@livekit/agents 1.0.15 → 1.0.17
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/cli.cjs +12 -12
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.d.cts +3 -3
- package/dist/cli.d.ts +3 -3
- package/dist/cli.d.ts.map +1 -1
- package/dist/cli.js +13 -13
- package/dist/cli.js.map +1 -1
- package/dist/inference/stt.cjs.map +1 -1
- package/dist/inference/stt.d.ts.map +1 -1
- package/dist/inference/stt.js +1 -1
- package/dist/inference/stt.js.map +1 -1
- package/dist/inference/tts.cjs.map +1 -1
- package/dist/inference/tts.d.cts +2 -1
- package/dist/inference/tts.d.ts +2 -1
- package/dist/inference/tts.d.ts.map +1 -1
- package/dist/inference/tts.js +1 -5
- package/dist/inference/tts.js.map +1 -1
- package/dist/llm/chat_context.cjs +78 -0
- package/dist/llm/chat_context.cjs.map +1 -1
- package/dist/llm/chat_context.d.cts +16 -0
- package/dist/llm/chat_context.d.ts +16 -0
- package/dist/llm/chat_context.d.ts.map +1 -1
- package/dist/llm/chat_context.js +78 -0
- package/dist/llm/chat_context.js.map +1 -1
- package/dist/llm/chat_context.test.cjs +531 -0
- package/dist/llm/chat_context.test.cjs.map +1 -1
- package/dist/llm/chat_context.test.js +531 -0
- package/dist/llm/chat_context.test.js.map +1 -1
- package/dist/llm/tool_context.cjs +40 -0
- package/dist/llm/tool_context.cjs.map +1 -1
- package/dist/llm/tool_context.d.cts +2 -0
- package/dist/llm/tool_context.d.ts +2 -0
- package/dist/llm/tool_context.d.ts.map +1 -1
- package/dist/llm/tool_context.js +38 -0
- package/dist/llm/tool_context.js.map +1 -1
- package/dist/metrics/base.cjs.map +1 -1
- package/dist/metrics/base.d.cts +7 -0
- package/dist/metrics/base.d.ts +7 -0
- package/dist/metrics/base.d.ts.map +1 -1
- package/dist/stt/stt.cjs +1 -1
- package/dist/stt/stt.cjs.map +1 -1
- package/dist/stt/stt.d.cts +7 -1
- package/dist/stt/stt.d.ts +7 -1
- package/dist/stt/stt.d.ts.map +1 -1
- package/dist/stt/stt.js +1 -1
- package/dist/stt/stt.js.map +1 -1
- package/dist/tts/tts.cjs +2 -4
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +3 -5
- package/dist/tts/tts.js.map +1 -1
- package/dist/voice/agent_activity.cjs +83 -8
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.cts +6 -2
- package/dist/voice/agent_activity.d.ts +6 -2
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +83 -8
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/agent_session.cjs +3 -2
- package/dist/voice/agent_session.cjs.map +1 -1
- package/dist/voice/agent_session.d.cts +2 -1
- package/dist/voice/agent_session.d.ts +2 -1
- package/dist/voice/agent_session.d.ts.map +1 -1
- package/dist/voice/agent_session.js +3 -2
- package/dist/voice/agent_session.js.map +1 -1
- package/dist/voice/audio_recognition.cjs +138 -16
- package/dist/voice/audio_recognition.cjs.map +1 -1
- package/dist/voice/audio_recognition.d.cts +11 -0
- package/dist/voice/audio_recognition.d.ts +11 -0
- package/dist/voice/audio_recognition.d.ts.map +1 -1
- package/dist/voice/audio_recognition.js +138 -16
- package/dist/voice/audio_recognition.js.map +1 -1
- package/dist/voice/room_io/_input.cjs.map +1 -1
- package/dist/voice/room_io/_input.d.ts.map +1 -1
- package/dist/voice/room_io/_input.js +0 -1
- package/dist/voice/room_io/_input.js.map +1 -1
- package/dist/worker.cjs +17 -11
- package/dist/worker.cjs.map +1 -1
- package/dist/worker.d.cts +16 -9
- package/dist/worker.d.ts +16 -9
- package/dist/worker.d.ts.map +1 -1
- package/dist/worker.js +16 -12
- package/dist/worker.js.map +1 -1
- package/package.json +1 -1
- package/src/cli.ts +17 -17
- package/src/inference/stt.ts +2 -1
- package/src/inference/tts.ts +2 -5
- package/src/llm/chat_context.test.ts +607 -0
- package/src/llm/chat_context.ts +106 -0
- package/src/llm/tool_context.ts +44 -0
- package/src/metrics/base.ts +7 -0
- package/src/stt/stt.ts +8 -1
- package/src/tts/tts.ts +7 -5
- package/src/voice/agent_activity.ts +119 -9
- package/src/voice/agent_session.ts +3 -1
- package/src/voice/audio_recognition.ts +235 -57
- package/src/voice/room_io/_input.ts +1 -1
- package/src/worker.ts +29 -18
|
@@ -25,6 +25,7 @@ export interface VoiceOptions {
|
|
|
25
25
|
minEndpointingDelay: number;
|
|
26
26
|
maxEndpointingDelay: number;
|
|
27
27
|
maxToolSteps: number;
|
|
28
|
+
preemptiveGeneration: boolean;
|
|
28
29
|
}
|
|
29
30
|
export type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;
|
|
30
31
|
export type AgentSessionCallbacks = {
|
|
@@ -113,7 +114,7 @@ export declare class AgentSession<UserData = UnknownUserData> extends AgentSessi
|
|
|
113
114
|
/** @internal */
|
|
114
115
|
_updateAgentState(state: AgentState): void;
|
|
115
116
|
/** @internal */
|
|
116
|
-
_updateUserState(state: UserState): void;
|
|
117
|
+
_updateUserState(state: UserState, _lastSpeakingTime?: number): void;
|
|
117
118
|
private onAudioInputChanged;
|
|
118
119
|
private onAudioOutputChanged;
|
|
119
120
|
private onTextOutputChanged;
|
|
@@ -25,6 +25,7 @@ export interface VoiceOptions {
|
|
|
25
25
|
minEndpointingDelay: number;
|
|
26
26
|
maxEndpointingDelay: number;
|
|
27
27
|
maxToolSteps: number;
|
|
28
|
+
preemptiveGeneration: boolean;
|
|
28
29
|
}
|
|
29
30
|
export type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;
|
|
30
31
|
export type AgentSessionCallbacks = {
|
|
@@ -113,7 +114,7 @@ export declare class AgentSession<UserData = UnknownUserData> extends AgentSessi
|
|
|
113
114
|
/** @internal */
|
|
114
115
|
_updateAgentState(state: AgentState): void;
|
|
115
116
|
/** @internal */
|
|
116
|
-
_updateUserState(state: UserState): void;
|
|
117
|
+
_updateUserState(state: UserState, _lastSpeakingTime?: number): void;
|
|
117
118
|
private onAudioInputChanged;
|
|
118
119
|
private onAudioOutputChanged;
|
|
119
120
|
private onTextOutputChanged;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent_session.d.ts","sourceRoot":"","sources":["../../src/voice/agent_session.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAC1D,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAE/B,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,KAAK,EAAE,GAAG,EAAE,aAAa,EAAE,kBAAkB,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC1F,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAE9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACnD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AAExC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AAC5D,OAAO,EACL,sBAAsB,EACtB,KAAK,UAAU,EACf,KAAK,sBAAsB,EAC3B,KAAK,UAAU,EACf,WAAW,EACX,KAAK,0BAA0B,EAC/B,KAAK,UAAU,EACf,KAAK,0BAA0B,EAC/B,KAAK,qBAAqB,EAC1B,KAAK,kBAAkB,EACvB,KAAK,yBAAyB,EAC9B,KAAK,SAAS,EACd,KAAK,qBAAqB,EAK3B,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAClD,OAAO,EAAU,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC3F,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,WAAW,YAAY;IAC3B,kBAAkB,EAAE,OAAO,CAAC;IAC5B,6BAA6B,EAAE,OAAO,CAAC;IACvC,uBAAuB,EAAE,MAAM,CAAC;IAChC,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,YAAY,EAAE,MAAM,CAAC;
|
|
1
|
+
{"version":3,"file":"agent_session.d.ts","sourceRoot":"","sources":["../../src/voice/agent_session.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAC1D,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAE/B,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,KAAK,EAAE,GAAG,EAAE,aAAa,EAAE,kBAAkB,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC1F,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAE9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACnD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AAExC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AAC5D,OAAO,EACL,sBAAsB,EACtB,KAAK,UAAU,EACf,KAAK,sBAAsB,EAC3B,KAAK,UAAU,EACf,WAAW,EACX,KAAK,0BAA0B,EAC/B,KAAK,UAAU,EACf,KAAK,0BAA0B,EAC/B,KAAK,qBAAqB,EAC1B,KAAK,kBAAkB,EACvB,KAAK,yBAAyB,EAC9B,KAAK,SAAS,EACd,KAAK,qBAAqB,EAK3B,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAClD,OAAO,EAAU,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC3F,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,WAAW,YAAY;IAC3B,kBAAkB,EAAE,OAAO,CAAC;IAC5B,6BAA6B,EAAE,OAAO,CAAC;IACvC,uBAAuB,EAAE,MAAM,CAAC;IAChC,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,YAAY,EAAE,MAAM,CAAC;IACrB,oBAAoB,EAAE,OAAO,CAAC;CAC/B;AAaD,MAAM,MAAM,iBAAiB,GAAG,KAAK,GAAG,KAAK,GAAG,cAAc,GAAG,QAAQ,GAAG,aAAa,CAAC;AAE1F,MAAM,MAAM,qBAAqB,GAAG;IAClC,CAAC,sBAAsB,CAAC,oBAAoB,CAAC,EAAE,CAAC,EAAE,EAAE,yBAAyB,KAAK,IAAI,CAAC;IACvF,CAAC,sBAAsB,CAAC,iBAAiB,CAAC,EAAE,CAAC,EAAE,EAAE,sBAAsB,KAAK,IAAI,CAAC;IACjF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,aAAa,CAAC,EAAE,CAAC,EAAE,EAAE,kBAAkB,KAAK,IAAI,CAAC;IACzE,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;IACzD,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;CAC1D,CAAC;AAEF,MAAM,MAAM,mBAAmB,CAAC,QAAQ,GAAG,eAAe,IAAI;IAC5D,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,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,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC,YAAY,CAAC,CAAC;CACtC,CAAC;2CAImC,aAAa,qBAAqB,CAAC;AAFxE,qBAAa,YAAY,CACvB,QAAQ,GAAG,eAAe,CAC1B,SAAQ,iBAA+D;IACvE,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAElC,QAAQ,CAAC,OAAO,EAAE,YAAY,CAAC;IAE/B,OAAO,CAAC,KAAK,CAAC,CAAQ;IACtB,OAAO,CAAC,QAAQ,CAAC,CAAgB;IACjC,OAAO,CAAC,YAAY,CAAC,CAAgB;IACrC,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,MAAM,CAAC,CAAS;IACxB,OAAO,CAAC,MAAM,CAAS;IAEvB,OAAO,CAAC,QAAQ,CAAc;IAC9B,OAAO,CAAC,SAAS,CAAuB;IACxC,OAAO,CAAC,WAAW,CAA8B;IAEjD,OAAO,CAAC,MAAM,CAAa;IAC3B,OAAO,CAAC,OAAO,CAAc;IAE7B,OAAO,CAAC,WAAW,CAA8B;gBAErC,IAAI,EAAE,mBAAmB,CAAC,QAAQ,CAAC;IA6C/C,IAAI,KAAK,IAAI,UAAU,CAEtB;IAED,IAAI,MAAM,IAAI,WAAW,CAExB;IAED,IAAI,QAAQ,IAAI,QAAQ,CAMvB;IAED,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,QAAQ,CAAC,KAAK,EAAE,QAAQ,EAE3B;IAEK,KAAK,CAAC,EACV,KAAK,EACL,IAAI,EACJ,YAAY,EACZ,aAAa,GACd,EAAE;QACD,KAAK,EAAE,KAAK,CAAC;QACb,IAAI,EAAE,IAAI,CAAC;QACX,YAAY,CAAC,EAAE,OAAO,CAAC,gBAAgB,CAAC,CAAC;QACzC,aAAa,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;KAC5C,GAAG,OAAO,CAAC,IAAI,CAAC;IAyDjB,WAAW,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI;IAQ/B,cAAc;IAQd,aAAa;IAOb,GAAG,CACD,IAAI,EAAE,MAAM,GAAG,cAAc,CAAC,MAAM,CAAC,EACrC,OAAO,CAAC,EAAE;QACR,KAAK,CAAC,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,YAAY,CAAC,EAAE,OAAO,CAAC;KACxB,GACA,YAAY;IAQf,SAAS;IAOT,aAAa,CAAC,OAAO,CAAC,EAAE;QACtB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,MAAM,CAAC;QACtB,UAAU,CAAC,EAAE,UAAU,CAAC;QACxB,kBAAkB,CAAC,EAAE,OAAO,CAAC;KAC9B,GAAG,YAAY;YAsBF,cAAc;IAmB5B,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,UAAU,IAAI,UAAU,CAE3B;IAED,IAAI,YAAY,IAAI,KAAK,CAMxB;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5B,gBAAgB;IAChB,UAAU,CAAC,EACT,MAAM,EACN,KAAa,EACb,KAAY,GACb,EAAE;QACD,MAAM,EAAE,WAAW,CAAC;QACpB,KAAK,CAAC,EAAE,OAAO,CAAC;QAChB,KAAK,CAAC,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI,CAAC;KACpE,GAAG,IAAI;IAOR,gBAAgB;IAChB,QAAQ,CAAC,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI;IAc1E,gBAAgB;IAChB,sBAAsB,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI;IAK/C,gBAAgB;IAChB,iBAAiB,CAAC,KAAK,EAAE,UAAU;IAanC,gBAAgB;IAChB,gBAAgB,CAAC,KAAK,EAAE,SAAS,EAAE,iBAAiB,CAAC,EAAE,MAAM;IAc7D,OAAO,CAAC,mBAAmB;IAU3B,OAAO,CAAC,oBAAoB;IAE5B,OAAO,CAAC,mBAAmB;YAEb,SAAS;CA6CxB"}
|
|
