generate TTS completion 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 { tts: this.#tts.label, attempt: i + 1, error },\n `failed to synthesize 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 mainTask = async () =>\n tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n // consuming the stream directly.\n //\n // A full refactor to native Web Streams was considered but is currently deferred.\n // The primary reason is to maintain architectural parity with the Python SDK,\n // which is a key design goal for the project. This ensures a consistent developer\n // experience across both platforms.\n //\n // For more context, see the discussion in GitHub issue # 844.\n protected async pumpInput() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n break;\n }\n this.pushText(value);\n }\n this.endInput();\n } catch (error) {\n this.logger.error(error, 'Error reading deferred input stream');\n } finally {\n reader.releaseLock();\n // Ensure output is closed when the stream ends\n if (!this.#monitorMetricsTask) {\n // No text was received, close the output directly\n this.output.close();\n }\n }\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n const emit = () => {\n if (this.#metricsPendingTexts.length) {\n const text = this.#metricsPendingTexts.shift()!;\n const duration = process.hrtime.bigint() - startTime;\n const roundedAudioDurationMs = Math.round(audioDurationMs);\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: text.length,\n audioDurationMs: roundedAudioDurationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#tts.label,\n streamed: false,\n };\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n }\n this.#tts.emit('metrics_collected', metrics);\n }\n };\n\n for await (const audio of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(audio);\n if (audio === SynthesizeStream.END_OF_STREAM) continue;\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n // TODO(AJS-102): use frame.durationMs once available in rtc-node\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n if (audio.final) {\n emit();\n }\n }\n\n if (requestId) {\n emit();\n }\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(text: ReadableStream<string>) {\n this.deferredInputStream.setSource(text);\n }\n\n /** Push a string of text to the TTS */\n /** @deprecated Use `updateInputStream` instead */\n pushText(text: string) {\n if (!this.#monitorMetricsTask) {\n this.#monitorMetricsTask = this.monitorMetrics();\n // Close output when metrics task completes\n this.#monitorMetricsTask.finally(() => this.output.close());\n }\n this.#metricsText += text;\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(text);\n }\n\n /** Flush the TTS, causing it to process all pending text */\n flush() {\n if (this.#metricsText) {\n this.#metricsPendingTexts.push(this.#metricsText);\n this.#metricsText = '';\n }\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n this.flush();\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n return this.output.next();\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): SynthesizeStream {\n return this;\n }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\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 ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n protected output = new AsyncIterableQueue<SynthesizedAudio>();\n protected closed = false;\n abstract label: string;\n #text: string;\n #tts: TTS;\n #ttsRequestSpan?: Span;\n private _connOptions: APIConnectOptions;\n private logger = log();\n\n protected abortController = new AbortController();\n\n constructor(\n text: string,\n tts: TTS,\n connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n abortSignal?: AbortSignal,\n ) {\n this.#text = text;\n this.#tts = tts;\n this._connOptions = connOptions;\n\n if (abortSignal) {\n abortSignal.addEventListener('abort', () => this.abortController.abort(), { once: true });\n }\n\n this.monitorMetrics();\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 Promise.resolve().then(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: false,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\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 generate TTS completion 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 { tts: this.#tts.label, attempt: i + 1, error },\n `failed to generate TTS completion, 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 async mainTask() {\n return tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n protected abstract run(): Promise<void>;\n\n get inputText(): string {\n return this.#text;\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n for await (const audio of this.queue) {\n this.output.put(audio);\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: this.#text.length,\n audioDurationMs: Math.round(audioDurationMs),\n cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n label: this.#tts.label,\n streamed: false,\n };\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n\n this.#tts.emit('metrics_collected', metrics);\n }\n\n /** Collect every frame into one in a single call */\n async collect(): Promise<AudioFrame> {\n const frames = [];\n for await (const event of this) {\n frames.push(event.frame);\n }\n return mergeFrames(frames);\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n 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](): ChunkedStream {\n return this;\n }\n}\n"],"mappings":"AAMA,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,iBAAiB,YAAY,cAAc;AACpD,SAAiC,6BAA6B,wBAAwB;AACtF,SAAS,oBAAoB,OAAO,aAAa,WAAW,eAAe;AAwDpE,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAkBA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,mBAAoE;AAAA,EAChF,QAAQ,IAAI,mBAEpB;AAAA,EACQ,SAAS,IAAI,mBAErB;AAAA,EACQ,SAAS;AAAA,EACT;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,SAAS,IAAI;AAAA,EAIrB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,cAAiC,6BAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,cAAc;AACnB,SAAK,sBAAsB,IAAI,uBAAuB;AACtD,SAAK,UAAU;AAEf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,UAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,UAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,WAAW,GAAG,KAAK;AACtD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,aAAa,CAAC;AAE1D,cAAI,KAAK,YAAY,aAAa,KAAK,CAAC,MAAM,WAAW;AACvD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,YAAY,UAAU;AAC1C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,YAAY,WAAW,CAAC;AAAA,cACjF,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,4CAA4C,aAAa;AAAA,YAC3D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,WAAW,YACjB,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,IAC/D,MAAM;AAAA,IACN,WAAW;AAAA,EACb,CAAC;AAAA,EAEK,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;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,YAAI,KAAK,iBAAiB;AACxB,eAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AAAA,QACxF;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AAEA,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AAEX,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,mBAAqC;AAAA,EACjD,SAAS,IAAI,mBAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EAEX,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,MACA,KACA,cAAiC,6BACjC,aACA;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,QAAI,aAAa;AACf,kBAAY,iBAAiB,SAAS,MAAM,KAAK,gBAAgB,MAAM,GAAG,EAAE,MAAM,KAAK,CAAC;AAAA,IAC1F;AAEA,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChF;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,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,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,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,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,WAAW;AACvB,WAAO,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,MACtE,MAAM;AAAA,MACN,WAAW;AAAA,IACb,CAAC;AAAA,EACH;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,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AACtF,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAEA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,WAAO,YAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,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,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/tts/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Span } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { recordException, traceTypes, tracer } from '../telemetry/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\nimport type { TimedString } from '../voice/io.js';\n\n/**\n * SynthesizedAudio is a packet of speech synthesis as returned by the TTS.\n */\nexport interface SynthesizedAudio {\n /** Request ID (one segment could be made up of multiple requests) */\n requestId: string;\n /** Segment ID, each segment is separated by a flush */\n segmentId: string;\n /** Synthesized audio frame */\n frame: AudioFrame;\n /** Current segment of the synthesized audio */\n deltaText?: string;\n /** Whether this is the last frame of the segment (streaming only) */\n final: boolean;\n /**\n * Timed transcripts associated with this audio packet (word-level timestamps).\n */\n timedTranscripts?: TimedString[];\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n streaming: boolean;\n // Whether this TTS supports aligned transcripts (word-level timestamps).\n alignedTranscript?: boolean;\n}\n\nexport interface TTSError {\n type: 'tts_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n ['metrics_collected']: (metrics: TTSMetrics) => void;\n ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech 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 TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n #capabilities: TTSCapabilities;\n #sampleRate: number;\n #numChannels: number;\n abstract label: string;\n\n constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n super();\n this.#capabilities = capabilities;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n }\n\n /** Returns this TTS's capabilities */\n get capabilities(): TTSCapabilities {\n return this.#capabilities;\n }\n\n /** Returns the sample rate of audio frames returned by this TTS */\n get sampleRate(): number {\n return this.#sampleRate;\n }\n\n /** Returns the channel count of audio frames returned by this TTS */\n get numChannels(): number {\n return this.#numChannels;\n }\n\n /**\n * Get the model name/identifier for this TTS 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 TTS 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 /**\n * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n */\n abstract synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream;\n\n /**\n * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\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 SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n protected queue = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected output = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected closed = false;\n protected connOptions: APIConnectOptions;\n protected abortController = new AbortController();\n\n private deferredInputStream: DeferredReadableStream<\n string | typeof SynthesizeStream.FLUSH_SENTINEL\n >;\n private logger = log();\n\n abstract label: string;\n\n #tts: TTS;\n #metricsPendingTexts: string[] = [];\n #metricsText = '';\n #monitorMetricsTask?: Promise<void>;\n #ttsRequestSpan?: Span;\n #inputTokens = 0;\n #outputTokens = 0;\n\n constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n this.#tts = tts;\n this.connOptions = connOptions;\n this.deferredInputStream = new DeferredReadableStream();\n this.pumpInput();\n\n this.abortController.signal.addEventListener('abort', () => {\n this.deferredInputStream.detachSource();\n // TODO (AJS-36) clean this up when we refactor with streams\n if (!this.input.closed) this.input.close();\n if (!this.output.closed) this.output.close();\n this.closed = true;\n });\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 _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: true,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this.connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\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 generate TTS completion 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 { tts: this.#tts.label, attempt: i + 1, error },\n `failed to synthesize 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 mainTask = async () =>\n tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n // consuming the stream directly.\n //\n // A full refactor to native Web Streams was considered but is currently deferred.\n // The primary reason is to maintain architectural parity with the Python SDK,\n // which is a key design goal for the project. This ensures a consistent developer\n // experience across both platforms.\n //\n // For more context, see the discussion in GitHub issue # 844.\n protected async pumpInput() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n break;\n }\n this.pushText(value);\n }\n this.endInput();\n } catch (error) {\n this.logger.error(error, 'Error reading deferred input stream');\n } finally {\n reader.releaseLock();\n // Ensure output is closed when the stream ends\n if (!this.#monitorMetricsTask) {\n // No text was received, close the output directly\n this.output.close();\n }\n }\n }\n\n /**\n * Set token usage for token-based TTS billing (e.g., OpenAI TTS).\n * Plugins should call this method to report token usage.\n */\n protected setTokenUsage({\n inputTokens = 0,\n outputTokens = 0,\n }: { inputTokens?: number; outputTokens?: number } = {}): void {\n this.#inputTokens = inputTokens;\n this.#outputTokens = outputTokens;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n const emit = () => {\n if (this.#metricsPendingTexts.length) {\n const text = this.#metricsPendingTexts.shift()!;\n const duration = process.hrtime.bigint() - startTime;\n const roundedAudioDurationMs = Math.round(audioDurationMs);\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: text.length,\n audioDurationMs: roundedAudioDurationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#tts.label,\n inputTokens: this.#inputTokens,\n outputTokens: this.#outputTokens,\n streamed: true,\n metadata: {\n modelProvider: this.#tts.provider,\n modelName: this.#tts.model,\n },\n };\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n }\n this.#tts.emit('metrics_collected', metrics);\n\n // Reset token usage after emitting metrics for the next segment\n this.#inputTokens = 0;\n this.#outputTokens = 0;\n }\n };\n\n for await (const audio of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(audio);\n if (audio === SynthesizeStream.END_OF_STREAM) continue;\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n // TODO(AJS-102): use frame.durationMs once available in rtc-node\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n if (audio.final) {\n emit();\n }\n }\n\n if (requestId) {\n emit();\n }\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(text: ReadableStream<string>) {\n this.deferredInputStream.setSource(text);\n }\n\n /** Push a string of text to the TTS */\n /** @deprecated Use `updateInputStream` instead */\n pushText(text: string) {\n if (!this.#monitorMetricsTask) {\n this.#monitorMetricsTask = this.monitorMetrics();\n // Close output when metrics task completes\n this.#monitorMetricsTask.finally(() => this.output.close());\n }\n this.#metricsText += text;\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(text);\n }\n\n /** Flush the TTS, causing it to process all pending text */\n flush() {\n if (this.#metricsText) {\n this.#metricsPendingTexts.push(this.#metricsText);\n this.#metricsText = '';\n }\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n this.flush();\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n return this.output.next();\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): SynthesizeStream {\n return this;\n }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\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 ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n protected output = new AsyncIterableQueue<SynthesizedAudio>();\n protected closed = false;\n abstract label: string;\n #text: string;\n #tts: TTS;\n #ttsRequestSpan?: Span;\n private _connOptions: APIConnectOptions;\n private logger = log();\n #inputTokens = 0;\n #outputTokens = 0;\n\n protected abortController = new AbortController();\n\n constructor(\n text: string,\n tts: TTS,\n connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n abortSignal?: AbortSignal,\n ) {\n this.#text = text;\n this.#tts = tts;\n this._connOptions = connOptions;\n\n if (abortSignal) {\n abortSignal.addEventListener('abort', () => this.abortController.abort(), { once: true });\n }\n\n this.monitorMetrics();\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 Promise.resolve().then(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: false,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\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 generate TTS completion 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 { tts: this.#tts.label, attempt: i + 1, error },\n `failed to generate TTS completion, 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 async mainTask() {\n return tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n protected abstract run(): Promise<void>;\n\n get inputText(): string {\n return this.#text;\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n /**\n * Set token usage for token-based TTS billing (e.g., OpenAI TTS).\n * Plugins should call this method to report token usage.\n */\n protected setTokenUsage({\n inputTokens = 0,\n outputTokens = 0,\n }: { inputTokens?: number; outputTokens?: number } = {}): void {\n this.#inputTokens = inputTokens;\n this.#outputTokens = outputTokens;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n for await (const audio of this.queue) {\n this.output.put(audio);\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: this.#text.length,\n audioDurationMs: Math.round(audioDurationMs),\n cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n label: this.#tts.label,\n inputTokens: this.#inputTokens,\n outputTokens: this.#outputTokens,\n streamed: false,\n metadata: {\n modelProvider: this.#tts.provider,\n modelName: this.#tts.model,\n },\n };\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n\n this.