@@ -25,7 +25,8 @@ const defaultVoiceOptions = {
|
|
|
25
25
|
minInterruptionWords: 0,
|
|
26
26
|
minEndpointingDelay: 500,
|
|
27
27
|
maxEndpointingDelay: 6e3,
|
|
28
|
-
maxToolSteps: 3
|
|
28
|
+
maxToolSteps: 3,
|
|
29
|
+
preemptiveGeneration: false
|
|
29
30
|
};
|
|
30
31
|
class AgentSession extends EventEmitter {
|
|
31
32
|
vad;
|
|
@@ -262,7 +263,7 @@ class AgentSession extends EventEmitter {
|
|
|
262
263
|
);
|
|
263
264
|
}
|
|
264
265
|
/** @internal */
|
|
265
|
-
_updateUserState(state) {
|
|
266
|
+
_updateUserState(state, _lastSpeakingTime) {
|
|
266
267
|
if (this.userState === state) {
|
|
267
268
|
return;
|
|
268
269
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/voice/agent_session.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, Room } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport {\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 { getJobContext } from '../job.js';\nimport { ChatContext, ChatMessage } from '../llm/chat_context.js';\nimport type { LLM, RealtimeModel, RealtimeModelError, ToolChoice } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport { log } from '../log.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS, TTSError } from '../tts/tts.js';\nimport type { VAD } from '../vad.js';\nimport type { Agent } from './agent.js';\nimport { AgentActivity } from './agent_activity.js';\nimport type { _TurnDetector } from './audio_recognition.js';\nimport {\n AgentSessionEventTypes,\n type AgentState,\n type AgentStateChangedEvent,\n type CloseEvent,\n CloseReason,\n type ConversationItemAddedEvent,\n type ErrorEvent,\n type FunctionToolsExecutedEvent,\n type MetricsCollectedEvent,\n type SpeechCreatedEvent,\n type UserInputTranscribedEvent,\n type UserState,\n type UserStateChangedEvent,\n createAgentStateChangedEvent,\n createCloseEvent,\n createConversationItemAddedEvent,\n createUserStateChangedEvent,\n} from './events.js';\nimport { AgentInput, AgentOutput } from './io.js';\nimport { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport interface VoiceOptions {\n allowInterruptions: boolean;\n discardAudioIfUninterruptible: boolean;\n minInterruptionDuration: number;\n minInterruptionWords: number;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n maxToolSteps: number;\n}\n\nconst defaultVoiceOptions: VoiceOptions = {\n allowInterruptions: true,\n discardAudioIfUninterruptible: true,\n minInterruptionDuration: 500,\n minInterruptionWords: 0,\n minEndpointingDelay: 500,\n maxEndpointingDelay: 6000,\n maxToolSteps: 3,\n} as const;\n\nexport type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;\n\nexport type AgentSessionCallbacks = {\n [AgentSessionEventTypes.UserInputTranscribed]: (ev: UserInputTranscribedEvent) => void;\n [AgentSessionEventTypes.AgentStateChanged]: (ev: AgentStateChangedEvent) => void;\n [AgentSessionEventTypes.UserStateChanged]: (ev: UserStateChangedEvent) => void;\n [AgentSessionEventTypes.ConversationItemAdded]: (ev: ConversationItemAddedEvent) => void;\n [AgentSessionEventTypes.FunctionToolsExecuted]: (ev: FunctionToolsExecutedEvent) => void;\n [AgentSessionEventTypes.MetricsCollected]: (ev: MetricsCollectedEvent) => void;\n [AgentSessionEventTypes.SpeechCreated]: (ev: SpeechCreatedEvent) => void;\n [AgentSessionEventTypes.Error]: (ev: ErrorEvent) => void;\n [AgentSessionEventTypes.Close]: (ev: CloseEvent) => void;\n};\n\nexport type AgentSessionOptions<UserData = UnknownUserData> = {\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n userData?: UserData;\n voiceOptions?: Partial<VoiceOptions>;\n};\n\nexport class AgentSession<\n UserData = UnknownUserData,\n> extends (EventEmitter as new () => TypedEmitter<AgentSessionCallbacks>) {\n vad?: VAD;\n stt?: STT;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n turnDetection?: TurnDetectionMode;\n\n readonly options: VoiceOptions;\n\n private agent?: Agent;\n private activity?: AgentActivity;\n private nextActivity?: AgentActivity;\n private started = false;\n private userState: UserState = 'listening';\n\n private roomIO?: RoomIO;\n private logger = log();\n\n private _chatCtx: ChatContext;\n private _userData: UserData | undefined;\n private _agentState: AgentState = 'initializing';\n\n private _input: AgentInput;\n private _output: AgentOutput;\n\n private closingTask: Promise<void> | null = null;\n\n constructor(opts: AgentSessionOptions<UserData>) {\n super();\n\n const {\n vad,\n stt,\n llm,\n tts,\n turnDetection,\n userData,\n voiceOptions = defaultVoiceOptions,\n } = opts;\n\n this.vad = vad;\n\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.turnDetection = turnDetection;\n this._userData = userData;\n\n // configurable IO\n this._input = new AgentInput(this.onAudioInputChanged);\n this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);\n\n // This is the \"global\" chat context, it holds the entire conversation history\n this._chatCtx = ChatContext.empty();\n this.options = { ...defaultVoiceOptions, ...voiceOptions };\n }\n\n get input(): AgentInput {\n return this._input;\n }\n\n get output(): AgentOutput {\n return this._output;\n }\n\n get userData(): UserData {\n if (this._userData === undefined) {\n throw new Error('Voice agent userData is not set');\n }\n\n return this._userData;\n }\n\n get history(): ChatContext {\n return this._chatCtx;\n }\n\n set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\n }\n\n if (this.output.audio && outputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio output is enabled but output.audio is already set, ignoring..',\n );\n }\n\n if (this.output.transcription && outputOptions?.transcriptionEnabled !== false) {\n this.logger.warn(\n 'RoomIO transcription output is enabled but output.transcription is already set, ignoring..',\n );\n }\n\n this.roomIO = new RoomIO({\n agentSession: this,\n room,\n inputOptions,\n outputOptions,\n });\n this.roomIO.start();\n\n const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\n }\n // TODO(AJS-265): add shutdown callback to job context\n tasks.push(this.updateActivity(this.agent));\n\n await Promise.allSettled(tasks);\n\n // Log used IO configuration\n this.logger.debug(\n `using audio io: ${this.input.audio ? '`' + this.input.audio.constructor.name + '`' : '(none)'} -> \\`AgentSession\\` -> ${this.output.audio ? '`' + this.output.audio.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug(\n `using transcript io: \\`AgentSession\\` -> ${this.output.transcription ? '`' + this.output.transcription.constructor.name + '`' : '(none)'}`,\n );\n\n this.started = true;\n this._updateAgentState('listening');\n }\n\n updateAgent(agent: Agent): void {\n this.agent = agent;\n\n if (this.started) {\n this.updateActivity(agent);\n }\n }\n\n commitUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n this.activity.commitUserTurn();\n }\n\n clearUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n this.activity.clearUserTurn();\n }\n\n say(\n text: string | ReadableStream<string>,\n options?: {\n audio?: ReadableStream<AudioFrame>;\n allowInterruptions?: boolean;\n addToChatCtx?: boolean;\n },\n ): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n return this.activity.say(text, options);\n }\n\n interrupt() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n return this.activity.interrupt();\n }\n\n generateReply(options?: {\n userInput?: string;\n instructions?: string;\n toolChoice?: ToolChoice;\n allowInterruptions?: boolean;\n }): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const userMessage = options?.userInput\n ? new ChatMessage({\n role: 'user',\n content: options.userInput,\n })\n : undefined;\n\n if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\n\n if (this.activity) {\n await this.activity.drain();\n await this.activity.close();\n }\n\n this.activity = this.nextActivity;\n this.nextActivity = undefined;\n\n await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n get chatCtx(): ChatContext {\n return this._chatCtx.copy();\n }\n\n get agentState(): AgentState {\n return this._agentState;\n }\n\n get currentAgent(): Agent {\n if (!this.agent) {\n throw new Error('AgentSession is not running');\n }\n\n return this.agent;\n }\n\n async close(): Promise<void> {\n await this.closeImpl(CloseReason.USER_INITIATED);\n }\n\n /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\n drain?: boolean;\n error?: RealtimeModelError | STTError | TTSError | LLMError | null;\n }): void {\n if (this.closingTask) {\n return;\n }\n this.closeImpl(reason, error, drain);\n }\n\n /** @internal */\n _onError(error: RealtimeModelError | STTError | TTSError | LLMError): void {\n if (this.closingTask || error.recoverable) {\n return;\n }\n\n this.logger.error(error, 'AgentSession is closing due to unrecoverable error');\n\n this.closingTask = (async () => {\n await this.closeImpl(CloseReason.ERROR, error);\n })().then(() => {\n this.closingTask = null;\n });\n }\n\n /** @internal */\n _conversationItemAdded(item: ChatMessage): void {\n this._chatCtx.insert(item);\n this.emit(AgentSessionEventTypes.ConversationItemAdded, createConversationItemAddedEvent(item));\n }\n\n /** @internal */\n _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\n this.emit(\n AgentSessionEventTypes.AgentStateChanged,\n createAgentStateChangedEvent(oldState, state),\n );\n }\n\n /** @internal */\n _updateUserState(state: UserState) {\n if (this.userState === state) {\n return;\n }\n\n const oldState = this.userState;\n this.userState = state;\n this.emit(\n AgentSessionEventTypes.UserStateChanged,\n createUserStateChangedEvent(oldState, state),\n );\n }\n\n // -- User changed input/output streams/sinks --\n private onAudioInputChanged(): void {\n if (!this.started) {\n return;\n }\n\n if (this.activity && this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n private onAudioOutputChanged(): void {}\n\n private onTextOutputChanged(): void {}\n\n private async closeImpl(\n reason: CloseReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\n // TODO(shubhra): force interrupt or wait for it to finish?