#tts.emit('metrics_collected', metrics);\n }\n\n /** Collect every frame into one in a single call */\n async collect(): Promise<AudioFrame> {\n const frames = [];\n for await (const event of this) {\n frames.push(event.frame);\n }\n return mergeFrames(frames);\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n 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](): ChunkedStream {\n return this;\n }\n}\n"],"mappings":"AAMA,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,iBAAiB,YAAY,cAAc;AACpD,SAAiC,6BAA6B,wBAAwB;AACtF,SAAS,oBAAoB,OAAO,aAAa,WAAW,eAAe;AAwDpE,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,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,EAkBA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,mBAAoE;AAAA,EAChF,QAAQ,IAAI,mBAEpB;AAAA,EACQ,SAAS,IAAI,mBAErB;AAAA,EACQ,SAAS;AAAA,EACT;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,SAAS,IAAI;AAAA,EAIrB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACA;AAAA,EACA,eAAe;AAAA,EACf,gBAAgB;AAAA,EAEhB,YAAY,KAAU,cAAiC,6BAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,cAAc;AACnB,SAAK,sBAAsB,IAAI,uBAAuB;AACtD,SAAK,UAAU;AAEf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,UAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,UAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,WAAW,GAAG,KAAK;AACtD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,aAAa,CAAC;AAE1D,cAAI,KAAK,YAAY,aAAa,KAAK,CAAC,MAAM,WAAW;AACvD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,YAAY,UAAU;AAC1C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,YAAY,WAAW,CAAC;AAAA,cACjF,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,4CAA4C,aAAa;AAAA,YAC3D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,WAAW,YACjB,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,IAC/D,MAAM;AAAA,IACN,WAAW;AAAA,EACb,CAAC;AAAA,EAEK,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;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMU,cAAc;AAAA,IACtB,cAAc;AAAA,IACd,eAAe;AAAA,EACjB,IAAqD,CAAC,GAAS;AAC7D,SAAK,eAAe;AACpB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,aAAa,KAAK;AAAA,UAClB,cAAc,KAAK;AAAA,UACnB,UAAU;AAAA,UACV,UAAU;AAAA,YACR,eAAe,KAAK,KAAK;AAAA,YACzB,WAAW,KAAK,KAAK;AAAA,UACvB;AAAA,QACF;AACA,YAAI,KAAK,iBAAiB;AACxB,eAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AAAA,QACxF;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAG3C,aAAK,eAAe;AACpB,aAAK,gBAAgB;AAAA,MACvB;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AAEA,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AAEX,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,mBAAqC;AAAA,EACjD,SAAS,IAAI,mBAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EACrB,eAAe;AAAA,EACf,gBAAgB;AAAA,EAEN,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,MACA,KACA,cAAiC,6BACjC,aACA;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,QAAI,aAAa;AACf,kBAAY,iBAAiB,SAAS,MAAM,KAAK,gBAAgB,MAAM,GAAG,EAAE,MAAM,KAAK,CAAC;AAAA,IAC1F;AAEA,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChF;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,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,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,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,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,WAAW;AACvB,WAAO,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,MACtE,MAAM;AAAA,MACN,WAAW;AAAA,IACb,CAAC;AAAA,EACH;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,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA;AAAA;AAAA;AAAA;AAAA,EAMU,cAAc;AAAA,IACtB,cAAc;AAAA,IACd,eAAe;AAAA,EACjB,IAAqD,CAAC,GAAS;AAC7D,SAAK,eAAe;AACpB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,cAAc,KAAK;AAAA,MACnB,UAAU;AAAA,MACV,UAAU;AAAA,QACR,eAAe,KAAK,KAAK;AAAA,QACzB,WAAW,KAAK,KAAK;AAAA,MACvB;AAAA,IACF;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AACtF,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAEA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,WAAO,YAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,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,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
|
package/dist/version.cjs
CHANGED
|
@@ -21,7 +21,7 @@ __export(version_exports, {
|
|
|
21
21
|
version: () => version
|
|
22
22
|
});
|
|
23
23
|
module.exports = __toCommonJS(version_exports);
|
|
24
|
-
const version = "1.0.
|
|
24
|
+
const version = "1.1.0-dev.0";
|
|
25
25
|
// Annotate the CommonJS export names for ESM import in node:
|
|
26
26
|
0 && (module.exports = {
|
|
27
27
|
version
|
package/dist/version.js
CHANGED
package/dist/voice/agent.cjs
CHANGED
|
@@ -41,6 +41,7 @@ var import_tts = require("../tts/index.cjs");
|
|
|
41
41
|
var import_types = require("../types.cjs");
|
|
42
42
|
var import_utils = require("../utils.cjs");
|
|
43
43
|
var import_agent_activity = require("./agent_activity.cjs");
|
|
44
|
+
var import_utils2 = require("./turn_config/utils.cjs");
|
|
44
45
|
const functionCallStorage = new import_node_async_hooks.AsyncLocalStorage();
|
|
45
46
|
const speechHandleStorage = new import_node_async_hooks.AsyncLocalStorage();
|
|
46
47
|
const activityTaskInfoStorage = /* @__PURE__ */ new WeakMap();
|
|
@@ -79,11 +80,13 @@ function isStopResponse(value) {
|
|
|
79
80
|
}
|
|
80
81
|
class Agent {
|
|
81
82
|
_id;
|
|
82
|
-
turnDetection;
|
|
83
83
|
_stt;
|
|
84
84
|
_vad;
|
|
85
85
|
_llm;
|
|
86
86
|
_tts;
|
|
87
|
+
turnHandling;
|
|
88
|
+
_interruptionDetection;
|
|
89
|
+
_allowInterruptions;
|
|
87
90
|
_useTtsAlignedTranscript;
|
|
88
91
|
/** @internal */
|
|
89
92
|
_agentActivity;
|
|
@@ -103,8 +106,11 @@ class Agent {
|
|
|
103
106
|
vad,
|
|
104
107
|
llm,
|
|
105
108
|
tts,
|
|
106
|
-
|
|
109
|
+
turnHandling,
|
|
110
|
+
useTtsAlignedTranscript,
|
|
111
|
+
allowInterruptions
|
|
107
112
|
}) {
|
|
113
|
+
var _a, _b;
|
|
108
114
|
if (id) {
|
|
109
115
|
this._id = id;
|
|
110
116
|
} else {
|
|
@@ -120,7 +126,11 @@ class Agent {
|
|
|
120
126
|
this._chatCtx = chatCtx ? chatCtx.copy({
|
|
121
127
|
toolCtx: this._tools
|
|
122
128
|
}) : import_llm.ChatContext.empty();
|
|
123
|
-
|
|
129
|
+
const migratedOptions = (0, import_utils2.migrateLegacyOptions)({
|
|
130
|
+
turnDetection,
|
|
131
|
+
options: { turnHandling, allowInterruptions }
|
|
132
|
+
});
|
|
133
|
+
this.turnHandling = migratedOptions.options.turnHandling;
|
|
124
134
|
this._vad = vad;
|
|
125
135
|
if (typeof stt === "string") {
|
|
126
136
|
this._stt = import_inference.STT.fromModelString(stt);
|
|
@@ -137,6 +147,10 @@ class Agent {
|
|
|
137
147
|
} else {
|
|
138
148
|
this._tts = tts;
|
|
139
149
|
}
|
|
150
|
+
this._interruptionDetection = (_a = this.turnHandling) == null ? void 0 : _a.interruption.mode;
|
|
151
|
+
if (((_b = this.turnHandling) == null ? void 0 : _b.interruption.mode) !== void 0) {
|
|
152
|
+
this._allowInterruptions = !!this.turnHandling.interruption.mode;
|
|
153
|
+
}
|
|
140
154
|
this._useTtsAlignedTranscript = useTtsAlignedTranscript;
|
|
141
155
|
this._agentActivity = void 0;
|
|
142
156
|
}
|
|
@@ -170,6 +184,12 @@ class Agent {
|
|
|
170
184
|
get session() {
|
|
171
185
|
return this.getActivityOrThrow().agentSession;
|
|
172
186
|
}
|
|
187
|
+
get interruptionDetection() {
|
|
188
|
+
return this._interruptionDetection;
|
|
189
|
+
}
|
|
190
|
+
get allowInterruptions() {
|
|
191
|
+
return this._allowInterruptions;
|
|
192
|
+
}
|
|
173
193
|
async onEnter() {
|
|
174
194
|
}
|
|
175
195
|
async onExit() {
|
|
@@ -233,7 +253,8 @@ class Agent {
|
|
|
233
253
|
}
|
|
234
254
|
const connOptions = activity.agentSession.connOptions.sttConnOptions;
|
|
235
255
|
const stream = wrappedStt.stream({ connOptions });
|
|
236
|
-
const audioInputStartedAt =
|
|
256
|
+
const audioInputStartedAt = activity.inputStartedAt ?? // Use input started at proxied from AudioRecognition if available
|
|
257
|
+
((_a = activity.agentSession._recorderIO) == null ? void 0 : _a.recordingStartedAt) ?? // Fallback to recording start time if available
|
|
237
258
|
activity.agentSession._startedAt ?? // Fallback to session start time
|
|
238
259
|
Date.now();
|
|
239
260
|
stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1e3;
|
package/dist/voice/agent.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall } from '../llm/index.js';\nimport {\n type ChatChunk,\n ChatContext,\n LLM,\n RealtimeModel,\n type ToolChoice,\n type ToolContext,\n} from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport { USERDATA_TIMED_TRANSCRIPT } from '../types.js';\nimport { Future, Task } from '../utils.js';\nimport type { VAD } from '../vad.js';\nimport { type AgentActivity, agentActivityStorage } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\nimport type { TimedString } from './io.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport const functionCallStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const speechHandleStorage = new AsyncLocalStorage<SpeechHandle>();\nconst activityTaskInfoStorage = new WeakMap<Task<any>, _ActivityTaskInfo>();\n\ntype _ActivityTaskInfo = {\n functionCall: FunctionCall | null;\n speechHandle: SpeechHandle | null;\n inlineTask: boolean;\n};\n\n/** @internal */\nexport function _setActivityTaskInfo<T>(\n task: Task<T>,\n options: {\n functionCall?: FunctionCall | null;\n speechHandle?: SpeechHandle | null;\n inlineTask?: boolean;\n },\n): void {\n const info = activityTaskInfoStorage.get(task) ?? {\n functionCall: null,\n speechHandle: null,\n inlineTask: false,\n };\n\n if (Object.hasOwn(options, 'functionCall')) {\n info.functionCall = options.functionCall ?? null;\n }\n if (Object.hasOwn(options, 'speechHandle')) {\n info.speechHandle = options.speechHandle ?? null;\n }\n if (Object.hasOwn(options, 'inlineTask')) {\n info.inlineTask = options.inlineTask ?? false;\n }\n\n activityTaskInfoStorage.set(task, info);\n}\n\n/** @internal */\nexport function _getActivityTaskInfo<T>(task: Task<T>): _ActivityTaskInfo | undefined {\n return activityTaskInfoStorage.get(task);\n}\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n constructor() {\n super();\n this.name = 'StopResponse';\n\n Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n value: true,\n });\n }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n return (\n value !