\n // it might be an audio played from the error callback\n }\n }\n await this.activity.drain();\n // wait any uninterruptible speech to finish\n await this.activity.currentSpeech?.waitForPlayout();\n this.activity.detachAudioInput();\n }\n\n // detach the inputs and outputs\n this.input.audio = null;\n this.output.audio = null;\n this.output.transcription = null;\n\n await this.roomIO?.close();\n this.roomIO = undefined;\n\n await this.activity?.close();\n this.activity = undefined;\n\n this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,qBAAqB;AAC9B,SAAS,aAAa,mBAAmB;AAGzC,SAAS,WAAW;AAMpB,SAAS,qBAAqB;AAE9B;AAAA,EACE;AAAA,EAIA;AAAA,EASA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,cAA6D;AActE,MAAM,sBAAoC;AAAA,EACxC,oBAAoB;AAAA,EACpB,+BAA+B;AAAA,EAC/B,yBAAyB;AAAA,EACzB,sBAAsB;AAAA,EACtB,qBAAqB;AAAA,EACrB,qBAAqB;AAAA,EACrB,cAAc;AAChB;AA0BO,MAAM,qBAEF,aAA+D;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAES;AAAA,EAED;AAAA,EACA;AAAA,EACA;AAAA,EACA,UAAU;AAAA,EACV,YAAuB;AAAA,EAEvB;AAAA,EACA,SAAS,IAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EAE5C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,SAAK,MAAM;AAEX,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,WAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,YAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,YAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAAA,EAC3D;AAAA,EAEA,IAAI,QAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,WAAqB;AACvB,QAAI,KAAK,cAAc,QAAW;AAChC,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,MAAM,cAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,UAAM,KAAK,KAAK,eAAe,KAAK,KAAK,CAAC;AAE1C,UAAM,QAAQ,WAAW,KAAK;AAG9B,SAAK,OAAO;AAAA,MACV,mBAAmB,KAAK,MAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,YAAY,OAAO,MAAM,QAAQ,2BAA2B,KAAK,OAAO,QAAQ,MAAM,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,QAAQ;AAAA,IACxM;AAEA,SAAK,OAAO;AAAA,MACV,4CAA4C,KAAK,OAAO,gBAAgB,MAAM,KAAK,OAAO,cAAc,YAAY,OAAO,MAAM,QAAQ;AAAA,IAC3I;AAEA,SAAK,UAAU;AACf,SAAK,kBAAkB,WAAW;AAAA,EACpC;AAAA,EAEA,YAAY,OAAoB;AAC9B,SAAK,QAAQ;AAEb,QAAI,KAAK,SAAS;AAChB,WAAK,eAAe,KAAK;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,iBAAiB;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,SAAK,SAAS,eAAe;AAAA,EAC/B;AAAA,EAEA,gBAAgB;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,SAAK,SAAS,cAAc;AAAA,EAC9B;AAAA,EAEA,IACE,MACA,SAKc;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;AAAA,EAEA,YAAY;AACV,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EAEA,cAAc,SAKG;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,eAAc,mCAAS,aACzB,IAAI,YAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAsB;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAuB;AAC3B,UAAM,KAAK,UAAU,YAAY,cAAc;AAAA,EACjD;AAAA;AAAA,EAGA,WAAW;AAAA,IACT;AAAA,IACA,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV,GAIS;AACP,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AACA,SAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,OAAkE;AACzE,QAAI,KAAK,eAAe,MAAM,aAAa;AACzC;AAAA,IACF;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,YAAY,OAAO,KAAK;AAAA,IAC/C,GAAG,EAAE,KAAK,MAAM;AACd,WAAK,cAAc;AAAA,IACrB,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,uBAAuB,MAAyB;AAC9C,SAAK,SAAS,OAAO,IAAI;AACzB,SAAK,KAAK,uBAAuB,uBAAuB,iCAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB;AACjC,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AACjB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,4BAA4B,UAAU,KAAK;AAAA,IAC7C;AAAA,EACF;AAAA;AAAA,EAGQ,sBAA4B;AAClC,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,YAAY,KAAK,OAAO,OAAO;AACtC,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEQ,uBAA6B;AAAA,EAAC;AAAA,EAE9B,sBAA4B;AAAA,EAAC;AAAA,EAErC,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AAvcnB;AAwcI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;AAGA,SAAK,MAAM,QAAQ;AACnB,SAAK,OAAO,QAAQ;AACpB,SAAK,OAAO,gBAAgB;AAE5B,YAAM,UAAK,WAAL,mBAAa;AACnB,SAAK,SAAS;AAEd,YAAM,UAAK,aAAL,mBAAe;AACrB,SAAK,WAAW;AAEhB,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
|
|
1
|
+
{"version":3,"sources":["../../src/voice/agent_session.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, Room } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport {\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 { getJobContext } from '../job.js';\nimport { ChatContext, ChatMessage } from '../llm/chat_context.js';\nimport type { LLM, RealtimeModel, RealtimeModelError, ToolChoice } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport { log } from '../log.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS, TTSError } from '../tts/tts.js';\nimport type { VAD } from '../vad.js';\nimport type { Agent } from './agent.js';\nimport { AgentActivity } from './agent_activity.js';\nimport type { _TurnDetector } from './audio_recognition.js';\nimport {\n AgentSessionEventTypes,\n type AgentState,\n type AgentStateChangedEvent,\n type CloseEvent,\n CloseReason,\n type ConversationItemAddedEvent,\n type ErrorEvent,\n type FunctionToolsExecutedEvent,\n type MetricsCollectedEvent,\n type SpeechCreatedEvent,\n type UserInputTranscribedEvent,\n type UserState,\n type UserStateChangedEvent,\n createAgentStateChangedEvent,\n createCloseEvent,\n createConversationItemAddedEvent,\n createUserStateChangedEvent,\n} from './events.js';\nimport { AgentInput, AgentOutput } from './io.js';\nimport { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport interface VoiceOptions {\n allowInterruptions: boolean;\n discardAudioIfUninterruptible: boolean;\n minInterruptionDuration: number;\n minInterruptionWords: number;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n maxToolSteps: number;\n preemptiveGeneration: boolean;\n}\n\nconst defaultVoiceOptions: VoiceOptions = {\n allowInterruptions: true,\n discardAudioIfUninterruptible: true,\n minInterruptionDuration: 500,\n minInterruptionWords: 0,\n minEndpointingDelay: 500,\n maxEndpointingDelay: 6000,\n maxToolSteps: 3,\n preemptiveGeneration: false,\n} as const;\n\nexport type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;\n\nexport type AgentSessionCallbacks = {\n [AgentSessionEventTypes.UserInputTranscribed]: (ev: UserInputTranscribedEvent) => void;\n [AgentSessionEventTypes.AgentStateChanged]: (ev: AgentStateChangedEvent) => void;\n [AgentSessionEventTypes.UserStateChanged]: (ev: UserStateChangedEvent) => void;\n [AgentSessionEventTypes.ConversationItemAdded]: (ev: ConversationItemAddedEvent) => void;\n [AgentSessionEventTypes.FunctionToolsExecuted]: (ev: FunctionToolsExecutedEvent) => void;\n [AgentSessionEventTypes.MetricsCollected]: (ev: MetricsCollectedEvent) => void;\n [AgentSessionEventTypes.SpeechCreated]: (ev: SpeechCreatedEvent) => void;\n [AgentSessionEventTypes.Error]: (ev: ErrorEvent) => void;\n [AgentSessionEventTypes.Close]: (ev: CloseEvent) => void;\n};\n\nexport type AgentSessionOptions<UserData = UnknownUserData> = {\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n userData?: UserData;\n voiceOptions?: Partial<VoiceOptions>;\n};\n\nexport class AgentSession<\n UserData = UnknownUserData,\n> extends (EventEmitter as new () => TypedEmitter<AgentSessionCallbacks>) {\n vad?: VAD;\n stt?: STT;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n turnDetection?: TurnDetectionMode;\n\n readonly options: VoiceOptions;\n\n private agent?: Agent;\n private activity?: AgentActivity;\n private nextActivity?: AgentActivity;\n private started = false;\n private userState: UserState = 'listening';\n\n private roomIO?: RoomIO;\n private logger = log();\n\n private _chatCtx: ChatContext;\n private _userData: UserData | undefined;\n private _agentState: AgentState = 'initializing';\n\n private _input: AgentInput;\n private _output: AgentOutput;\n\n private closingTask: Promise<void> | null = null;\n\n constructor(opts: AgentSessionOptions<UserData>) {\n super();\n\n const {\n vad,\n stt,\n llm,\n tts,\n turnDetection,\n userData,\n voiceOptions = defaultVoiceOptions,\n } = opts;\n\n this.vad = vad;\n\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.turnDetection = turnDetection;\n this._userData = userData;\n\n // configurable IO\n this._input = new AgentInput(this.onAudioInputChanged);\n this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);\n\n // This is the \"global\" chat context, it holds the entire conversation history\n this._chatCtx = ChatContext.empty();\n this.options = { ...defaultVoiceOptions, ...voiceOptions };\n }\n\n get input(): AgentInput {\n return this._input;\n }\n\n get output(): AgentOutput {\n return this._output;\n }\n\n get userData(): UserData {\n if (this._userData === undefined) {\n throw new Error('Voice agent userData is not set');\n }\n\n return this._userData;\n }\n\n get history(): ChatContext {\n return this._chatCtx;\n }\n\n set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\n }\n\n if (this.output.audio && outputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio output is enabled but output.audio is already set, ignoring..',\n );\n }\n\n if (this.output.transcription && outputOptions?.transcriptionEnabled !== false) {\n this.logger.warn(\n 'RoomIO transcription output is enabled but output.transcription is already set, ignoring..',\n );\n }\n\n this.roomIO = new RoomIO({\n agentSession: this,\n room,\n inputOptions,\n outputOptions,\n });\n this.roomIO.start();\n\n const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\n }\n // TODO(AJS-265): add shutdown callback to job context\n tasks.push(this.updateActivity(this.agent));\n\n await Promise.allSettled(tasks);\n\n // Log used IO configuration\n this.logger.debug(\n `using audio io: ${this.input.audio ? '`' + this.input.audio.constructor.name + '`' : '(none)'} -> \\`AgentSession\\` -> ${this.output.audio ? '`' + this.output.audio.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug(\n `using transcript io: \\`AgentSession\\` -> ${this.output.transcription ? '`' + this.output.transcription.constructor.name + '`' : '(none)'}`,\n );\n\n this.started = true;\n this._updateAgentState('listening');\n }\n\n updateAgent(agent: Agent): void {\n this.agent = agent;\n\n if (this.started) {\n this.updateActivity(agent);\n }\n }\n\n commitUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n this.activity.commitUserTurn();\n }\n\n clearUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n this.activity.clearUserTurn();\n }\n\n say(\n text: string | ReadableStream<string>,\n options?: {\n audio?: ReadableStream<AudioFrame>;\n allowInterruptions?: boolean;\n addToChatCtx?: boolean;\n },\n ): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n return this.activity.say(text, options);\n }\n\n interrupt() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n return this.activity.interrupt();\n }\n\n generateReply(options?: {\n userInput?: string;\n instructions?: string;\n toolChoice?: ToolChoice;\n allowInterruptions?: boolean;\n }): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const userMessage = options?.userInput\n ? new ChatMessage({\n role: 'user',\n content: options.userInput,\n })\n : undefined;\n\n if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\n\n if (this.activity) {\n await this.activity.drain();\n await this.activity.close();\n }\n\n this.activity = this.nextActivity;\n this.nextActivity = undefined;\n\n await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n get chatCtx(): ChatContext {\n return this._chatCtx.copy();\n }\n\n get agentState(): AgentState {\n return this._agentState;\n }\n\n get currentAgent(): Agent {\n if (!this.agent) {\n throw new Error('AgentSession is not running');\n }\n\n return this.agent;\n }\n\n async close(): Promise<void> {\n await this.closeImpl(CloseReason.USER_INITIATED);\n }\n\n /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\n drain?: boolean;\n error?: RealtimeModelError | STTError | TTSError | LLMError | null;\n }): void {\n if (this.closingTask) {\n return;\n }\n this.closeImpl(reason, error, drain);\n }\n\n /** @internal */\n _onError(error: RealtimeModelError | STTError | TTSError | LLMError): void {\n if (this.closingTask || error.recoverable) {\n return;\n }\n\n this.logger.error(error, 'AgentSession is closing due to unrecoverable error');\n\n this.closingTask = (async () => {\n await this.closeImpl(CloseReason.ERROR, error);\n })().then(() => {\n this.closingTask = null;\n });\n }\n\n /** @internal */\n _conversationItemAdded(item: ChatMessage): void {\n this._chatCtx.insert(item);\n this.emit(AgentSessionEventTypes.ConversationItemAdded, createConversationItemAddedEvent(item));\n }\n\n /** @internal */\n _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\n this.emit(\n AgentSessionEventTypes.AgentStateChanged,\n createAgentStateChangedEvent(oldState, state),\n );\n }\n\n /** @internal */\n _updateUserState(state: UserState, _lastSpeakingTime?: number) {\n if (this.userState === state) {\n return;\n }\n\n const oldState = this.userState;\n this.userState = state;\n this.emit(\n AgentSessionEventTypes.UserStateChanged,\n createUserStateChangedEvent(oldState, state),\n );\n }\n\n // -- User changed input/output streams/sinks --\n private onAudioInputChanged(): void {\n if (!this.started) {\n return;\n }\n\n if (this.activity && this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n private onAudioOutputChanged(): void {}\n\n private onTextOutputChanged(): void {}\n\n private async closeImpl(\n reason: CloseReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\n // TODO(shubhra): force interrupt or wait for it to finish?