== undefined &&\n value !== null &&\n typeof value === 'object' &&\n STOP_RESPONSE_SYMBOL in value\n );\n}\n\nexport interface ModelSettings {\n /** The tool choice to use when calling the LLM. */\n toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n id?: string;\n instructions: string;\n chatCtx?: ChatContext;\n tools?: ToolContext<UserData>;\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n allowInterruptions?: boolean;\n minConsecutiveSpeechDelay?: number;\n useTtsAlignedTranscript?: boolean;\n}\n\nexport class Agent<UserData = any> {\n private _id: string;\n private turnDetection?: TurnDetectionMode;\n private _stt?: STT;\n private _vad?: VAD;\n private _llm?: LLM | RealtimeModel;\n private _tts?: TTS;\n private _useTtsAlignedTranscript?: boolean;\n\n /** @internal */\n _agentActivity?: AgentActivity;\n\n /** @internal */\n _chatCtx: ChatContext;\n\n /** @internal */\n _instructions: string;\n\n /** @internal */\n _tools?: ToolContext<UserData>;\n\n constructor({\n id,\n instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n useTtsAlignedTranscript,\n }: AgentOptions<UserData>) {\n if (id) {\n this._id = id;\n } else {\n // Convert class name to snake_case\n const className = this.constructor.name;\n if (className === 'Agent') {\n this._id = 'default_agent';\n } else {\n this._id = className\n .replace(/([A-Z])/g, '_$1')\n .toLowerCase()\n .replace(/^_/, '');\n }\n }\n\n this._instructions = instructions;\n this._tools = { ...tools };\n this._chatCtx = chatCtx\n ? chatCtx.copy({\n toolCtx: this._tools,\n })\n : ChatContext.empty();\n\n this.turnDetection = turnDetection;\n this._vad = vad;\n\n if (typeof stt === 'string') {\n this._stt = InferenceSTT.fromModelString(stt);\n } else {\n this._stt = stt;\n }\n\n if (typeof llm === 'string') {\n this._llm = InferenceLLM.fromModelString(llm);\n } else {\n this._llm = llm;\n }\n\n if (typeof tts === 'string') {\n this._tts = InferenceTTS.fromModelString(tts);\n } else {\n this._tts = tts;\n }\n\n this._useTtsAlignedTranscript = useTtsAlignedTranscript;\n\n this._agentActivity = undefined;\n }\n\n get vad(): VAD | undefined {\n return this._vad;\n }\n\n get stt(): STT | undefined {\n return this._stt;\n }\n\n get llm(): LLM | RealtimeModel | undefined {\n return this._llm;\n }\n\n get tts(): TTS | undefined {\n return this._tts;\n }\n\n get useTtsAlignedTranscript(): boolean | undefined {\n return this._useTtsAlignedTranscript;\n }\n\n get chatCtx(): ReadonlyChatContext {\n return new ReadonlyChatContext(this._chatCtx.items);\n }\n\n get id(): string {\n return this._id;\n }\n\n get instructions(): string {\n return this._instructions;\n }\n\n get toolCtx(): ToolContext<UserData> {\n return { ...this._tools };\n }\n\n get session(): AgentSession<UserData> {\n return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n }\n\n async onEnter(): Promise<void> {}\n\n async onExit(): Promise<void> {}\n\n async transcriptionNode(\n text: ReadableStream<string | TimedString>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return Agent.default.transcriptionNode(this, text, modelSettings);\n }\n\n async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n async sttNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n return Agent.default.sttNode(this, audio, modelSettings);\n }\n\n async llmNode(\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n }\n\n async ttsNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.ttsNode(this, text, modelSettings);\n }\n\n async realtimeAudioOutputNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n }\n\n // realtime_audio_output_node\n\n getActivityOrThrow(): AgentActivity {\n if (!this._agentActivity) {\n throw new Error('Agent activity not found');\n }\n return this._agentActivity;\n }\n\n async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n if (!this._agentActivity) {\n this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n return;\n }\n\n this._agentActivity.updateChatCtx(chatCtx);\n }\n\n // TODO(parity): Add when AgentConfigUpdate is ported to ChatContext.\n async updateTools(tools: ToolContext): Promise<void> {\n if (!this._agentActivity) {\n this._tools = { ...tools };\n this._chatCtx = this._chatCtx.copy({ toolCtx: this._tools });\n return;\n }\n\n await this._agentActivity.updateTools(tools);\n }\n\n static default = {\n async sttNode(\n agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.stt) {\n throw new Error('sttNode called but no STT node is available');\n }\n\n let wrappedStt = activity.stt;\n\n if (!wrappedStt.capabilities.streaming) {\n const vad = agent.vad || activity.vad;\n if (!vad) {\n throw new Error(\n 'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n );\n }\n wrappedStt = new STTStreamAdapter(wrappedStt, vad);\n }\n\n const connOptions = activity.agentSession.connOptions.sttConnOptions;\n const stream = wrappedStt.stream({ connOptions });\n\n // Set startTimeOffset to provide linear timestamps across reconnections\n const audioInputStartedAt =\n activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n activity.agentSession._startedAt ?? // Fallback to session start time\n Date.now(); // Fallback to current time\n\n stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n stream.updateInputStream(audio);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.detachInputStream();\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n } finally {\n // Always clean up the STT stream, whether it ends naturally or is cancelled\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async llmNode(\n agent: Agent,\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.llm) {\n throw new Error('llmNode called but no LLM node is available');\n }\n\n if (!(activity.llm instanceof LLM)) {\n throw new Error(\n 'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n );\n }\n\n const { toolChoice } = modelSettings;\n const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n connOptions,\n });\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async ttsNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.tts) {\n throw new Error('ttsNode called but no TTS node is available');\n }\n\n let wrappedTts = activity.tts;\n\n if (!activity.tts.capabilities.streaming) {\n wrappedTts = new TTSStreamAdapter(wrappedTts, new BasicSentenceTokenizer());\n }\n\n const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n const stream = wrappedTts.stream({ connOptions });\n stream.updateInputStream(text);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n if (chunk === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n // Attach timed transcripts to frame.userdata\n if (chunk.timedTranscripts && chunk.timedTranscripts.length > 0) {\n chunk.frame.userdata[USERDATA_TIMED_TRANSCRIPT] = chunk.timedTranscripts;\n }\n controller.enqueue(chunk.frame);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async transcriptionNode(\n agent: Agent,\n text: ReadableStream<string | TimedString>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return text;\n },\n\n async realtimeAudioOutputNode(\n _agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return audio;\n },\n };\n}\n\nexport class AgentTask<ResultT = unknown, UserData = any> extends Agent<UserData> {\n private started = false;\n private future = new Future<ResultT>();\n\n #logger = log();\n\n get done(): boolean {\n return this.future.done;\n }\n\n complete(result: ResultT | Error): void {\n if (this.future.done) {\n throw new Error(`${this.constructor.name} is already done`);\n }\n\n if (result instanceof Error) {\n this.future.reject(result);\n } else {\n this.future.resolve(result);\n }\n\n const speechHandle = speechHandleStorage.getStore();\n if (speechHandle) {\n speechHandle._maybeRunFinalOutput = result;\n }\n }\n\n async run(): Promise<ResultT> {\n if (this.started) {\n throw new Error(\n `Task ${this.constructor.name} has already started and cannot be awaited multiple times`,\n );\n }\n this.started = true;\n\n const currentTask = Task.current();\n if (!currentTask) {\n throw new Error(`${this.constructor.name} must be executed inside a Task context`);\n }\n\n const taskInfo = _getActivityTaskInfo(currentTask);\n if (!taskInfo || !taskInfo.inlineTask) {\n throw new Error(\n `${this.constructor.name} should only be awaited inside function tools or the onEnter/onExit methods of an Agent`,\n );\n }\n\n const speechHandle = speechHandleStorage.getStore();\n const oldActivity = agentActivityStorage.getStore();\n if (!oldActivity) {\n throw new Error(`${this.constructor.name} must be executed inside an AgentActivity context`);\n }\n\n currentTask.addDoneCallback(() => {\n if (this.future.done) return;\n\n // If the Task finished before the AgentTask was completed, complete the AgentTask with an error.\n this.#logger.error(`The Task finished before ${this.constructor.name} was completed.`);\n this.complete(new Error(`The Task finished before ${this.constructor.name} was completed.`));\n });\n\n const oldAgent = oldActivity.agent;\n const session = oldActivity.agentSession;\n\n const blockedTasks: Task<any>[] = [currentTask];\n const onEnterTask = oldActivity._onEnterTask;\n\n if (onEnterTask && !onEnterTask.done && onEnterTask !== currentTask) {\n blockedTasks.push(onEnterTask);\n }\n\n if (\n taskInfo.functionCall &&\n oldActivity.llm instanceof RealtimeModel &&\n !oldActivity.llm.capabilities.manualFunctionCalls\n ) {\n this.