\n // it might be an audio played from the error callback\n }\n }\n await this.activity.drain();\n // wait any uninterruptible speech to finish\n await this.activity.currentSpeech?.waitForPlayout();\n this.activity.detachAudioInput();\n }\n\n // detach the inputs and outputs\n this.input.audio = null;\n this.output.audio = null;\n this.output.transcription = null;\n\n await this.roomIO?.close();\n this.roomIO = undefined;\n\n await this.activity?.close();\n this.activity = undefined;\n\n this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,qBAAqB;AAC9B,SAAS,aAAa,mBAAmB;AAGzC,SAAS,WAAW;AAMpB,SAAS,qBAAqB;AAE9B;AAAA,EACE;AAAA,EAIA;AAAA,EASA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,cAA6D;AAetE,MAAM,sBAAoC;AAAA,EACxC,oBAAoB;AAAA,EACpB,+BAA+B;AAAA,EAC/B,yBAAyB;AAAA,EACzB,sBAAsB;AAAA,EACtB,qBAAqB;AAAA,EACrB,qBAAqB;AAAA,EACrB,cAAc;AAAA,EACd,sBAAsB;AACxB;AA0BO,MAAM,qBAEF,aAA+D;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAES;AAAA,EAED;AAAA,EACA;AAAA,EACA;AAAA,EACA,UAAU;AAAA,EACV,YAAuB;AAAA,EAEvB;AAAA,EACA,SAAS,IAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EAE5C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,SAAK,MAAM;AAEX,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,WAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,YAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,YAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAAA,EAC3D;AAAA,EAEA,IAAI,QAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,WAAqB;AACvB,QAAI,KAAK,cAAc,QAAW;AAChC,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,MAAM,cAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,UAAM,KAAK,KAAK,eAAe,KAAK,KAAK,CAAC;AAE1C,UAAM,QAAQ,WAAW,KAAK;AAG9B,SAAK,OAAO;AAAA,MACV,mBAAmB,KAAK,MAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,YAAY,OAAO,MAAM,QAAQ,2BAA2B,KAAK,OAAO,QAAQ,MAAM,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,QAAQ;AAAA,IACxM;AAEA,SAAK,OAAO;AAAA,MACV,4CAA4C,KAAK,OAAO,gBAAgB,MAAM,KAAK,OAAO,cAAc,YAAY,OAAO,MAAM,QAAQ;AAAA,IAC3I;AAEA,SAAK,UAAU;AACf,SAAK,kBAAkB,WAAW;AAAA,EACpC;AAAA,EAEA,YAAY,OAAoB;AAC9B,SAAK,QAAQ;AAEb,QAAI,KAAK,SAAS;AAChB,WAAK,eAAe,KAAK;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,iBAAiB;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,SAAK,SAAS,eAAe;AAAA,EAC/B;AAAA,EAEA,gBAAgB;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,SAAK,SAAS,cAAc;AAAA,EAC9B;AAAA,EAEA,IACE,MACA,SAKc;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;AAAA,EAEA,YAAY;AACV,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EAEA,cAAc,SAKG;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,eAAc,mCAAS,aACzB,IAAI,YAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAsB;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAuB;AAC3B,UAAM,KAAK,UAAU,YAAY,cAAc;AAAA,EACjD;AAAA;AAAA,EAGA,WAAW;AAAA,IACT;AAAA,IACA,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV,GAIS;AACP,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AACA,SAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,OAAkE;AACzE,QAAI,KAAK,eAAe,MAAM,aAAa;AACzC;AAAA,IACF;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,YAAY,OAAO,KAAK;AAAA,IAC/C,GAAG,EAAE,KAAK,MAAM;AACd,WAAK,cAAc;AAAA,IACrB,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,uBAAuB,MAAyB;AAC9C,SAAK,SAAS,OAAO,IAAI;AACzB,SAAK,KAAK,uBAAuB,uBAAuB,iCAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB,mBAA4B;AAC7D,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AACjB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,4BAA4B,UAAU,KAAK;AAAA,IAC7C;AAAA,EACF;AAAA;AAAA,EAGQ,sBAA4B;AAClC,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,YAAY,KAAK,OAAO,OAAO;AACtC,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEQ,uBAA6B;AAAA,EAAC;AAAA,EAE9B,sBAA4B;AAAA,EAAC;AAAA,EAErC,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AAzcnB;AA0cI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;AAGA,SAAK,MAAM,QAAQ;AACnB,SAAK,OAAO,QAAQ;AACpB,SAAK,OAAO,gBAAgB;AAE5B,YAAM,UAAK,WAAL,mBAAa;AACnB,SAAK,SAAS;AAEd,YAAM,UAAK,aAAL,mBAAe;AACrB,SAAK,WAAW;AAEhB,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
|
|
@@ -45,7 +45,10 @@ class AudioRecognition {
|
|
|
45
45
|
lastFinalTranscriptTime = 0;
|
|
46
46
|
audioTranscript = "";
|
|
47
47
|
audioInterimTranscript = "";
|
|
48
|
-
|
|
48
|
+
audioPreflightTranscript = "";
|
|
49
|
+
finalTranscriptConfidence = [];
|
|
50
|
+
lastSpeakingTime;
|
|
51
|
+
speechStartTime;
|
|
49
52
|
userTurnCommitted = false;
|
|
50
53
|
speaking = false;
|
|
51
54
|
sampleRate;
|
|
@@ -93,7 +96,7 @@ class AudioRecognition {
|
|
|
93
96
|
});
|
|
94
97
|
}
|
|
95
98
|
async onSTTEvent(ev) {
|
|
96
|
-
var _a, _b, _c, _d, _e, _f, _g, _h, _i;
|
|
99
|
+
var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r;
|
|
97
100
|
if (this.turnDetectionMode === "manual" && this.userTurnCommitted && (this.bounceEOUTask === void 0 || this.bounceEOUTask.done || ev.type == import_stt.SpeechEventType.INTERIM_TRANSCRIPT)) {
|
|
98
101
|
this.logger.debug(
|
|
99
102
|
{
|
|
@@ -110,7 +113,8 @@ class AudioRecognition {
|
|
|
110
113
|
case import_stt.SpeechEventType.FINAL_TRANSCRIPT:
|
|
111
114
|
this.hooks.onFinalTranscript(ev);
|
|
112
115
|
const transcript = (_c = (_b = ev.alternatives) == null ? void 0 : _b[0]) == null ? void 0 : _c.text;
|
|
113
|
-
|
|
116
|
+
const confidence = ((_e = (_d = ev.alternatives) == null ? void 0 : _d[0]) == null ? void 0 : _e.confidence) ?? 0;
|
|
117
|
+
this.lastLanguage = (_g = (_f = ev.alternatives) == null ? void 0 : _f[0]) == null ? void 0 : _g.language;
|
|
114
118
|
if (!transcript) {
|
|
115
119
|
return;
|
|
116
120
|
}
|
|
@@ -124,26 +128,112 @@ class AudioRecognition {
|
|
|
124
128
|
this.lastFinalTranscriptTime = Date.now();
|
|
125
129
|
this.audioTranscript += ` ${transcript}`;
|
|
126
130
|
this.audioTranscript = this.audioTranscript.trimStart();
|
|
131
|
+
this.finalTranscriptConfidence.push(confidence);
|
|
132
|
+
const transcriptChanged = this.audioTranscript !== this.audioPreflightTranscript;
|
|
127
133
|
this.audioInterimTranscript = "";
|
|
128
|
-
|
|
129
|
-
|
|
130
|
-
|
|
134
|
+
this.audioPreflightTranscript = "";
|
|
135
|
+
if (!this.vad || this.lastSpeakingTime === void 0) {
|
|
136
|
+
this.lastSpeakingTime = Date.now();
|
|
137
|
+
}
|
|
138
|
+
if (this.vadBaseTurnDetection || this.userTurnCommitted) {
|
|
139
|
+
if (transcriptChanged) {
|
|
140
|
+
this.logger.debug(
|
|
141
|
+
{ transcript: this.audioTranscript },
|
|
142
|
+
"triggering preemptive generation (FINAL_TRANSCRIPT)"
|
|
143
|
+
);
|
|
144
|
+
this.hooks.onPreemptiveGeneration({
|
|
145
|
+
newTranscript: this.audioTranscript,
|
|
146
|
+
transcriptConfidence: this.finalTranscriptConfidence.length > 0 ? this.finalTranscriptConfidence.reduce((a, b) => a + b, 0) / this.finalTranscriptConfidence.length : 0
|
|
147
|
+
});
|
|
131
148
|
}
|
|
132
|
-
if (this.
|
|
149
|
+
if (!this.speaking) {
|
|
133
150
|
const chatCtx = this.hooks.retrieveChatCtx();
|
|
134
151
|
this.logger.debug("running EOU detection on stt FINAL_TRANSCRIPT");
|
|
135
152
|
this.runEOUDetection(chatCtx);
|
|
136
153
|
}
|
|
137
154
|
}
|
|
138
155
|
break;
|
|
156
|
+
case import_stt.SpeechEventType.PREFLIGHT_TRANSCRIPT:
|
|
157
|
+
this.hooks.onInterimTranscript(ev);
|
|
158
|
+
const preflightTranscript = ((_i = (_h = ev.alternatives) == null ? void 0 : _h[0]) == null ? void 0 : _i.text) ?? "";
|
|
159
|
+
const preflightConfidence = ((_k = (_j = ev.alternatives) == null ? void 0 : _j[0]) == null ? void 0 : _k.confidence) ?? 0;
|
|
160
|
+
const preflightLanguage = (_m = (_l = ev.alternatives) == null ? void 0 : _l[0]) == null ? void 0 : _m.language;
|
|
161
|
+
const MIN_LANGUAGE_DETECTION_LENGTH = 5;
|
|
162
|
+
if (!this.lastLanguage || preflightLanguage && preflightTranscript.length > MIN_LANGUAGE_DETECTION_LENGTH) {
|
|
163
|
+
this.lastLanguage = preflightLanguage;
|
|
164
|
+
}
|
|
165
|
+
if (!preflightTranscript) {
|
|
166
|
+
return;
|
|
167
|
+
}
|
|
168
|
+
this.logger.debug(
|
|
169
|
+
{
|
|
170
|
+
user_transcript: preflightTranscript,
|
|
171
|
+
language: this.lastLanguage
|
|
172
|
+
},
|
|
173
|
+
"received user preflight transcript"
|
|
174
|
+
);
|
|
175
|
+
this.lastFinalTranscriptTime = Date.now();
|
|
176
|
+
this.audioPreflightTranscript = `${this.audioTranscript} ${preflightTranscript}`.trimStart();
|
|
177
|
+
this.audioInterimTranscript = preflightTranscript;
|
|
178
|
+
if (!this.vad || this.lastSpeakingTime === void 0) {
|
|
179
|
+
this.lastSpeakingTime = Date.now();
|
|
180
|
+
}
|
|
181
|
+
if (this.turnDetectionMode !== "manual" || this.userTurnCommitted) {
|
|
182
|
+
const confidenceVals = [...this.finalTranscriptConfidence, preflightConfidence];
|
|
183
|
+
this.logger.debug(
|
|
184
|
+
{
|
|
185
|
+
transcript: this.audioPreflightTranscript.length > 100 ? this.audioPreflightTranscript.slice(0, 100) + "..." : this.audioPreflightTranscript
|
|
186
|
+
},
|
|
187
|
+
"triggering preemptive generation (PREFLIGHT_TRANSCRIPT)"
|
|
188
|
+
);
|
|
189
|
+
this.hooks.onPreemptiveGeneration({
|
|
190
|
+
newTranscript: this.audioPreflightTranscript,
|
|
191
|
+
transcriptConfidence: confidenceVals.length > 0 ? confidenceVals.reduce((a, b) => a + b, 0) / confidenceVals.length : 0
|
|
192
|
+
});
|
|
193
|
+
}
|
|
194
|
+
break;
|
|
139
195
|
case import_stt.SpeechEventType.INTERIM_TRANSCRIPT:
|
|
140
|
-
this.logger.debug({ transcript: (
|
|
196
|
+
this.logger.debug({ transcript: (_o = (_n = ev.alternatives) == null ? void 0 : _n[0]) == null ? void 0 : _o.text }, "interim transcript");
|
|
141
197
|
this.hooks.onInterimTranscript(ev);
|
|
142
|
-
this.audioInterimTranscript = ((
|
|
198
|
+
this.audioInterimTranscript = ((_q = (_p = ev.alternatives) == null ? void 0 : _p[0]) == null ? void 0 : _q.text) ?? "";
|
|
199
|
+
break;
|
|
200
|
+
case import_stt.SpeechEventType.START_OF_SPEECH:
|
|
201
|
+
if (this.turnDetectionMode !== "stt") break;
|
|
202
|
+