#logger.error(\n `Realtime model does not support resuming function calls from chat context, ` +\n `using AgentTask inside a function tool may have unexpected behavior.`,\n );\n }\n\n await session._updateActivity(this, {\n previousActivity: 'pause',\n newActivity: 'start',\n blockedTasks,\n });\n\n let runState = session._globalRunState;\n if (speechHandle && runState && !runState.done()) {\n // Only unwatch the parent speech handle if there are other handles keeping the run alive.\n // When watchedHandleCount is 1 (only the parent), unwatching would drop it to 0 and\n // mark the run done prematurely — before function_call_output and assistant message arrive.\n if (runState._watchedHandleCount() > 1) {\n runState._unwatchHandle(speechHandle);\n }\n // it is OK to call _markDoneIfNeeded here, the above _updateActivity will call onEnter\n // and newly added handles keep the run alive.\n runState._markDoneIfNeeded();\n }\n\n try {\n return await this.future.await;\n } finally {\n // runState could have changed after future resolved\n runState = session._globalRunState;\n\n if (session.currentAgent !== this) {\n this.#logger.warn(\n `${this.constructor.name} completed, but the agent has changed in the meantime. ` +\n `Ignoring handoff to the previous agent, likely due to AgentSession.updateAgent being invoked.`,\n );\n await oldActivity.close();\n } else {\n if (speechHandle && runState && !runState.done()) {\n runState._watchHandle(speechHandle);\n }\n\n const mergedChatCtx = oldAgent._chatCtx.merge(this._chatCtx, {\n excludeFunctionCall: true,\n excludeInstructions: true,\n });\n oldAgent._chatCtx.items = mergedChatCtx.items;\n\n await session._updateActivity(oldAgent, {\n previousActivity: 'close',\n newActivity: 'resume',\n waitOnEnter: false,\n });\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,8BAAkC;AAClC,iBAA+B;AAC/B,uBAOO;AACP,0BAAoC;AAEpC,iBAOO;AACP,iBAAoB;AAEpB,iBAAkD;AAClD,mBAA4D;AAE5D,iBAAoE;AACpE,mBAA0C;AAC1C,mBAA6B;AAE7B,4BAAyD;AAKlD,MAAM,sBAAsB,IAAI,0CAAmD;AACnF,MAAM,sBAAsB,IAAI,0CAAgC;AACvE,MAAM,0BAA0B,oBAAI,QAAsC;AASnE,SAAS,qBACd,MACA,SAKM;AACN,QAAM,OAAO,wBAAwB,IAAI,IAAI,KAAK;AAAA,IAChD,cAAc;AAAA,IACd,cAAc;AAAA,IACd,YAAY;AAAA,EACd;AAEA,MAAI,OAAO,OAAO,SAAS,cAAc,GAAG;AAC1C,SAAK,eAAe,QAAQ,gBAAgB;AAAA,EAC9C;AACA,MAAI,OAAO,OAAO,SAAS,cAAc,GAAG;AAC1C,SAAK,eAAe,QAAQ,gBAAgB;AAAA,EAC9C;AACA,MAAI,OAAO,OAAO,SAAS,YAAY,GAAG;AACxC,SAAK,aAAa,QAAQ,cAAc;AAAA,EAC1C;AAEA,0BAAwB,IAAI,MAAM,IAAI;AACxC;AAGO,SAAS,qBAAwB,MAA8C;AACpF,SAAO,wBAAwB,IAAI,IAAI;AACzC;AACO,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAsBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,uBAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAA,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,2BAA2B;AAEhC,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,0BAA+C;AACjD,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,wCAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACsD;AACtD,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA;AAAA,EAGA,MAAM,YAAY,OAAmC;AACnD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,SAAS,EAAE,GAAG,MAAM;AACzB,WAAK,WAAW,KAAK,SAAS,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;AAC3D;AAAA,IACF;AAEA,UAAM,KAAK,eAAe,YAAY,KAAK;AAAA,EAC7C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AAhU5D;AAiUM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,aAAa,SAAS;AAE1B,UAAI,CAAC,WAAW,aAAa,WAAW;AACtC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,qBAAa,IAAI,WAAAC,cAAiB,YAAY,GAAG;AAAA,MACnD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,WAAW,OAAO,EAAE,YAAY,CAAC;AAGhD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,iBAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,aAAa,SAAS;AAE1B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,qBAAa,IAAI,WAAAC,cAAiB,YAAY,IAAI,aAAAC,kBAAuB,CAAC;AAAA,MAC5E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,WAAW,OAAO,EAAE,YAAY,CAAC;AAChD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,cACF;AAEA,kBAAI,MAAM,oBAAoB,MAAM,iBAAiB,SAAS,GAAG;AAC/D,sBAAM,MAAM,SAAS,sCAAyB,IAAI,MAAM;AAAA,cAC1D;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACsD;AACtD,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEO,MAAM,kBAAqD,MAAgB;AAAA,EACxE,UAAU;AAAA,EACV,SAAS,IAAI,oBAAgB;AAAA,EAErC,cAAU,gBAAI;AAAA,EAEd,IAAI,OAAgB;AAClB,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,SAAS,QAA+B;AACtC,QAAI,KAAK,OAAO,MAAM;AACpB,YAAM,IAAI,MAAM,GAAG,KAAK,YAAY,IAAI,kBAAkB;AAAA,IAC5D;AAEA,QAAI,kBAAkB,OAAO;AAC3B,WAAK,OAAO,OAAO,MAAM;AAAA,IAC3B,OAAO;AACL,WAAK,OAAO,QAAQ,MAAM;AAAA,IAC5B;AAEA,UAAM,eAAe,oBAAoB,SAAS;AAClD,QAAI,cAAc;AAChB,mBAAa,uBAAuB;AAAA,IACtC;AAAA,EACF;AAAA,EAEA,MAAM,MAAwB;AAC5B,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI;AAAA,QACR,QAAQ,KAAK,YAAY,IAAI;AAAA,MAC/B;AAAA,IACF;AACA,SAAK,UAAU;AAEf,UAAM,cAAc,kBAAK,QAAQ;AACjC,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,GAAG,KAAK,YAAY,IAAI,yCAAyC;AAAA,IACnF;AAEA,UAAM,WAAW,qBAAqB,WAAW;AACjD,QAAI,CAAC,YAAY,CAAC,SAAS,YAAY;AACrC,YAAM,IAAI;AAAA,QACR,GAAG,KAAK,YAAY,IAAI;AAAA,MAC1B;AAAA,IACF;AAEA,UAAM,eAAe,oBAAoB,SAAS;AAClD,UAAM,cAAc,2CAAqB,SAAS;AAClD,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,GAAG,KAAK,YAAY,IAAI,mDAAmD;AAAA,IAC7F;AAEA,gBAAY,gBAAgB,MAAM;AAChC,UAAI,KAAK,OAAO,KAAM;AAGtB,WAAK,QAAQ,MAAM,4BAA4B,KAAK,YAAY,IAAI,iBAAiB;AACrF,WAAK,SAAS,IAAI,MAAM,4BAA4B,KAAK,YAAY,IAAI,iBAAiB,CAAC;AAAA,IAC7F,CAAC;AAED,UAAM,WAAW,YAAY;AAC7B,UAAM,UAAU,YAAY;AAE5B,UAAM,eAA4B,CAAC,WAAW;AAC9C,UAAM,cAAc,YAAY;AAEhC,QAAI,eAAe,CAAC,YAAY,QAAQ,gBAAgB,aAAa;AACnE,mBAAa,KAAK,WAAW;AAAA,IAC/B;AAEA,QACE,SAAS,gBACT,YAAY,eAAe,4BAC3B,CAAC,YAAY,IAAI,aAAa,qBAC9B;AACA,WAAK,QAAQ;AAAA,QACX;AAAA,MAEF;AAAA,IACF;AAEA,UAAM,QAAQ,gBAAgB,MAAM;AAAA,MAClC,kBAAkB;AAAA,MAClB,aAAa;AAAA,MACb;AAAA,IACF,CAAC;AAED,QAAI,WAAW,QAAQ;AACvB,QAAI,gBAAgB,YAAY,CAAC,SAAS,KAAK,GAAG;AAIhD,UAAI,SAAS,oBAAoB,IAAI,GAAG;AACtC,iBAAS,eAAe,YAAY;AAAA,MACtC;AAGA,eAAS,kBAAkB;AAAA,IAC7B;AAEA,QAAI;AACF,aAAO,MAAM,KAAK,OAAO;AAAA,IAC3B,UAAE;AAEA,iBAAW,QAAQ;AAEnB,UAAI,QAAQ,iBAAiB,MAAM;AACjC,aAAK,QAAQ;AAAA,UACX,GAAG,KAAK,YAAY,IAAI;AAAA,QAE1B;AACA,cAAM,YAAY,MAAM;AAAA,MAC1B,OAAO;AACL,YAAI,gBAAgB,YAAY,CAAC,SAAS,KAAK,GAAG;AAChD,mBAAS,aAAa,YAAY;AAAA,QACpC;AAEA,cAAM,gBAAgB,SAAS,SAAS,MAAM,KAAK,UAAU;AAAA,UAC3D,qBAAqB;AAAA,UACrB,qBAAqB;AAAA,QACvB,CAAC;AACD,iBAAS,SAAS,QAAQ,cAAc;AAExC,cAAM,QAAQ,gBAAgB,UAAU;AAAA,UACtC,kBAAkB;AAAA,UAClB,aAAa;AAAA,UACb,aAAa;AAAA,QACf,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF;AACF;","names":["InferenceSTT","InferenceLLM","InferenceTTS","STTStreamAdapter","TTSStreamAdapter","BasicSentenceTokenizer"]}
|
|
1
|
+
{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall } from '../llm/index.js';\nimport {\n type ChatChunk,\n ChatContext,\n LLM,\n RealtimeModel,\n type ToolChoice,\n type ToolContext,\n} from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport { USERDATA_TIMED_TRANSCRIPT } from '../types.js';\nimport { Future, Task } from '../utils.js';\nimport type { VAD } from '../vad.js';\nimport { type AgentActivity, agentActivityStorage } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\nimport type { TimedString } from './io.js';\nimport type { SpeechHandle } from './speech_handle.js';\nimport type { InterruptionOptions } from './turn_config/interruption.js';\nimport type { TurnHandlingOptions } from './turn_config/turn_handling.js';\nimport { migrateLegacyOptions } from './turn_config/utils.js';\n\nexport const functionCallStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const speechHandleStorage = new AsyncLocalStorage<SpeechHandle>();\nconst activityTaskInfoStorage = new WeakMap<Task<any>, _ActivityTaskInfo>();\n\ntype _ActivityTaskInfo = {\n functionCall: FunctionCall | null;\n speechHandle: SpeechHandle | null;\n inlineTask: boolean;\n};\n\n/** @internal */\nexport function _setActivityTaskInfo<T>(\n task: Task<T>,\n options: {\n functionCall?: FunctionCall | null;\n speechHandle?: SpeechHandle | null;\n inlineTask?: boolean;\n },\n): void {\n const info = activityTaskInfoStorage.get(task) ?? {\n functionCall: null,\n speechHandle: null,\n inlineTask: false,\n };\n\n if (Object.hasOwn(options, 'functionCall')) {\n info.functionCall = options.functionCall ?? null;\n }\n if (Object.hasOwn(options, 'speechHandle')) {\n info.speechHandle = options.speechHandle ?? null;\n }\n if (Object.hasOwn(options, 'inlineTask')) {\n info.inlineTask = options.inlineTask ?? false;\n }\n\n activityTaskInfoStorage.set(task, info);\n}\n\n/** @internal */\nexport function _getActivityTaskInfo<T>(task: Task<T>): _ActivityTaskInfo | undefined {\n return activityTaskInfoStorage.get(task);\n}\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n constructor() {\n super();\n this.name = 'StopResponse';\n\n Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n value: true,\n });\n }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n return (\n value !== undefined &&\n value !