this.hooks.onStartOfSpeech({
|
|
203
|
+
type: import_vad.VADEventType.START_OF_SPEECH,
|
|
204
|
+
samplesIndex: 0,
|
|
205
|
+
timestamp: Date.now(),
|
|
206
|
+
speechDuration: 0,
|
|
207
|
+
silenceDuration: 0,
|
|
208
|
+
frames: [],
|
|
209
|
+
probability: 0,
|
|
210
|
+
inferenceDuration: 0,
|
|
211
|
+
speaking: true,
|
|
212
|
+
rawAccumulatedSilence: 0,
|
|
213
|
+
rawAccumulatedSpeech: 0
|
|
214
|
+
});
|
|
215
|
+
this.speaking = true;
|
|
216
|
+
this.lastSpeakingTime = Date.now();
|
|
217
|
+
(_r = this.bounceEOUTask) == null ? void 0 : _r.cancel();
|
|
143
218
|
break;
|
|
144
219
|
case import_stt.SpeechEventType.END_OF_SPEECH:
|
|
145
220
|
if (this.turnDetectionMode !== "stt") break;
|
|
221
|
+
this.hooks.onEndOfSpeech({
|
|
222
|
+
type: import_vad.VADEventType.END_OF_SPEECH,
|
|
223
|
+
samplesIndex: 0,
|
|
224
|
+
timestamp: Date.now(),
|
|
225
|
+
speechDuration: 0,
|
|
226
|
+
silenceDuration: 0,
|
|
227
|
+
frames: [],
|
|
228
|
+
probability: 0,
|
|
229
|
+
inferenceDuration: 0,
|
|
230
|
+
speaking: false,
|
|
231
|
+
rawAccumulatedSilence: 0,
|
|
232
|
+
rawAccumulatedSpeech: 0
|
|
233
|
+
});
|
|
234
|
+
this.speaking = false;
|
|
146
235
|
this.userTurnCommitted = true;
|
|
236
|
+
this.lastSpeakingTime = Date.now();
|
|
147
237
|
if (!this.speaking) {
|
|
148
238
|
const chatCtx = this.hooks.retrieveChatCtx();
|
|
149
239
|
this.logger.debug("running EOU detection on stt END_OF_SPEECH");
|
|
@@ -171,7 +261,7 @@ class AudioRecognition {
|
|
|
171
261
|
// disable EOU model if manual turn detection enabled
|
|
172
262
|
this.audioTranscript && this.turnDetectionMode !== "manual" ? this.turnDetector : void 0
|
|
173
263
|
);
|
|
174
|
-
const bounceEOUTask = (lastSpeakingTime) => async (controller) => {
|
|
264
|
+
const bounceEOUTask = (lastSpeakingTime, lastFinalTranscriptTime, speechStartTime) => async (controller) => {
|
|
175
265
|
let endpointingDelay = this.minEndpointingDelay;
|
|
176
266
|
if (turnDetector) {
|
|
177
267
|
this.logger.debug("Running turn detector model");
|
|
@@ -198,21 +288,46 @@ class AudioRecognition {
|
|
|
198
288
|
}
|
|
199
289
|
}
|
|
200
290
|
}
|
|
201
|
-
|
|
202
|
-
|
|
291
|
+
let extraSleep = endpointingDelay;
|
|
292
|
+
if (lastSpeakingTime !== void 0) {
|
|
293
|
+
extraSleep += lastSpeakingTime - Date.now();
|
|
294
|
+
}
|
|
295
|
+
if (extraSleep > 0) {
|
|
296
|
+
await (0, import_utils.delay)(Math.max(extraSleep, 0), { signal: controller.signal });
|
|
297
|
+
}
|
|
203
298
|
this.logger.debug({ transcript: this.audioTranscript }, "end of user turn");
|
|
299
|
+
const confidenceAvg = this.finalTranscriptConfidence.length > 0 ? this.finalTranscriptConfidence.reduce((a, b) => a + b, 0) / this.finalTranscriptConfidence.length : 0;
|
|
300
|
+
let startedSpeakingAt;
|
|
301
|
+
let stoppedSpeakingAt;
|
|
302
|
+
let transcriptionDelay;
|
|
303
|
+
let endOfUtteranceDelay;
|
|
304
|
+
if (lastFinalTranscriptTime !== 0 && lastSpeakingTime !== void 0 && speechStartTime !== void 0) {
|
|
305
|
+
startedSpeakingAt = speechStartTime;
|
|
306
|
+
stoppedSpeakingAt = lastSpeakingTime;
|
|
307
|
+
transcriptionDelay = Math.max(lastFinalTranscriptTime - lastSpeakingTime, 0);
|
|
308
|
+
endOfUtteranceDelay = Date.now() - lastSpeakingTime;
|
|
309
|
+
}
|
|
204
310
|
const committed = await this.hooks.onEndOfTurn({
|
|
205
311
|
newTranscript: this.audioTranscript,
|
|
206
|
-
|
|
207
|
-
|
|
312
|
+
transcriptConfidence: confidenceAvg,
|
|
313
|
+
transcriptionDelay: transcriptionDelay ?? 0,
|
|
314
|
+
endOfUtteranceDelay: endOfUtteranceDelay ?? 0,
|
|
315
|
+
startedSpeakingAt,
|
|
316
|
+
stoppedSpeakingAt
|
|
208
317
|
});
|
|
209
318
|
if (committed) {
|
|
210
319
|
this.audioTranscript = "";
|
|
320
|
+
this.finalTranscriptConfidence = [];
|
|
321
|
+
this.lastSpeakingTime = void 0;
|
|
322
|
+
this.lastFinalTranscriptTime = 0;
|
|
323
|
+
this.speechStartTime = void 0;
|
|
211
324
|
}
|
|
212
325
|
this.userTurnCommitted = false;
|
|
213
326
|
};
|
|
214
327
|
(_a = this.bounceEOUTask) == null ? void 0 : _a.cancel();
|
|
215
|
-
this.bounceEOUTask = import_utils.Task.from(
|
|
328
|
+
this.bounceEOUTask = import_utils.Task.from(
|
|
329
|
+
bounceEOUTask(this.lastSpeakingTime, this.lastFinalTranscriptTime, this.speechStartTime)
|
|
330
|
+
);
|
|
216
331
|
this.bounceEOUTask.result.then(() => {
|
|
217
332
|
this.logger.debug("EOU detection task completed");
|
|
218
333
|
}).catch((err) => {
|
|
@@ -292,12 +407,17 @@ class AudioRecognition {
|
|
|
292
407
|
break;
|
|
293
408
|
case import_vad.VADEventType.INFERENCE_DONE:
|
|
294
409
|
this.hooks.onVADInferenceDone(ev);
|
|
410
|
+
if (ev.rawAccumulatedSpeech > 0) {
|
|
411
|
+
this.lastSpeakingTime = Date.now();
|
|
412
|
+
if (this.speechStartTime === void 0) {
|
|
413
|
+
this.speechStartTime = Date.now();
|
|
414
|
+
}
|
|
415
|
+
}
|
|
295
416
|
break;
|
|
296
417
|
case import_vad.VADEventType.END_OF_SPEECH:
|
|
297
418
|
this.logger.debug("VAD task: END_OF_SPEECH");
|
|
298
419
|
this.hooks.onEndOfSpeech(ev);
|
|
299
420
|
this.speaking = false;
|
|
300
|
-
this.lastSpeakingTime = Date.now() - ev.silenceDuration;
|
|
301
421
|
if (this.vadBaseTurnDetection || this.turnDetectionMode === "stt" && this.userTurnCommitted) {
|
|
302
422
|
const chatCtx = this.hooks.retrieveChatCtx();
|
|
303
423
|
this.runEOUDetection(chatCtx);
|
|
@@ -321,6 +441,8 @@ class AudioRecognition {
|
|
|
321
441
|
var _a;
|
|
322
442
|
this.audioTranscript = "";
|
|
323
443
|
this.audioInterimTranscript = "";
|
|
444
|
+
this.audioPreflightTranscript = "";
|
|
445
|
+
this.finalTranscriptConfidence = [];
|
|
324
446
|
this.userTurnCommitted = false;
|
|
325
447
|
(_a = this.sttTask) == null ? void 0 : _a.cancelAndWait().finally(() => {
|
|
326
448
|
this.sttTask = import_utils.Task.from(({ signal }) => this.createSttTask(this.stt, signal));
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/voice/audio_recognition.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ReadableStream } from 'node:stream/web';\nimport { type ChatContext } from '../llm/chat_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream, isStreamReaderReleaseError } from '../stream/deferred_stream.js';\nimport { IdentityTransform } from '../stream/identity_transform.js';\nimport { mergeReadableStreams } from '../stream/merge_readable_streams.js';\nimport { type SpeechEvent, SpeechEventType } from '../stt/stt.js';\nimport { Task, delay } from '../utils.js';\nimport { type VAD, type VADEvent, VADEventType } from '../vad.js';\nimport type { TurnDetectionMode } from './agent_session.js';\nimport type { STTNode } from './io.js';\n\nexport interface EndOfTurnInfo {\n newTranscript: string;\n transcriptionDelay: number;\n endOfUtteranceDelay: number;\n}\n\nexport interface RecognitionHooks {\n onStartOfSpeech: (ev: VADEvent) => void;\n onVADInferenceDone: (ev: VADEvent) => void;\n onEndOfSpeech: (ev: VADEvent) => void;\n onInterimTranscript: (ev: SpeechEvent) => void;\n onFinalTranscript: (ev: SpeechEvent) => void;\n onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;\n\n retrieveChatCtx: () => ChatContext;\n}\n\nexport interface _TurnDetector {\n unlikelyThreshold: (language?: string) => Promise<number | undefined>;\n supportsLanguage: (language?: string) => Promise<boolean>;\n predictEndOfTurn(chatCtx: ChatContext): Promise<number>;\n}\n\nexport interface AudioRecognitionOptions {\n recognitionHooks: RecognitionHooks;\n stt?: STTNode;\n vad?: VAD;\n turnDetector?: _TurnDetector;\n turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n}\n\nexport class AudioRecognition {\n private hooks: RecognitionHooks;\n private stt?: STTNode;\n private vad?: VAD;\n private turnDetector?: _TurnDetector;\n private turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n private minEndpointingDelay: number;\n private maxEndpointingDelay: number;\n private lastLanguage?: string;\n\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private lastFinalTranscriptTime = 0;\n private audioTranscript = '';\n private audioInterimTranscript = '';\n private lastSpeakingTime = 0;\n private userTurnCommitted = false;\n private speaking = false;\n private sampleRate?: number;\n\n private vadInputStream: ReadableStream<AudioFrame>;\n private sttInputStream: ReadableStream<AudioFrame>;\n private silenceAudioTransform = new IdentityTransform<AudioFrame>();\n private silenceAudioWriter: WritableStreamDefaultWriter<AudioFrame>;\n\n // all cancellable tasks\n private bounceEOUTask?: Task<void>;\n private commitUserTurnTask?: Task<void>;\n private vadTask?: Task<void>;\n private sttTask?: Task<void>;\n\n constructor(opts: AudioRecognitionOptions) {\n this.hooks = opts.recognitionHooks;\n this.stt = opts.stt;\n this.vad = opts.vad;\n this.turnDetector = opts.turnDetector;\n this.turnDetectionMode = opts.turnDetectionMode;\n this.minEndpointingDelay = opts.minEndpointingDelay;\n this.maxEndpointingDelay = opts.maxEndpointingDelay;\n this.lastLanguage = undefined;\n\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();\n this.vadInputStream = vadInputStream;\n this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);\n this.silenceAudioWriter = this.silenceAudioTransform.writable.getWriter();\n }\n\n /**\n * Current transcript of the user's speech, including interim transcript if available.\n */\n get currentTranscript(): string {\n if (this.audioInterimTranscript) {\n return `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n }\n return this.audioTranscript;\n }\n\n async start() {\n this.vadTask = Task.from(({ signal }) => this.createVadTask(this.vad, signal));\n this.vadTask.result.catch((err) => {\n this.logger.error(`Error running VAD task: ${err}`);\n });\n\n this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n this.sttTask.result.catch((err) => {\n this.logger.error(`Error running STT task: ${err}`);\n });\n }\n\n private async onSTTEvent(ev: SpeechEvent) {\n if (\n this.turnDetectionMode === 'manual' &&\n this.userTurnCommitted &&\n (this.bounceEOUTask === undefined ||\n this.bounceEOUTask.done ||\n ev.type == SpeechEventType.INTERIM_TRANSCRIPT)\n ) {\n // ignore stt event if user turn already committed and EOU task is done\n // or it's an interim transcript\n this.logger.debug(\n {\n userTurnCommitted: this.userTurnCommitted,\n eouTaskDone: this.bounceEOUTask?.done,\n evType: ev.type,\n turnDetectionMode: this.turnDetectionMode,\n },\n 'ignoring stt event',\n );\n return;\n }\n\n switch (ev.type) {\n case SpeechEventType.FINAL_TRANSCRIPT:\n this.hooks.onFinalTranscript(ev);\n const transcript = ev.alternatives?.[0]?.text;\n this.lastLanguage = ev.alternatives?.[0]?.language;\n\n if (!transcript) {\n // stt final transcript received but no transcript\n return;\n }\n\n this.logger.debug(\n {\n user_transcript: transcript,\n language: this.lastLanguage,\n },\n 'received user transcript',\n );\n\n this.lastFinalTranscriptTime = Date.now();\n this.audioTranscript += ` ${transcript}`;\n this.audioTranscript = this.audioTranscript.trimStart();\n this.audioInterimTranscript = '';\n\n if (!this.speaking) {\n if (!this.vad) {\n // Copied from python agents:\n // vad disabled, use stt timestamp\n // TODO: this would screw up transcription latency metrics\n // but we'll live with it for now.