== null &&\n typeof value === 'object' &&\n STOP_RESPONSE_SYMBOL in value\n );\n}\n\nexport interface ModelSettings {\n /** The tool choice to use when calling the LLM. */\n toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n id?: string;\n instructions: string;\n chatCtx?: ChatContext;\n tools?: ToolContext<UserData>;\n /** @deprecated use turnHandling instead */\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n allowInterruptions?: boolean;\n minConsecutiveSpeechDelay?: number;\n turnHandling?: TurnHandlingOptions;\n useTtsAlignedTranscript?: boolean;\n}\n\nexport class Agent<UserData = any> {\n private _id: string;\n private _stt?: STT;\n private _vad?: VAD;\n private _llm?: LLM | RealtimeModel;\n private _tts?: TTS;\n private turnHandling?: TurnHandlingOptions;\n private _interruptionDetection: InterruptionOptions['mode'];\n private _allowInterruptions?: boolean;\n private _useTtsAlignedTranscript?: boolean;\n\n /** @internal */\n _agentActivity?: AgentActivity;\n\n /** @internal */\n _chatCtx: ChatContext;\n\n /** @internal */\n _instructions: string;\n\n /** @internal */\n _tools?: ToolContext<UserData>;\n\n constructor({\n id,\n instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n turnHandling,\n useTtsAlignedTranscript,\n allowInterruptions,\n }: AgentOptions<UserData>) {\n if (id) {\n this._id = id;\n } else {\n // Convert class name to snake_case\n const className = this.constructor.name;\n if (className === 'Agent') {\n this._id = 'default_agent';\n } else {\n this._id = className\n .replace(/([A-Z])/g, '_$1')\n .toLowerCase()\n .replace(/^_/, '');\n }\n }\n\n this._instructions = instructions;\n this._tools = { ...tools };\n this._chatCtx = chatCtx\n ? chatCtx.copy({\n toolCtx: this._tools,\n })\n : ChatContext.empty();\n\n const migratedOptions = migrateLegacyOptions({\n turnDetection,\n options: { turnHandling, allowInterruptions },\n });\n this.turnHandling = migratedOptions.options.turnHandling;\n\n this._vad = vad;\n\n if (typeof stt === 'string') {\n this._stt = InferenceSTT.fromModelString(stt);\n } else {\n this._stt = stt;\n }\n\n if (typeof llm === 'string') {\n this._llm = InferenceLLM.fromModelString(llm);\n } else {\n this._llm = llm;\n }\n\n if (typeof tts === 'string') {\n this._tts = InferenceTTS.fromModelString(tts);\n } else {\n this._tts = tts;\n }\n\n this._interruptionDetection = this.turnHandling?.interruption.mode;\n if (this.turnHandling?.interruption.mode !== undefined) {\n this._allowInterruptions = !!this.turnHandling.interruption.mode;\n }\n this._useTtsAlignedTranscript = useTtsAlignedTranscript;\n\n this._agentActivity = undefined;\n }\n\n get vad(): VAD | undefined {\n return this._vad;\n }\n\n get stt(): STT | undefined {\n return this._stt;\n }\n\n get llm(): LLM | RealtimeModel | undefined {\n return this._llm;\n }\n\n get tts(): TTS | undefined {\n return this._tts;\n }\n\n get useTtsAlignedTranscript(): boolean | undefined {\n return this._useTtsAlignedTranscript;\n }\n\n get chatCtx(): ReadonlyChatContext {\n return new ReadonlyChatContext(this._chatCtx.items);\n }\n\n get id(): string {\n return this._id;\n }\n\n get instructions(): string {\n return this._instructions;\n }\n\n get toolCtx(): ToolContext<UserData> {\n return { ...this._tools };\n }\n\n get session(): AgentSession<UserData> {\n return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n }\n\n get interruptionDetection(): InterruptionOptions['mode'] {\n return this._interruptionDetection;\n }\n\n get allowInterruptions(): boolean | undefined {\n return this._allowInterruptions;\n }\n\n async onEnter(): Promise<void> {}\n\n async onExit(): Promise<void> {}\n\n async transcriptionNode(\n text: ReadableStream<string | TimedString>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return Agent.default.transcriptionNode(this, text, modelSettings);\n }\n\n async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n async sttNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n return Agent.default.sttNode(this, audio, modelSettings);\n }\n\n async llmNode(\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n }\n\n async ttsNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.ttsNode(this, text, modelSettings);\n }\n\n async realtimeAudioOutputNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n }\n\n // realtime_audio_output_node\n\n getActivityOrThrow(): AgentActivity {\n if (!this._agentActivity) {\n throw new Error('Agent activity not found');\n }\n return this._agentActivity;\n }\n\n async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n if (!this._agentActivity) {\n this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n return;\n }\n\n this._agentActivity.updateChatCtx(chatCtx);\n }\n\n // TODO(parity): Add when AgentConfigUpdate is ported to ChatContext.\n async updateTools(tools: ToolContext): Promise<void> {\n if (!this._agentActivity) {\n this._tools = { ...tools };\n this._chatCtx = this._chatCtx.copy({ toolCtx: this._tools });\n return;\n }\n\n await this._agentActivity.updateTools(tools);\n }\n\n static default = {\n async sttNode(\n agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.stt) {\n throw new Error('sttNode called but no STT node is available');\n }\n\n let wrappedStt = activity.stt;\n\n if (!wrappedStt.capabilities.streaming) {\n const vad = agent.vad || activity.vad;\n if (!vad) {\n throw new Error(\n 'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n );\n }\n wrappedStt = new STTStreamAdapter(wrappedStt, vad);\n }\n\n const connOptions = activity.agentSession.connOptions.sttConnOptions;\n const stream = wrappedStt.stream({ connOptions });\n\n // Set startTimeOffset to provide linear timestamps across reconnections\n const audioInputStartedAt =\n activity.inputStartedAt ?? // Use input started at proxied from AudioRecognition if available\n activity.agentSession._recorderIO?.recordingStartedAt ?? // Fallback to recording start time if available\n activity.agentSession._startedAt ?? // Fallback to session start time\n Date.now(); // Fallback to current time\n\n stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n stream.updateInputStream(audio);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.detachInputStream();\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n } finally {\n // Always clean up the STT stream, whether it ends naturally or is cancelled\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async llmNode(\n agent: Agent,\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.llm) {\n throw new Error('llmNode called but no LLM node is available');\n }\n\n if (!(activity.llm instanceof LLM)) {\n throw new Error(\n 'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n );\n }\n\n const { toolChoice } = modelSettings;\n const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n connOptions,\n });\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async ttsNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.tts) {\n throw new Error('ttsNode called but no TTS node is available');\n }\n\n let wrappedTts = activity.tts;\n\n if (!activity.tts.capabilities.streaming) {\n wrappedTts = new TTSStreamAdapter(wrappedTts, new BasicSentenceTokenizer());\n }\n\n const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n const stream = wrappedTts.stream({ connOptions });\n stream.updateInputStream(text);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n if (chunk === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n // Attach timed transcripts to frame.userdata\n if (chunk.timedTranscripts && chunk.timedTranscripts.length > 0) {\n chunk.frame.userdata[USERDATA_TIMED_TRANSCRIPT] = chunk.timedTranscripts;\n }\n controller.enqueue(chunk.frame);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async transcriptionNode(\n agent: Agent,\n text: ReadableStream<string | TimedString>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return text;\n },\n\n async realtimeAudioOutputNode(\n _agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return audio;\n },\n };\n}\n\nexport class AgentTask<ResultT = unknown, UserData = any> extends Agent<UserData> {\n private started = false;\n private future = new Future<ResultT>();\n\n #logger = log();\n\n get done(): boolean {\n return this.future.done;\n }\n\n complete(result: ResultT | Error): void {\n if (this.future.done) {\n throw new Error(`${this.constructor.name} is already done`);\n }\n\n if (result instanceof Error) {\n this.future.reject(result);\n } else {\n this.future.resolve(result);\n }\n\n const speechHandle = speechHandleStorage.getStore();\n if (speechHandle) {\n speechHandle._maybeRunFinalOutput = result;\n }\n }\n\n async run(): Promise<ResultT> {\n if (this.started) {\n throw new Error(\n `Task ${this.constructor.name} has already started and cannot be awaited multiple times`,\n );\n }\n this.started = true;\n\n const currentTask = Task.current();\n if (!currentTask) {\n throw new Error(`${this.constructor.name} must be executed inside a Task context`);\n }\n\n const taskInfo = _getActivityTaskInfo(currentTask);\n if (!taskInfo || !taskInfo.inlineTask) {\n throw new Error(\n `${this.constructor.name} should only be awaited inside function tools or the onEnter/onExit methods of an Agent`,\n );\n }\n\n const speechHandle = speechHandleStorage.getStore();\n const oldActivity = agentActivityStorage.getStore();\n if (!oldActivity) {\n throw new Error(`${this.constructor.name} must be executed inside an AgentActivity context`);\n }\n\n currentTask.addDoneCallback(() => {\n if (this.future.done) return;\n\n // If the Task finished before the AgentTask was completed, complete the AgentTask with an error.\n this.#logger.error(`The Task finished before ${this.constructor.name} was completed.`);\n this.complete(new Error(`The Task finished before ${this.constructor.name} was completed.`));\n });\n\n const oldAgent = oldActivity.agent;\n const session = oldActivity.agentSession;\n\n const blockedTasks: Task<any>[] = [currentTask];\n const onEnterTask = oldActivity._onEnterTask;\n\n if (onEnterTask && !onEnterTask.done && onEnterTask !== currentTask) {\n blockedTasks.push(onEnterTask);\n }\n\n if (\n taskInfo.functionCall &&\n oldActivity.llm instanceof RealtimeModel &&\n !oldActivity.llm.capabilities.manualFunctionCalls\n ) {\n this.#logger.error(\n `Realtime model does not support resuming function calls from chat context, ` +\n `using AgentTask inside a function tool may have unexpected behavior.`,\n );\n }\n\n await session._updateActivity(this, {\n previousActivity: 'pause',\n newActivity: 'start',\n blockedTasks,\n });\n\n let runState = session._globalRunState;\n if (speechHandle && runState && !runState.done()) {\n // Only unwatch the parent speech handle if there are other handles keeping the run alive.\n // When watchedHandleCount is 1 (only the parent), unwatching would drop it to 0 and\n // mark the run done prematurely — before function_call_output and assistant message arrive.\n if (runState._watchedHandleCount() > 1) {\n runState._unwatchHandle(speechHandle);\n }\n // it is OK to call _markDoneIfNeeded here, the above _updateActivity will call onEnter\n // and newly added handles keep the run alive.\n runState._markDoneIfNeeded();\n }\n\n try {\n return await this.future.await;\n } finally {\n // runState could have changed after future resolved\n runState = session._globalRunState;\n\n if (session.currentAgent !== this) {\n this.#logger.warn(\n `${this.constructor.name} completed, but the agent has changed in the meantime. ` +\n `Ignoring handoff to the previous agent, likely due to AgentSession.updateAgent being invoked.`,\n );\n await oldActivity.close();\n } else {\n if (speechHandle && runState && !runState.done()) {\n runState._watchHandle(speechHandle);\n }\n\n const mergedChatCtx = oldAgent._chatCtx.merge(this._chatCtx, {\n excludeFunctionCall: true,\n excludeInstructions: true,\n });\n oldAgent._chatCtx.items = mergedChatCtx.items;\n\n await session._updateActivity(oldAgent, {\n previousActivity: 'close',\n newActivity: 'resume',\n waitOnEnter: false,\n });\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,8BAAkC;AAClC,iBAA+B;AAC/B,uBAOO;AACP,0BAAoC;AAEpC,iBAOO;AACP,iBAAoB;AAEpB,iBAAkD;AAClD,mBAA4D;AAE5D,iBAAoE;AACpE,mBAA0C;AAC1C,mBAA6B;AAE7B,4BAAyD;AAMzD,IAAAA,gBAAqC;AAE9B,MAAM,sBAAsB,IAAI,0CAAmD;AACnF,MAAM,sBAAsB,IAAI,0CAAgC;AACvE,MAAM,0BAA0B,oBAAI,QAAsC;AASnE,SAAS,qBACd,MACA,SAKM;AACN,QAAM,OAAO,wBAAwB,IAAI,IAAI,KAAK;AAAA,IAChD,cAAc;AAAA,IACd,cAAc;AAAA,IACd,YAAY;AAAA,EACd;AAEA,MAAI,OAAO,OAAO,SAAS,cAAc,GAAG;AAC1C,SAAK,eAAe,QAAQ,gBAAgB;AAAA,EAC9C;AACA,MAAI,OAAO,OAAO,SAAS,cAAc,GAAG;AAC1C,SAAK,eAAe,QAAQ,gBAAgB;AAAA,EAC9C;AACA,MAAI,OAAO,OAAO,SAAS,YAAY,GAAG;AACxC,SAAK,aAAa,QAAQ,cAAc;AAAA,EAC1C;AAEA,0BAAwB,IAAI,MAAM,IAAI;AACxC;AAGO,SAAS,qBAAwB,MAA8C;AACpF,SAAO,wBAAwB,IAAI,IAAI;AACzC;AACO,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAwBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AAnK7B;AAoKI,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,uBAAY,MAAM;AAEtB,UAAM,sBAAkB,oCAAqB;AAAA,MAC3C;AAAA,MACA,SAAS,EAAE,cAAc,mBAAmB;AAAA,IAC9C,CAAC;AACD,SAAK,eAAe,gBAAgB,QAAQ;AAE5C,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,0BAAyB,UAAK,iBAAL,mBAAmB,aAAa;AAC9D,UAAI,UAAK,iBAAL,mBAAmB,aAAa,UAAS,QAAW;AACtD,WAAK,sBAAsB,CAAC,CAAC,KAAK,aAAa,aAAa;AAAA,IAC9D;AACA,SAAK,2BAA2B;AAEhC,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,0BAA+C;AACjD,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,wCAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,IAAI,wBAAqD;AACvD,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,qBAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACsD;AACtD,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA;AAAA,EAGA,MAAM,YAAY,OAAmC;AACnD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,SAAS,EAAE,GAAG,MAAM;AACzB,WAAK,WAAW,KAAK,SAAS,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;AAC3D;AAAA,IACF;AAEA,UAAM,KAAK,eAAe,YAAY,KAAK;AAAA,EAC7C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AA1V5D;AA2VM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,aAAa,SAAS;AAE1B,UAAI,CAAC,WAAW,aAAa,WAAW;AACtC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,qBAAa,IAAI,WAAAC,cAAiB,YAAY,GAAG;AAAA,MACnD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,WAAW,OAAO,EAAE,YAAY,CAAC;AAGhD,YAAM,sBACJ,SAAS;AAAA,QACT,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,iBAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,aAAa,SAAS;AAE1B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,qBAAa,IAAI,WAAAC,cAAiB,YAAY,IAAI,aAAAC,kBAAuB,CAAC;AAAA,MAC5E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,WAAW,OAAO,EAAE,YAAY,CAAC;AAChD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,cACF;AAEA,kBAAI,MAAM,oBAAoB,MAAM,iBAAiB,SAAS,GAAG;AAC/D,sBAAM,MAAM,SAAS,sCAAyB,IAAI,MAAM;AAAA,cAC1D;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACsD;AACtD,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEO,MAAM,kBAAqD,MAAgB;AAAA,EACxE,UAAU;AAAA,EACV,SAAS,IAAI,oBAAgB;AAAA,EAErC,cAAU,gBAAI;AAAA,EAEd,IAAI,OAAgB;AAClB,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,SAAS,QAA+B;AACtC,QAAI,KAAK,OAAO,MAAM;AACpB,YAAM,IAAI,MAAM,GAAG,KAAK,YAAY,IAAI,kBAAkB;AAAA,IAC5D;AAEA,QAAI,kBAAkB,OAAO;AAC3B,WAAK,OAAO,OAAO,MAAM;AAAA,IAC3B,OAAO;AACL,WAAK,OAAO,QAAQ,MAAM;AAAA,IAC5B;AAEA,UAAM,eAAe,oBAAoB,SAAS;AAClD,QAAI,cAAc;AAChB,mBAAa,uBAAuB;AAAA,IACtC;AAAA,EACF;AAAA,EAEA,MAAM,MAAwB;AAC5B,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI;AAAA,QACR,QAAQ,KAAK,YAAY,IAAI;AAAA,MAC/B;AAAA,IACF;AACA,SAAK,UAAU;AAEf,UAAM,cAAc,kBAAK,QAAQ;AACjC,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,GAAG,KAAK,YAAY,IAAI,yCAAyC;AAAA,IACnF;AAEA,UAAM,WAAW,qBAAqB,WAAW;AACjD,QAAI,CAAC,YAAY,CAAC,SAAS,YAAY;AACrC,YAAM,IAAI;AAAA,QACR,GAAG,KAAK,YAAY,IAAI;AAAA,MAC1B;AAAA,IACF;AAEA,UAAM,eAAe,oBAAoB,SAAS;AAClD,UAAM,cAAc,2CAAqB,SAAS;AAClD,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,GAAG,KAAK,YAAY,IAAI,mDAAmD;AAAA,IAC7F;AAEA,gBAAY,gBAAgB,MAAM;AAChC,UAAI,KAAK,OAAO,KAAM;AAGtB,WAAK,QAAQ,MAAM,4BAA4B,KAAK,YAAY,IAAI,iBAAiB;AACrF,WAAK,SAAS,IAAI,MAAM,4BAA4B,KAAK,YAAY,IAAI,iBAAiB,CAAC;AAAA,IAC7F,CAAC;AAED,UAAM,WAAW,YAAY;AAC7B,UAAM,UAAU,YAAY;AAE5B,UAAM,eAA4B,CAAC,WAAW;AAC9C,UAAM,cAAc,YAAY;AAEhC,QAAI,eAAe,CAAC,YAAY,QAAQ,gBAAgB,aAAa;AACnE,mBAAa,KAAK,WAAW;AAAA,IAC/B;AAEA,QACE,SAAS,gBACT,YAAY,eAAe,4BAC3B,CAAC,YAAY,IAAI,aAAa,qBAC9B;AACA,WAAK,QAAQ;AAAA,QACX;AAAA,MAEF;AAAA,IACF;AAEA,UAAM,QAAQ,gBAAgB,MAAM;AAAA,MAClC,kBAAkB;AAAA,MAClB,aAAa;AAAA,MACb;AAAA,IACF,CAAC;AAED,QAAI,WAAW,QAAQ;AACvB,QAAI,gBAAgB,YAAY,CAAC,SAAS,KAAK,GAAG;AAIhD,UAAI,SAAS,oBAAoB,IAAI,GAAG;AACtC,iBAAS,eAAe,YAAY;AAAA,MACtC;AAGA,eAAS,kBAAkB;AAAA,IAC7B;AAEA,QAAI;AACF,aAAO,MAAM,KAAK,OAAO;AAAA,IAC3B,UAAE;AAEA,iBAAW,QAAQ;AAEnB,UAAI,QAAQ,iBAAiB,MAAM;AACjC,aAAK,QAAQ;AAAA,UACX,GAAG,KAAK,YAAY,IAAI;AAAA,QAE1B;AACA,cAAM,YAAY,MAAM;AAAA,MAC1B,OAAO;AACL,YAAI,gBAAgB,YAAY,CAAC,SAAS,KAAK,GAAG;AAChD,mBAAS,aAAa,YAAY;AAAA,QACpC;AAEA,cAAM,gBAAgB,SAAS,SAAS,MAAM,KAAK,UAAU;AAAA,UAC3D,qBAAqB;AAAA,UACrB,qBAAqB;AAAA,QACvB,CAAC;AACD,iBAAS,SAAS,QAAQ,cAAc;AAExC,cAAM,QAAQ,gBAAgB,UAAU;AAAA,UACtC,kBAAkB;AAAA,UAClB,aAAa;AAAA,UACb,aAAa;AAAA,QACf,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF;AACF;","names":["import_utils","InferenceSTT","InferenceLLM","InferenceTTS","STTStreamAdapter","TTSStreamAdapter","BasicSentenceTokenizer"]}
|
package/dist/voice/agent.d.cts
CHANGED
|
@@ -15,6 +15,8 @@ import { type AgentActivity } from './agent_activity.js';
|
|
|
15
15
|
import type { AgentSession, TurnDetectionMode } from './agent_session.js';
|
|
16
16
|
import type { TimedString } from './io.js';
|
|
17
17
|
import type { SpeechHandle } from './speech_handle.js';
|
|
18
|
+
import type { InterruptionOptions } from './turn_config/interruption.js';
|
|
19
|
+
import type { TurnHandlingOptions } from './turn_config/turn_handling.js';
|
|
18
20
|
export declare const functionCallStorage: AsyncLocalStorage<{
|
|
19
21
|
functionCall?: FunctionCall | undefined;
|
|
20
22
|
}>;
|
|
@@ -46,6 +48,7 @@ export interface AgentOptions<UserData> {
|
|
|
46
48
|
instructions: string;
|
|
47
49
|
chatCtx?: ChatContext;
|
|
48
50
|
tools?: ToolContext<UserData>;
|
|
51
|
+
/** @deprecated use turnHandling instead */
|
|
49
52
|
turnDetection?: TurnDetectionMode;
|
|
50
53
|
stt?: STT | STTModelString;
|
|
51
54
|
vad?: VAD;
|
|
@@ -53,15 +56,18 @@ export interface AgentOptions<UserData> {
|
|
|
53
56
|
tts?: TTS | TTSModelString;
|
|
54
57
|
allowInterruptions?: boolean;
|
|
55
58
|
minConsecutiveSpeechDelay?: number;
|
|
59
|
+
turnHandling?: TurnHandlingOptions;
|
|
56
60
|
useTtsAlignedTranscript?: boolean;
|
|
57
61
|
}
|
|
58
62
|
export declare class Agent<UserData = any> {
|
|
59
63
|
private _id;
|
|
60
|
-
private turnDetection?;
|
|
61
64
|
private _stt?;
|
|
62
65
|
private _vad?;
|
|
63
66
|
private _llm?;
|
|
64
67
|
private _tts?;
|
|
68
|
+
private turnHandling?;
|
|
69
|
+
private _interruptionDetection;
|
|
70
|
+
private _allowInterruptions?;
|
|
65
71
|
private _useTtsAlignedTranscript?;
|
|
66
72
|
/** @internal */
|
|
67
73
|
_agentActivity?: AgentActivity;
|
|
@@ -71,7 +77,7 @@ export declare class Agent<UserData = any> {
|
|
|
71
77
|
_instructions: string;
|
|
72
78
|
/** @internal */
|
|
73
79
|
_tools?: ToolContext<UserData>;
|
|
74
|
-
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, useTtsAlignedTranscript, }: AgentOptions<UserData>);
|
|
80
|
+
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, turnHandling, useTtsAlignedTranscript, allowInterruptions, }: AgentOptions<UserData>);
|
|
75
81
|
get vad(): VAD | undefined;
|
|
76
82
|
get stt(): STT | undefined;
|
|
77
83
|
get llm(): LLM | RealtimeModel | undefined;
|
|
@@ -82,6 +88,8 @@ export declare class Agent<UserData = any> {