\n // the correct way is to ensure STT fires SpeechEventType.END_OF_SPEECH\n // and using that timestamp for _last_speaking_time\n this.lastSpeakingTime = Date.now();\n }\n\n if (this.vadBaseTurnDetection || this.userTurnCommitted) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on stt FINAL_TRANSCRIPT');\n this.runEOUDetection(chatCtx);\n }\n }\n break;\n case SpeechEventType.INTERIM_TRANSCRIPT:\n this.logger.debug({ transcript: ev.alternatives?.[0]?.text }, 'interim transcript');\n this.hooks.onInterimTranscript(ev);\n this.audioInterimTranscript = ev.alternatives?.[0]?.text ?? '';\n break;\n case SpeechEventType.END_OF_SPEECH:\n if (this.turnDetectionMode !== 'stt') break;\n this.userTurnCommitted = true;\n\n if (!this.speaking) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on stt END_OF_SPEECH');\n this.runEOUDetection(chatCtx);\n }\n }\n }\n\n private runEOUDetection(chatCtx: ChatContext) {\n this.logger.debug(\n {\n stt: this.stt,\n audioTranscript: this.audioTranscript,\n turnDetectionMode: this.turnDetectionMode,\n },\n 'running EOU detection',\n );\n\n if (this.stt && !this.audioTranscript && this.turnDetectionMode !== 'manual') {\n // stt enabled but no transcript yet\n this.logger.debug('skipping EOU detection');\n return;\n }\n\n chatCtx = chatCtx.copy();\n chatCtx.addMessage({ role: 'user', content: this.audioTranscript });\n\n const turnDetector =\n // disable EOU model if manual turn detection enabled\n this.audioTranscript && this.turnDetectionMode !== 'manual' ? this.turnDetector : undefined;\n\n const bounceEOUTask = (lastSpeakingTime: number) => async (controller: AbortController) => {\n let endpointingDelay = this.minEndpointingDelay;\n\n // TODO(AJS-74): need to support actual turn detection model plugins for following code to run\n if (turnDetector) {\n this.logger.debug('Running turn detector model');\n if (!turnDetector.supportsLanguage(this.lastLanguage)) {\n this.logger.debug(`Turn detector does not support language ${this.lastLanguage}`);\n } else {\n const endOfTurnProbability = await turnDetector.predictEndOfTurn(chatCtx);\n this.logger.debug(\n { endOfTurnProbability, language: this.lastLanguage },\n 'end of turn probability',\n );\n\n const unlikelyThreshold = await turnDetector.unlikelyThreshold(this.lastLanguage);\n this.logger.debug(\n {\n unlikelyThreshold,\n endOfTurnProbability,\n language: this.lastLanguage,\n transcript: this.audioTranscript,\n },\n 'EOU Detection',\n );\n\n if (unlikelyThreshold && endOfTurnProbability < unlikelyThreshold) {\n endpointingDelay = this.maxEndpointingDelay;\n }\n }\n }\n\n const extraSleep = lastSpeakingTime + endpointingDelay - Date.now();\n // add delay to see if there's a potential upcoming EOU task that cancels this one\n await delay(Math.max(extraSleep, 0), { signal: controller.signal });\n\n this.logger.debug({ transcript: this.audioTranscript }, 'end of user turn');\n\n const committed = await this.hooks.onEndOfTurn({\n newTranscript: this.audioTranscript,\n transcriptionDelay: Math.max(this.lastFinalTranscriptTime - lastSpeakingTime, 0),\n endOfUtteranceDelay: Date.now() - lastSpeakingTime,\n });\n\n if (committed) {\n // clear the transcript if the user turn was committed\n this.audioTranscript = '';\n }\n\n this.userTurnCommitted = false;\n };\n\n // cancel any existing EOU task\n this.bounceEOUTask?.cancel();\n this.bounceEOUTask = Task.from(bounceEOUTask(this.lastSpeakingTime));\n\n this.bounceEOUTask.result\n .then(() => {\n this.logger.debug('EOU detection task completed');\n })\n .catch((err: unknown) => {\n if (err instanceof Error && err.message.includes('This operation was aborted')) {\n // ignore aborted errors\n return;\n }\n this.logger.error(err, 'Error in EOU detection task:');\n });\n }\n\n private async createSttTask(stt: STTNode | undefined, signal: AbortSignal) {\n if (!stt) return;\n\n this.logger.debug('createSttTask: create stt stream from stt node');\n\n const sttStream = await stt(this.sttInputStream, {});\n\n if (signal.aborted || sttStream === null) return;\n\n if (sttStream instanceof ReadableStream) {\n const reader = sttStream.getReader();\n\n signal.addEventListener('abort', async () => {\n try {\n reader.releaseLock();\n await sttStream?.cancel();\n } catch (e) {\n this.logger.debug('createSttTask: error during abort handler:', e);\n }\n });\n\n try {\n while (true) {\n if (signal.aborted) break;\n\n const { done, value: ev } = await reader.read();\n if (done) break;\n\n if (typeof ev === 'string') {\n throw new Error('STT node must yield SpeechEvent');\n } else {\n await this.onSTTEvent(ev);\n }\n }\n } catch (e) {\n if (isStreamReaderReleaseError(e)) {\n return;\n }\n this.logger.error({ error: e }, 'createSttTask: error reading sttStream');\n } finally {\n reader.releaseLock();\n try {\n await sttStream.cancel();\n } catch (e) {\n this.logger.debug(\n 'createSttTask: error cancelling sttStream (may already be cancelled):',\n e,\n );\n }\n }\n }\n }\n\n private async createVadTask(vad: VAD | undefined, signal: AbortSignal) {\n if (!vad) return;\n\n const vadStream = vad.stream();\n vadStream.updateInputStream(this.vadInputStream);\n\n const abortHandler = () => {\n vadStream.detachInputStream();\n vadStream.close();\n signal.removeEventListener('abort', abortHandler);\n };\n signal.addEventListener('abort', abortHandler);\n\n try {\n for await (const ev of vadStream) {\n if (signal.aborted) break;\n\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.logger.debug('VAD task: START_OF_SPEECH');\n this.hooks.onStartOfSpeech(ev);\n this.speaking = true;\n\n // Capture sample rate from the first VAD event if not already set\n if (ev.frames.length > 0 && ev.frames[0]) {\n this.sampleRate = ev.frames[0].sampleRate;\n }\n\n this.bounceEOUTask?.cancel();\n break;\n case VADEventType.INFERENCE_DONE:\n this.hooks.onVADInferenceDone(ev);\n break;\n case VADEventType.END_OF_SPEECH:\n this.logger.debug('VAD task: END_OF_SPEECH');\n this.hooks.onEndOfSpeech(ev);\n this.speaking = false;\n // when VAD fires END_OF_SPEECH, it already waited for the silence_duration\n this.lastSpeakingTime = Date.now() - ev.silenceDuration;\n\n if (\n this.vadBaseTurnDetection ||\n (this.turnDetectionMode === 'stt' && this.userTurnCommitted)\n ) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.runEOUDetection(chatCtx);\n }\n break;\n }\n }\n } catch (e) {\n this.logger.error(e, 'Error in VAD task');\n } finally {\n this.logger.debug('VAD task closed');\n }\n }\n\n setInputAudioStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputAudioStream() {\n this.deferredInputStream.detachSource();\n }\n\n clearUserTurn() {\n this.audioTranscript = '';\n this.audioInterimTranscript = '';\n this.userTurnCommitted = false;\n\n this.sttTask?.cancelAndWait().finally(() => {\n this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n this.sttTask.result.catch((err) => {\n this.logger.error(`Error running STT task: ${err}`);\n });\n });\n }\n\n commitUserTurn(audioDetached: boolean) {\n const commitUserTurnTask =\n (delayDuration: number = 500) =>\n async (controller: AbortController) => {\n if (Date.now() - this.lastFinalTranscriptTime > delayDuration) {\n // flush the stt by pushing silence\n if (audioDetached && this.sampleRate !== undefined) {\n const numSamples = Math.floor(this.sampleRate * 0.5);\n const silence = new Int16Array(numSamples * 2);\n const silenceFrame = new AudioFrame(silence, this.sampleRate, 1, numSamples);\n this.silenceAudioWriter.write(silenceFrame);\n }\n\n // wait for the final transcript to be available\n await delay(delayDuration, { signal: controller.signal });\n }\n\n if (this.audioInterimTranscript) {\n // append interim transcript in case the final transcript is not ready\n this.audioTranscript = `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n }\n this.audioInterimTranscript = '';\n\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on commitUserTurn');\n this.runEOUDetection(chatCtx);\n this.userTurnCommitted = true;\n };\n\n // cancel any existing commit user turn task\n this.commitUserTurnTask?.cancel();\n this.commitUserTurnTask = Task.from(commitUserTurnTask());\n\n this.commitUserTurnTask.result\n .then(() => {\n this.logger.debug('User turn committed');\n })\n .catch((err: unknown) => {\n this.logger.error(err, 'Error in user turn commit task:');\n });\n }\n\n async close() {\n this.detachInputAudioStream();\n await this.commitUserTurnTask?.cancelAndWait();\n await this.sttTask?.cancelAndWait();\n await this.vadTask?.cancelAndWait();\n await this.bounceEOUTask?.cancelAndWait();\n }\n\n private get vadBaseTurnDetection() {\n return ['vad', undefined].includes(this.turnDetectionMode);\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAA2B;AAE3B,iBAA+B;AAC/B,0BAAiC;AACjC,iBAAoB;AACpB,6BAAmE;AACnE,gCAAkC;AAClC,oCAAqC;AACrC,iBAAkD;AAClD,mBAA4B;AAC5B,iBAAsD;AAqC/C,MAAM,iBAAiB;AAAA,EACpB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA,aAAS,gBAAI;AAAA,EACb,0BAA0B;AAAA,EAC1B,kBAAkB;AAAA,EAClB,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EACpB,WAAW;AAAA,EACX;AAAA,EAEA;AAAA,EACA;AAAA,EACA,wBAAwB,IAAI,4CAA8B;AAAA,EAC1D;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,MAA+B;AACzC,SAAK,QAAQ,KAAK;AAClB,SAAK,MAAM,KAAK;AAChB,SAAK,MAAM,KAAK;AAChB,SAAK,eAAe,KAAK;AACzB,SAAK,oBAAoB,KAAK;AAC9B,SAAK,sBAAsB,KAAK;AAChC,SAAK,sBAAsB,KAAK;AAChC,SAAK,eAAe;AAEpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,UAAM,CAAC,gBAAgB,cAAc,IAAI,KAAK,oBAAoB,OAAO,IAAI;AAC7E,SAAK,iBAAiB;AACtB,SAAK,qBAAiB,oDAAqB,gBAAgB,KAAK,sBAAsB,QAAQ;AAC9F,SAAK,qBAAqB,KAAK,sBAAsB,SAAS,UAAU;AAAA,EAC1E;AAAA;AAAA;AAAA;AAAA,EAKA,IAAI,oBAA4B;AAC9B,QAAI,KAAK,wBAAwB;AAC/B,aAAO,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,IACvE;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAED,SAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,WAAW,IAAiB;AAxH5C;AAyHI,QACE,KAAK,sBAAsB,YAC3B,KAAK,sBACJ,KAAK,kBAAkB,UACtB,KAAK,cAAc,QACnB,GAAG,QAAQ,2BAAgB,qBAC7B;AAGA,WAAK,OAAO;AAAA,QACV;AAAA,UACE,mBAAmB,KAAK;AAAA,UACxB,cAAa,UAAK,kBAAL,mBAAoB;AAAA,UACjC,QAAQ,GAAG;AAAA,UACX,mBAAmB,KAAK;AAAA,QAC1B;AAAA,QACA;AAAA,MACF;AACA;AAAA,IACF;AAEA,YAAQ,GAAG,MAAM;AAAA,MACf,KAAK,2BAAgB;AACnB,aAAK,MAAM,kBAAkB,EAAE;AAC/B,cAAM,cAAa,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AACzC,aAAK,gBAAe,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AAE1C,YAAI,CAAC,YAAY;AAEf;AAAA,QACF;AAEA,aAAK,OAAO;AAAA,UACV;AAAA,YACE,iBAAiB;AAAA,YACjB,UAAU,KAAK;AAAA,UACjB;AAAA,UACA;AAAA,QACF;AAEA,aAAK,0BAA0B,KAAK,IAAI;AACxC,aAAK,mBAAmB,IAAI,UAAU;AACtC,aAAK,kBAAkB,KAAK,gBAAgB,UAAU;AACtD,aAAK,yBAAyB;AAE9B,YAAI,CAAC,KAAK,UAAU;AAClB,cAAI,CAAC,KAAK,KAAK;AAOb,iBAAK,mBAAmB,KAAK,IAAI;AAAA,UACnC;AAEA,cAAI,KAAK,wBAAwB,KAAK,mBAAmB;AACvD,kBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,iBAAK,OAAO,MAAM,+CAA+C;AACjE,iBAAK,gBAAgB,OAAO;AAAA,UAC9B;AAAA,QACF;AACA;AAAA,MACF,KAAK,2BAAgB;AACnB,aAAK,OAAO,MAAM,EAAE,aAAY,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,KAAK,GAAG,oBAAoB;AAClF,aAAK,MAAM,oBAAoB,EAAE;AACjC,aAAK,2BAAyB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,SAAQ;AAC5D;AAAA,MACF,KAAK,2BAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC,aAAK,oBAAoB;AAEzB,YAAI,CAAC,KAAK,UAAU;AAClB,gBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,eAAK,OAAO,MAAM,4CAA4C;AAC9D,eAAK,gBAAgB,OAAO;AAAA,QAC9B;AAAA,IACJ;AAAA,EACF;AAAA,EAEQ,gBAAgB,SAAsB;AAzMhD;AA0MI,SAAK,OAAO;AAAA,MACV;AAAA,QACE,KAAK,KAAK;AAAA,QACV,iBAAiB,KAAK;AAAA,QACtB,mBAAmB,KAAK;AAAA,MAC1B;AAAA,MACA;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,CAAC,KAAK,mBAAmB,KAAK,sBAAsB,UAAU;AAE5E,WAAK,OAAO,MAAM,wBAAwB;AAC1C;AAAA,IACF;AAEA,cAAU,QAAQ,KAAK;AACvB,YAAQ,WAAW,EAAE,MAAM,QAAQ,SAAS,KAAK,gBAAgB,CAAC;AAElE,UAAM;AAAA;AAAA,MAEJ,KAAK,mBAAmB,KAAK,sBAAsB,WAAW,KAAK,eAAe;AAAA;AAEpF,UAAM,gBAAgB,CAAC,qBAA6B,OAAO,eAAgC;AACzF,UAAI,mBAAmB,KAAK;AAG5B,UAAI,cAAc;AAChB,aAAK,OAAO,MAAM,6BAA6B;AAC/C,YAAI,CAAC,aAAa,iBAAiB,KAAK,YAAY,GAAG;AACrD,eAAK,OAAO,MAAM,2CAA2C,KAAK,YAAY,EAAE;AAAA,QAClF,OAAO;AACL,gBAAM,uBAAuB,MAAM,aAAa,iBAAiB,OAAO;AACxE,eAAK,OAAO;AAAA,YACV,EAAE,sBAAsB,UAAU,KAAK,aAAa;AAAA,YACpD;AAAA,UACF;AAEA,gBAAM,oBAAoB,MAAM,aAAa,kBAAkB,KAAK,YAAY;AAChF,eAAK,OAAO;AAAA,YACV;AAAA,cACE;AAAA,cACA;AAAA,cACA,UAAU,KAAK;AAAA,cACf,YAAY,KAAK;AAAA,YACnB;AAAA,YACA;AAAA,UACF;AAEA,cAAI,qBAAqB,uBAAuB,mBAAmB;AACjE,+BAAmB,KAAK;AAAA,UAC1B;AAAA,QACF;AAAA,MACF;AAEA,YAAM,aAAa,mBAAmB,mBAAmB,KAAK,IAAI;AAElE,gBAAM,oBAAM,KAAK,IAAI,YAAY,CAAC,GAAG,EAAE,QAAQ,WAAW,OAAO,CAAC;AAElE,WAAK,OAAO,MAAM,EAAE,YAAY,KAAK,gBAAgB,GAAG,kBAAkB;AAE1E,YAAM,YAAY,MAAM,KAAK,MAAM,YAAY;AAAA,QAC7C,eAAe,KAAK;AAAA,QACpB,oBAAoB,KAAK,IAAI,KAAK,0BAA0B,kBAAkB,CAAC;AAAA,QAC/E,qBAAqB,KAAK,IAAI,IAAI;AAAA,MACpC,CAAC;AAED,UAAI,WAAW;AAEb,aAAK,kBAAkB;AAAA,MACzB;AAEA,WAAK,oBAAoB;AAAA,IAC3B;AAGA,eAAK,kBAAL,mBAAoB;AACpB,SAAK,gBAAgB,kBAAK,KAAK,cAAc,KAAK,gBAAgB,CAAC;AAEnE,SAAK,cAAc,OAChB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,8BAA8B;AAAA,IAClD,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,UAAI,eAAe,SAAS,IAAI,QAAQ,SAAS,4BAA4B,GAAG;AAE9E;AAAA,MACF;AACA,WAAK,OAAO,MAAM,KAAK,8BAA8B;AAAA,IACvD,CAAC;AAAA,EACL;AAAA,EAEA,MAAc,cAAc,KAA0B,QAAqB;AACzE,QAAI,CAAC,IAAK;AAEV,SAAK,OAAO,MAAM,gDAAgD;AAElE,UAAM,YAAY,MAAM,IAAI,KAAK,gBAAgB,CAAC,CAAC;AAEnD,QAAI,OAAO,WAAW,cAAc,KAAM;AAE1C,QAAI,qBAAqB,2BAAgB;AACvC,YAAM,SAAS,UAAU,UAAU;AAEnC,aAAO,iBAAiB,SAAS,YAAY;AAC3C,YAAI;AACF,iBAAO,YAAY;AACnB,iBAAM,uCAAW;AAAA,QACnB,SAAS,GAAG;AACV,eAAK,OAAO,MAAM,8CAA8C,CAAC;AAAA,QACnE;AAAA,MACF,CAAC;AAED,UAAI;AACF,eAAO,MAAM;AACX,cAAI,OAAO,QAAS;AAEpB,gBAAM,EAAE,MAAM,OAAO,GAAG,IAAI,MAAM,OAAO,KAAK;AAC9C,cAAI,KAAM;AAEV,cAAI,OAAO,OAAO,UAAU;AAC1B,kBAAM,IAAI,MAAM,iCAAiC;AAAA,UACnD,OAAO;AACL,kBAAM,KAAK,WAAW,EAAE;AAAA,UAC1B;AAAA,QACF;AAAA,MACF,SAAS,GAAG;AACV,gBAAI,mDAA2B,CAAC,GAAG;AACjC;AAAA,QACF;AACA,aAAK,OAAO,MAAM,EAAE,OAAO,EAAE,GAAG,wCAAwC;AAAA,MAC1E,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,UAAU,OAAO;AAAA,QACzB,SAAS,GAAG;AACV,eAAK,OAAO;AAAA,YACV;AAAA,YACA;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,cAAc,KAAsB,QAAqB;AA1VzE;AA2VI,QAAI,CAAC,IAAK;AAEV,UAAM,YAAY,IAAI,OAAO;AAC7B,cAAU,kBAAkB,KAAK,cAAc;AAE/C,UAAM,eAAe,MAAM;AACzB,gBAAU,kBAAkB;AAC5B,gBAAU,MAAM;AAChB,aAAO,oBAAoB,SAAS,YAAY;AAAA,IAClD;AACA,WAAO,iBAAiB,SAAS,YAAY;AAE7C,QAAI;AACF,uBAAiB,MAAM,WAAW;AAChC,YAAI,OAAO,QAAS;AAEpB,gBAAQ,GAAG,MAAM;AAAA,UACf,KAAK,wBAAa;AAChB,iBAAK,OAAO,MAAM,2BAA2B;AAC7C,iBAAK,MAAM,gBAAgB,EAAE;AAC7B,iBAAK,WAAW;AAGhB,gBAAI,GAAG,OAAO,SAAS,KAAK,GAAG,OAAO,CAAC,GAAG;AACxC,mBAAK,aAAa,GAAG,OAAO,CAAC,EAAE;AAAA,YACjC;AAEA,uBAAK,kBAAL,mBAAoB;AACpB;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,MAAM,mBAAmB,EAAE;AAChC;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,OAAO,MAAM,yBAAyB;AAC3C,iBAAK,MAAM,cAAc,EAAE;AAC3B,iBAAK,WAAW;AAEhB,iBAAK,mBAAmB,KAAK,IAAI,IAAI,GAAG;AAExC,gBACE,KAAK,wBACJ,KAAK,sBAAsB,SAAS,KAAK,mBAC1C;AACA,oBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,mBAAK,gBAAgB,OAAO;AAAA,YAC9B;AACA;AAAA,QACJ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,GAAG,mBAAmB;AAAA,IAC1C,UAAE;AACA,WAAK,OAAO,MAAM,iBAAiB;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,oBAAoB,aAAyC;AAC3D,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,yBAAyB;AACvB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA,EAEA,gBAAgB;AA3ZlB;AA4ZI,SAAK,kBAAkB;AACvB,SAAK,yBAAyB;AAC9B,SAAK,oBAAoB;AAEzB,eAAK,YAAL,mBAAc,gBAAgB,QAAQ,MAAM;AAC1C,WAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,WAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,aAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,MACpD,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,eAAe,eAAwB;AAxazC;AAyaI,UAAM,qBACJ,CAAC,gBAAwB,QACzB,OAAO,eAAgC;AACrC,UAAI,KAAK,IAAI,IAAI,KAAK,0BAA0B,eAAe;AAE7D,YAAI,iBAAiB,KAAK,eAAe,QAAW;AAClD,gBAAM,aAAa,KAAK,MAAM,KAAK,aAAa,GAAG;AACnD,gBAAM,UAAU,IAAI,WAAW,aAAa,CAAC;AAC7C,gBAAM,eAAe,IAAI,2BAAW,SAAS,KAAK,YAAY,GAAG,UAAU;AAC3E,eAAK,mBAAmB,MAAM,YAAY;AAAA,QAC5C;AAGA,kBAAM,oBAAM,eAAe,EAAE,QAAQ,WAAW,OAAO,CAAC;AAAA,MAC1D;AAEA,UAAI,KAAK,wBAAwB;AAE/B,aAAK,kBAAkB,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,MACvF;AACA,WAAK,yBAAyB;AAE9B,YAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,WAAK,OAAO,MAAM,yCAAyC;AAC3D,WAAK,gBAAgB,OAAO;AAC5B,WAAK,oBAAoB;AAAA,IAC3B;AAGF,eAAK,uBAAL,mBAAyB;AACzB,SAAK,qBAAqB,kBAAK,KAAK,mBAAmB,CAAC;AAExD,SAAK,mBAAmB,OACrB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,qBAAqB;AAAA,IACzC,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,WAAK,OAAO,MAAM,KAAK,iCAAiC;AAAA,IAC1D,CAAC;AAAA,EACL;AAAA,EAEA,MAAM,QAAQ;AAldhB;AAmdI,SAAK,uBAAuB;AAC5B,YAAM,UAAK,uBAAL,mBAAyB;AAC/B,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,kBAAL,mBAAoB;AAAA,EAC5B;AAAA,EAEA,IAAY,uBAAuB;AACjC,WAAO,CAAC,OAAO,MAAS,EAAE,SAAS,KAAK,iBAAiB;AAAA,EAC3D;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/voice/audio_recognition.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ReadableStream } from 'node:stream/web';\nimport { type ChatContext } from '../llm/chat_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream, isStreamReaderReleaseError } from '../stream/deferred_stream.js';\nimport { IdentityTransform } from '../stream/identity_transform.js';\nimport { mergeReadableStreams } from '../stream/merge_readable_streams.js';\nimport { type SpeechEvent, SpeechEventType } from '../stt/stt.js';\nimport { Task, delay } from '../utils.js';\nimport { type VAD, type VADEvent, VADEventType } from '../vad.js';\nimport type { TurnDetectionMode } from './agent_session.js';\nimport type { STTNode } from './io.js';\n\nexport interface EndOfTurnInfo {\n newTranscript: string;\n transcriptConfidence: number;\n transcriptionDelay: number;\n endOfUtteranceDelay: number;\n startedSpeakingAt: number | undefined;\n stoppedSpeakingAt: number | undefined;\n}\n\nexport interface PreemptiveGenerationInfo {\n newTranscript: string;\n transcriptConfidence: number;\n}\n\nexport interface RecognitionHooks {\n onStartOfSpeech: (ev: VADEvent) => void;\n onVADInferenceDone: (ev: VADEvent) => void;\n onEndOfSpeech: (ev: VADEvent) => void;\n onInterimTranscript: (ev: SpeechEvent) => void;\n onFinalTranscript: (ev: SpeechEvent) => void;\n onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;\n onPreemptiveGeneration: (info: PreemptiveGenerationInfo) => void;\n\n retrieveChatCtx: () => ChatContext;\n}\n\nexport interface _TurnDetector {\n unlikelyThreshold: (language?: string) => Promise<number | undefined>;\n supportsLanguage: (language?: string) => Promise<boolean>;\n predictEndOfTurn(chatCtx: ChatContext): Promise<number>;\n}\n\nexport interface AudioRecognitionOptions {\n recognitionHooks: RecognitionHooks;\n stt?: STTNode;\n vad?: VAD;\n turnDetector?: _TurnDetector;\n turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n}\n\nexport class AudioRecognition {\n private hooks: RecognitionHooks;\n private stt?: STTNode;\n private vad?: VAD;\n private turnDetector?: _TurnDetector;\n private turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n private minEndpointingDelay: number;\n private maxEndpointingDelay: number;\n private lastLanguage?: string;\n\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private lastFinalTranscriptTime = 0;\n private audioTranscript = '';\n private audioInterimTranscript = '';\n private audioPreflightTranscript = '';\n private finalTranscriptConfidence: number[] = [];\n private lastSpeakingTime: number | undefined;\n private speechStartTime: number | undefined;\n private userTurnCommitted = false;\n private speaking = false;\n private sampleRate?: number;\n\n private vadInputStream: ReadableStream<AudioFrame>;\n private sttInputStream: ReadableStream<AudioFrame>;\n private silenceAudioTransform = new IdentityTransform<AudioFrame>();\n private silenceAudioWriter: WritableStreamDefaultWriter<AudioFrame>;\n\n // all cancellable tasks\n private bounceEOUTask?: Task<void>;\n private commitUserTurnTask?: Task<void>;\n private vadTask?: Task<void>;\n private sttTask?: Task<void>;\n\n constructor(opts: AudioRecognitionOptions) {\n this.hooks = opts.recognitionHooks;\n this.stt = opts.stt;\n this.vad = opts.vad;\n this.turnDetector = opts.turnDetector;\n this.turnDetectionMode = opts.turnDetectionMode;\n this.minEndpointingDelay = opts.minEndpointingDelay;\n this.maxEndpointingDelay = opts.maxEndpointingDelay;\n this.lastLanguage = undefined;\n\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();\n this.vadInputStream = vadInputStream;\n this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);\n this.silenceAudioWriter = this.silenceAudioTransform.writable.getWriter();\n }\n\n /**\n * Current transcript of the user's speech, including interim transcript if available.\n */\n get currentTranscript(): string {\n if (this.audioInterimTranscript) {\n return `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n }\n return this.audioTranscript;\n }\n\n async start() {\n this.vadTask = Task.from(({ signal }) => this.createVadTask(this.vad, signal));\n this.vadTask.result.catch((err) => {\n this.logger.error(`Error running VAD task: ${err}`);\n });\n\n this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n this.sttTask.result.catch((err) => {\n this.logger.error(`Error running STT task: ${err}`);\n });\n }\n\n private async onSTTEvent(ev: SpeechEvent) {\n if (\n this.turnDetectionMode === 'manual' &&\n this.userTurnCommitted &&\n (this.bounceEOUTask === undefined ||\n this.bounceEOUTask.done ||\n ev.type == SpeechEventType.INTERIM_TRANSCRIPT)\n ) {\n // ignore stt event if user turn already committed and EOU task is done\n // or it's an interim transcript\n this.logger.debug(\n {\n userTurnCommitted: this.userTurnCommitted,\n eouTaskDone: this.bounceEOUTask?.done,\n evType: ev.type,\n turnDetectionMode: this.turnDetectionMode,\n },\n 'ignoring stt event',\n );\n return;\n }\n\n switch (ev.type) {\n case SpeechEventType.FINAL_TRANSCRIPT:\n this.hooks.onFinalTranscript(ev);\n const transcript = ev.alternatives?.[0]?.text;\n const confidence = ev.alternatives?.[0]?.confidence ?? 0;\n this.lastLanguage = ev.alternatives?.[0]?.language;\n\n if (!transcript) {\n // stt final transcript received but no transcript\n return;\n }\n\n this.logger.debug(\n {\n user_transcript: transcript,\n language: this.lastLanguage,\n },\n 'received user transcript',\n );\n\n this.lastFinalTranscriptTime = Date.now();\n this.audioTranscript += ` ${transcript}`;\n this.audioTranscript = this.audioTranscript.trimStart();\n this.finalTranscriptConfidence.push(confidence);\n const transcriptChanged = this.audioTranscript !