|
|
|
82
88
|
get instructions(): string;
|
|
83
89
|
get toolCtx(): ToolContext<UserData>;
|
|
84
90
|
get session(): AgentSession<UserData>;
|
|
91
|
+
get interruptionDetection(): InterruptionOptions['mode'];
|
|
92
|
+
get allowInterruptions(): boolean | undefined;
|
|
85
93
|
onEnter(): Promise<void>;
|
|
86
94
|
onExit(): Promise<void>;
|
|
87
95
|
transcriptionNode(text: ReadableStream<string | TimedString>, modelSettings: ModelSettings): Promise<ReadableStream<string | TimedString> | null>;
|
package/dist/voice/agent.d.ts
CHANGED
|
@@ -15,6 +15,8 @@ import { type AgentActivity } from './agent_activity.js';
|
|
|
15
15
|
import type { AgentSession, TurnDetectionMode } from './agent_session.js';
|
|
16
16
|
import type { TimedString } from './io.js';
|
|
17
17
|
import type { SpeechHandle } from './speech_handle.js';
|
|
18
|
+
import type { InterruptionOptions } from './turn_config/interruption.js';
|
|
19
|
+
import type { TurnHandlingOptions } from './turn_config/turn_handling.js';
|
|
18
20
|
export declare const functionCallStorage: AsyncLocalStorage<{
|
|
19
21
|
functionCall?: FunctionCall | undefined;
|
|
20
22
|
}>;
|
|
@@ -46,6 +48,7 @@ export interface AgentOptions<UserData> {
|
|
|
46
48
|
instructions: string;
|
|
47
49
|
chatCtx?: ChatContext;
|
|
48
50
|
tools?: ToolContext<UserData>;
|
|
51
|
+
/** @deprecated use turnHandling instead */
|
|
49
52
|
turnDetection?: TurnDetectionMode;
|
|
50
53
|
stt?: STT | STTModelString;
|
|
51
54
|
vad?: VAD;
|
|
@@ -53,15 +56,18 @@ export interface AgentOptions<UserData> {
|
|
|
53
56
|
tts?: TTS | TTSModelString;
|
|
54
57
|
allowInterruptions?: boolean;
|
|
55
58
|
minConsecutiveSpeechDelay?: number;
|
|
59
|
+
turnHandling?: TurnHandlingOptions;
|
|
56
60
|
useTtsAlignedTranscript?: boolean;
|
|
57
61
|
}
|
|
58
62
|
export declare class Agent<UserData = any> {
|
|
59
63
|
private _id;
|
|
60
|
-
private turnDetection?;
|
|
61
64
|
private _stt?;
|
|
62
65
|
private _vad?;
|
|
63
66
|
private _llm?;
|
|
64
67
|
private _tts?;
|
|
68
|
+
private turnHandling?;
|
|
69
|
+
private _interruptionDetection;
|
|
70
|
+
private _allowInterruptions?;
|
|
65
71
|
private _useTtsAlignedTranscript?;
|
|
66
72
|
/** @internal */
|
|
67
73
|
_agentActivity?: AgentActivity;
|
|
@@ -71,7 +77,7 @@ export declare class Agent<UserData = any> {
|
|
|
71
77
|
_instructions: string;
|
|
72
78
|
/** @internal */
|
|
73
79
|
_tools?: ToolContext<UserData>;
|
|
74
|
-
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, useTtsAlignedTranscript, }: AgentOptions<UserData>);
|
|
80
|
+
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, turnHandling, useTtsAlignedTranscript, allowInterruptions, }: AgentOptions<UserData>);
|
|
75
81
|
get vad(): VAD | undefined;
|
|
76
82
|
get stt(): STT | undefined;
|
|
77
83
|
get llm(): LLM | RealtimeModel | undefined;
|
|
@@ -82,6 +88,8 @@ export declare class Agent<UserData = any> {
|
|
|
82
88
|
get instructions(): string;
|
|
83
89
|
get toolCtx(): ToolContext<UserData>;
|
|
84
90
|
get session(): AgentSession<UserData>;
|
|
91
|
+
get interruptionDetection(): InterruptionOptions['mode'];
|
|
92
|
+
get allowInterruptions(): boolean | undefined;
|
|
85
93
|
onEnter(): Promise<void>;
|
|
86
94
|
onExit(): Promise<void>;
|
|
87
95
|
transcriptionNode(text: ReadableStream<string | TimedString>, modelSettings: ModelSettings): Promise<ReadableStream<string | TimedString> | null>;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent.d.ts","sourceRoot":"","sources":["../../src/voice/agent.ts"],"names":[],"mappings":";;AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAC/B,OAAO,EAAE,mBAAmB,EAAE,MAAM,wBAAwB,CAAC;AAC7D,OAAO,KAAK,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAC;AACjE,OAAO,EACL,KAAK,SAAS,EACd,WAAW,EACX,GAAG,EACH,aAAa,EACb,KAAK,UAAU,EACf,KAAK,WAAW,EACjB,MAAM,iBAAiB,CAAC;AAEzB,OAAO,KAAK,EAAE,GAAG,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAGxD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAG3C,OAAO,EAAU,IAAI,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,EAAE,KAAK,aAAa,EAAwB,MAAM,qBAAqB,CAAC;AAC/E,OAAO,KAAK,EAAE,YAAY,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC1E,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAC3C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;
|
|
1
|
+
{"version":3,"file":"agent.d.ts","sourceRoot":"","sources":["../../src/voice/agent.ts"],"names":[],"mappings":";;AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAC/B,OAAO,EAAE,mBAAmB,EAAE,MAAM,wBAAwB,CAAC;AAC7D,OAAO,KAAK,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAC;AACjE,OAAO,EACL,KAAK,SAAS,EACd,WAAW,EACX,GAAG,EACH,aAAa,EACb,KAAK,UAAU,EACf,KAAK,WAAW,EACjB,MAAM,iBAAiB,CAAC;AAEzB,OAAO,KAAK,EAAE,GAAG,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAGxD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAG3C,OAAO,EAAU,IAAI,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,EAAE,KAAK,aAAa,EAAwB,MAAM,qBAAqB,CAAC;AAC/E,OAAO,KAAK,EAAE,YAAY,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC1E,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAC3C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACvD,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,+BAA+B,CAAC;AACzE,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,gCAAgC,CAAC;AAG1E,eAAO,MAAM,mBAAmB;;EAA2D,CAAC;AAC5F,eAAO,MAAM,mBAAmB,iCAAwC,CAAC;AAGzE,KAAK,iBAAiB,GAAG;IACvB,YAAY,EAAE,YAAY,GAAG,IAAI,CAAC;IAClC,YAAY,EAAE,YAAY,GAAG,IAAI,CAAC;IAClC,UAAU,EAAE,OAAO,CAAC;CACrB,CAAC;AAEF,gBAAgB;AAChB,wBAAgB,oBAAoB,CAAC,CAAC,EACpC,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC,EACb,OAAO,EAAE;IACP,YAAY,CAAC,EAAE,YAAY,GAAG,IAAI,CAAC;IACnC,YAAY,CAAC,EAAE,YAAY,GAAG,IAAI,CAAC;IACnC,UAAU,CAAC,EAAE,OAAO,CAAC;CACtB,GACA,IAAI,CAkBN;AAED,gBAAgB;AAChB,wBAAgB,oBAAoB,CAAC,CAAC,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC,GAAG,iBAAiB,GAAG,SAAS,CAEpF;AACD,eAAO,MAAM,oBAAoB,eAAyB,CAAC;AAE3D,qBAAa,YAAa,SAAQ,KAAK;;CAStC;AAED,wBAAgB,cAAc,CAAC,KAAK,EAAE,OAAO,GAAG,KAAK,IAAI,YAAY,CAOpE;AAED,MAAM,WAAW,aAAa;IAC5B,mDAAmD;IACnD,UAAU,CAAC,EAAE,UAAU,CAAC;CACzB;AAED,MAAM,WAAW,YAAY,CAAC,QAAQ;IACpC,EAAE,CAAC,EAAE,MAAM,CAAC;IACZ,YAAY,EAAE,MAAM,CAAC;IACrB,OAAO,CAAC,EAAE,WAAW,CAAC;IACtB,KAAK,CAAC,EAAE,WAAW,CAAC,QAAQ,CAAC,CAAC;IAC9B,2CAA2C;IAC3C,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,GAAG,SAAS,CAAC;IACtC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,kBAAkB,CAAC,EAAE,OAAO,CAAC;IAC7B,yBAAyB,CAAC,EAAE,MAAM,CAAC;IACnC,YAAY,CAAC,EAAE,mBAAmB,CAAC;IACnC,uBAAuB,CAAC,EAAE,OAAO,CAAC;CACnC;AAED,qBAAa,KAAK,CAAC,QAAQ,GAAG,GAAG;IAC/B,OAAO,CAAC,GAAG,CAAS;IACpB,OAAO,CAAC,IAAI,CAAC,CAAM;IACnB,OAAO,CAAC,IAAI,CAAC,CAAM;IACnB,OAAO,CAAC,IAAI,CAAC,CAAsB;IACnC,OAAO,CAAC,IAAI,CAAC,CAAM;IACnB,OAAO,CAAC,YAAY,CAAC,CAAsB;IAC3C,OAAO,CAAC,sBAAsB,CAA8B;IAC5D,OAAO,CAAC,mBAAmB,CAAC,CAAU;IACtC,OAAO,CAAC,wBAAwB,CAAC,CAAU;IAE3C,gBAAgB;IAChB,cAAc,CAAC,EAAE,aAAa,CAAC;IAE/B,gBAAgB;IAChB,QAAQ,EAAE,WAAW,CAAC;IAEtB,gBAAgB;IAChB,aAAa,EAAE,MAAM,CAAC;IAEtB,gBAAgB;IAChB,MAAM,CAAC,EAAE,WAAW,CAAC,QAAQ,CAAC,CAAC;gBAEnB,EACV,EAAE,EACF,YAAY,EACZ,OAAO,EACP,KAAK,EACL,aAAa,EACb,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,YAAY,EACZ,uBAAuB,EACvB,kBAAkB,GACnB,EAAE,YAAY,CAAC,QAAQ,CAAC;IA2DzB,IAAI,GAAG,IAAI,GAAG,GAAG,SAAS,CAEzB;IAED,IAAI,GAAG,IAAI,GAAG,GAAG,SAAS,CAEzB;IAED,IAAI,GAAG,IAAI,GAAG,GAAG,aAAa,GAAG,SAAS,CAEzC;IAED,IAAI,GAAG,IAAI,GAAG,GAAG,SAAS,CAEzB;IAED,IAAI,uBAAuB,IAAI,OAAO,GAAG,SAAS,CAEjD;IAED,IAAI,OAAO,IAAI,mBAAmB,CAEjC;IAED,IAAI,EAAE,IAAI,MAAM,CAEf;IAED,IAAI,YAAY,IAAI,MAAM,CAEzB;IAED,IAAI,OAAO,IAAI,WAAW,CAAC,QAAQ,CAAC,CAEnC;IAED,IAAI,OAAO,IAAI,YAAY,CAAC,QAAQ,CAAC,CAEpC;IAED,IAAI,qBAAqB,IAAI,mBAAmB,CAAC,MAAM,CAAC,CAEvD;IAED,IAAI,kBAAkB,IAAI,OAAO,GAAG,SAAS,CAE5C;IAEK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAExB,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvB,iBAAiB,CACrB,IAAI,EAAE,cAAc,CAAC,MAAM,GAAG,WAAW,CAAC,EAC1C,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,MAAM,GAAG,WAAW,CAAC,GAAG,IAAI,CAAC;IAIjD,mBAAmB,CAAC,QAAQ,EAAE,WAAW,EAAE,WAAW,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAEnF,OAAO,CACX,KAAK,EAAE,cAAc,CAAC,UAAU,CAAC,EACjC,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,WAAW,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;IAIjD,OAAO,CACX,OAAO,EAAE,WAAW,EACpB,OAAO,EAAE,WAAW,EACpB,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;IAI/C,OAAO,CACX,IAAI,EAAE,cAAc,CAAC,MAAM,CAAC,EAC5B,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;IAIvC,uBAAuB,CAC3B,KAAK,EAAE,cAAc,CAAC,UAAU,CAAC,EACjC,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;IAM7C,kBAAkB,IAAI,aAAa;IAO7B,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAUlD,WAAW,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAUpD,MAAM,CAAC,OAAO;uBAEH,KAAK,SACL,eAAe,UAAU,CAAC,kBACjB,aAAa,GAC5B,QAAQ,eAAe,WAAW,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;uBA2D9C,KAAK,WACH,WAAW,WACX,WAAW,iBACL,aAAa,GAC3B,QAAQ,eAAe,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;uBAiD5C,KAAK,QACN,eAAe,MAAM,CAAC,kBACZ,aAAa,GAC5B,QAAQ,eAAe,UAAU,CAAC,GAAG,IAAI,CAAC;iCAgDpC,KAAK,QACN,eAAe,MAAM,GAAG,WAAW,CAAC,kBAC1B,aAAa,GAC5B,QAAQ,eAAe,MAAM,GAAG,WAAW,CAAC,GAAG,IAAI,CAAC;wCAK7C,KAAK,SACN,eAAe,UAAU,CAAC,kBACjB,aAAa,GAC5B,QAAQ,eAAe,UAAU,CAAC,GAAG,IAAI,CAAC;MAG7C;CACH;AAED,qBAAa,SAAS,CAAC,OAAO,GAAG,OAAO,EAAE,QAAQ,GAAG,GAAG,CAAE,SAAQ,KAAK,CAAC,QAAQ,CAAC;;IAC/E,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,MAAM,CAAyB;IAIvC,IAAI,IAAI,IAAI,OAAO,CAElB;IAED,QAAQ,CAAC,MAAM,EAAE,OAAO,GAAG,KAAK,GAAG,IAAI;IAiBjC,GAAG,IAAI,OAAO,CAAC,OAAO,CAAC;CAyG9B"}
|