== this.audioPreflightTranscript;\n this.audioInterimTranscript = '';\n this.audioPreflightTranscript = '';\n\n if (!this.vad || this.lastSpeakingTime === undefined) {\n // vad disabled, use stt timestamp\n // TODO: this would screw up transcription latency metrics\n // but we'll live with it for now.\n // the correct way is to ensure STT fires SpeechEventType.END_OF_SPEECH\n // and using that timestamp for lastSpeakingTime\n this.lastSpeakingTime = Date.now();\n }\n\n if (this.vadBaseTurnDetection || this.userTurnCommitted) {\n if (transcriptChanged) {\n this.logger.debug(\n { transcript: this.audioTranscript },\n 'triggering preemptive generation (FINAL_TRANSCRIPT)',\n );\n this.hooks.onPreemptiveGeneration({\n newTranscript: this.audioTranscript,\n transcriptConfidence:\n this.finalTranscriptConfidence.length > 0\n ? this.finalTranscriptConfidence.reduce((a, b) => a + b, 0) /\n this.finalTranscriptConfidence.length\n : 0,\n });\n }\n\n if (!this.speaking) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on stt FINAL_TRANSCRIPT');\n this.runEOUDetection(chatCtx);\n }\n }\n break;\n case SpeechEventType.PREFLIGHT_TRANSCRIPT:\n this.hooks.onInterimTranscript(ev);\n const preflightTranscript = ev.alternatives?.[0]?.text ?? '';\n const preflightConfidence = ev.alternatives?.[0]?.confidence ?? 0;\n const preflightLanguage = ev.alternatives?.[0]?.language;\n\n const MIN_LANGUAGE_DETECTION_LENGTH = 5;\n if (\n !this.lastLanguage ||\n (preflightLanguage && preflightTranscript.length > MIN_LANGUAGE_DETECTION_LENGTH)\n ) {\n this.lastLanguage = preflightLanguage;\n }\n\n if (!preflightTranscript) {\n return;\n }\n\n this.logger.debug(\n {\n user_transcript: preflightTranscript,\n language: this.lastLanguage,\n },\n 'received user preflight transcript',\n );\n\n // still need to increment it as it's used for turn detection,\n this.lastFinalTranscriptTime = Date.now();\n // preflight transcript includes all pre-committed transcripts (including final transcript from the previous STT run)\n this.audioPreflightTranscript =\n `${this.audioTranscript} ${preflightTranscript}`.trimStart();\n this.audioInterimTranscript = preflightTranscript;\n\n if (!this.vad || this.lastSpeakingTime === undefined) {\n // vad disabled, use stt timestamp\n this.lastSpeakingTime = Date.now();\n }\n\n if (this.turnDetectionMode !== 'manual' || this.userTurnCommitted) {\n const confidenceVals = [...this.finalTranscriptConfidence, preflightConfidence];\n this.logger.debug(\n {\n transcript:\n this.audioPreflightTranscript.length > 100\n ? this.audioPreflightTranscript.slice(0, 100) + '...'\n : this.audioPreflightTranscript,\n },\n 'triggering preemptive generation (PREFLIGHT_TRANSCRIPT)',\n );\n this.hooks.onPreemptiveGeneration({\n newTranscript: this.audioPreflightTranscript,\n transcriptConfidence:\n confidenceVals.length > 0\n ? confidenceVals.reduce((a, b) => a + b, 0) / confidenceVals.length\n : 0,\n });\n }\n break;\n case SpeechEventType.INTERIM_TRANSCRIPT:\n this.logger.debug({ transcript: ev.alternatives?.[0]?.text }, 'interim transcript');\n this.hooks.onInterimTranscript(ev);\n this.audioInterimTranscript = ev.alternatives?.[0]?.text ?? '';\n break;\n case SpeechEventType.START_OF_SPEECH:\n if (this.turnDetectionMode !== 'stt') break;\n this.hooks.onStartOfSpeech({\n type: VADEventType.START_OF_SPEECH,\n samplesIndex: 0,\n timestamp: Date.now(),\n speechDuration: 0,\n silenceDuration: 0,\n frames: [],\n probability: 0,\n inferenceDuration: 0,\n speaking: true,\n rawAccumulatedSilence: 0,\n rawAccumulatedSpeech: 0,\n });\n this.speaking = true;\n this.lastSpeakingTime = Date.now();\n\n this.bounceEOUTask?.cancel();\n break;\n case SpeechEventType.END_OF_SPEECH:\n if (this.turnDetectionMode !== 'stt') break;\n this.hooks.onEndOfSpeech({\n type: VADEventType.END_OF_SPEECH,\n samplesIndex: 0,\n timestamp: Date.now(),\n speechDuration: 0,\n silenceDuration: 0,\n frames: [],\n probability: 0,\n inferenceDuration: 0,\n speaking: false,\n rawAccumulatedSilence: 0,\n rawAccumulatedSpeech: 0,\n });\n this.speaking = false;\n this.userTurnCommitted = true;\n this.lastSpeakingTime = Date.now();\n\n if (!this.speaking) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on stt END_OF_SPEECH');\n this.runEOUDetection(chatCtx);\n }\n }\n }\n\n private runEOUDetection(chatCtx: ChatContext) {\n this.logger.debug(\n {\n stt: this.stt,\n audioTranscript: this.audioTranscript,\n turnDetectionMode: this.turnDetectionMode,\n },\n 'running EOU detection',\n );\n\n if (this.stt && !this.audioTranscript && this.turnDetectionMode !== 'manual') {\n // stt enabled but no transcript yet\n this.logger.debug('skipping EOU detection');\n return;\n }\n\n chatCtx = chatCtx.copy();\n chatCtx.addMessage({ role: 'user', content: this.audioTranscript });\n\n const turnDetector =\n // disable EOU model if manual turn detection enabled\n this.audioTranscript && this.turnDetectionMode !== 'manual' ? this.turnDetector : undefined;\n\n const bounceEOUTask =\n (\n lastSpeakingTime: number | undefined,\n lastFinalTranscriptTime: number,\n speechStartTime: number | undefined,\n ) =>\n async (controller: AbortController) => {\n let endpointingDelay = this.minEndpointingDelay;\n\n if (turnDetector) {\n this.logger.debug('Running turn detector model');\n if (!turnDetector.supportsLanguage(this.lastLanguage)) {\n this.logger.debug(`Turn detector does not support language ${this.lastLanguage}`);\n } else {\n const endOfTurnProbability = await turnDetector.predictEndOfTurn(chatCtx);\n this.logger.debug(\n { endOfTurnProbability, language: this.lastLanguage },\n 'end of turn probability',\n );\n\n const unlikelyThreshold = await turnDetector.unlikelyThreshold(this.lastLanguage);\n this.logger.debug(\n {\n unlikelyThreshold,\n endOfTurnProbability,\n language: this.lastLanguage,\n transcript: this.audioTranscript,\n },\n 'EOU Detection',\n );\n\n if (unlikelyThreshold && endOfTurnProbability < unlikelyThreshold) {\n endpointingDelay = this.maxEndpointingDelay;\n }\n }\n }\n\n let extraSleep = endpointingDelay;\n if (lastSpeakingTime !== undefined) {\n extraSleep += lastSpeakingTime - Date.now();\n }\n\n if (extraSleep > 0) {\n // add delay to see if there's a potential upcoming EOU task that cancels this one\n await delay(Math.max(extraSleep, 0), { signal: controller.signal });\n }\n\n this.logger.debug({ transcript: this.audioTranscript }, 'end of user turn');\n\n const confidenceAvg =\n this.finalTranscriptConfidence.length > 0\n ? this.finalTranscriptConfidence.reduce((a, b) => a + b, 0) /\n this.finalTranscriptConfidence.length\n : 0;\n\n let startedSpeakingAt: number | undefined;\n let stoppedSpeakingAt: number | undefined;\n let transcriptionDelay: number | undefined;\n let endOfUtteranceDelay: number | undefined;\n\n // sometimes, we can't calculate the metrics because VAD was unreliable.\n // in this case, we just ignore the calculation, it's better than providing likely wrong values\n if (\n lastFinalTranscriptTime !== 0 &&\n lastSpeakingTime !== undefined &&\n speechStartTime !== undefined\n ) {\n startedSpeakingAt = speechStartTime;\n stoppedSpeakingAt = lastSpeakingTime;\n transcriptionDelay = Math.max(lastFinalTranscriptTime - lastSpeakingTime, 0);\n endOfUtteranceDelay = Date.now() - lastSpeakingTime;\n }\n\n const committed = await this.hooks.onEndOfTurn({\n newTranscript: this.audioTranscript,\n transcriptConfidence: confidenceAvg,\n transcriptionDelay: transcriptionDelay ?? 0,\n endOfUtteranceDelay: endOfUtteranceDelay ?? 0,\n startedSpeakingAt,\n stoppedSpeakingAt,\n });\n\n if (committed) {\n // clear the transcript if the user turn was committed\n this.audioTranscript = '';\n this.finalTranscriptConfidence = [];\n this.lastSpeakingTime = undefined;\n this.lastFinalTranscriptTime = 0;\n this.speechStartTime = undefined;\n }\n\n this.userTurnCommitted = false;\n };\n\n // cancel any existing EOU task\n this.bounceEOUTask?.cancel();\n // copy the values before awaiting (the values can change)\n this.bounceEOUTask = Task.from(\n bounceEOUTask(this.lastSpeakingTime, this.lastFinalTranscriptTime, this.speechStartTime),\n );\n\n this.bounceEOUTask.result\n .then(() => {\n this.logger.debug('EOU detection task completed');\n })\n .catch((err: unknown) => {\n if (err instanceof Error && err.message.includes('This operation was aborted')) {\n // ignore aborted errors\n return;\n }\n this.logger.error(err, 'Error in EOU detection task:');\n });\n }\n\n private async createSttTask(stt: STTNode | undefined, signal: AbortSignal) {\n if (!stt) return;\n\n this.logger.debug('createSttTask: create stt stream from stt node');\n\n const sttStream = await stt(this.sttInputStream, {});\n\n if (signal.aborted || sttStream === null) return;\n\n if (sttStream instanceof ReadableStream) {\n const reader = sttStream.getReader();\n\n signal.addEventListener('abort', async () => {\n try {\n reader.releaseLock();\n await sttStream?.cancel();\n } catch (e) {\n this.logger.debug('createSttTask: error during abort handler:', e);\n }\n });\n\n try {\n while (true) {\n if (signal.aborted) break;\n\n const { done, value: ev } = await reader.read();\n if (done) break;\n\n if (typeof ev === 'string') {\n throw new Error('STT node must yield SpeechEvent');\n } else {\n await this.onSTTEvent(ev);\n }\n }\n } catch (e) {\n if (isStreamReaderReleaseError(e)) {\n return;\n }\n this.logger.error({ error: e }, 'createSttTask: error reading sttStream');\n } finally {\n reader.releaseLock();\n try {\n await sttStream.cancel();\n } catch (e) {\n this.logger.debug(\n 'createSttTask: error cancelling sttStream (may already be cancelled):',\n e,\n );\n }\n }\n }\n }\n\n private async createVadTask(vad: VAD | undefined, signal: AbortSignal) {\n if (!vad) return;\n\n const vadStream = vad.stream();\n vadStream.updateInputStream(this.vadInputStream);\n\n const abortHandler = () => {\n vadStream.detachInputStream();\n vadStream.close();\n signal.removeEventListener('abort', abortHandler);\n };\n signal.addEventListener('abort', abortHandler);\n\n try {\n for await (const ev of vadStream) {\n if (signal.aborted) break;\n\n switch (ev.type) {\n case VADEventType.START_OF_SPEECH:\n this.logger.debug('VAD task: START_OF_SPEECH');\n this.hooks.onStartOfSpeech(ev);\n this.speaking = true;\n\n // Capture sample rate from the first VAD event if not already set\n if (ev.frames.length > 0 && ev.frames[0]) {\n this.sampleRate = ev.frames[0].sampleRate;\n }\n\n this.bounceEOUTask?.cancel();\n break;\n case VADEventType.INFERENCE_DONE:\n this.hooks.onVADInferenceDone(ev);\n // for metrics, get the \"earliest\" signal of speech as possible\n if (ev.rawAccumulatedSpeech > 0.0) {\n this.lastSpeakingTime = Date.now();\n\n if (this.speechStartTime === undefined) {\n this.speechStartTime = Date.now();\n }\n }\n break;\n case VADEventType.END_OF_SPEECH:\n this.logger.debug('VAD task: END_OF_SPEECH');\n this.hooks.onEndOfSpeech(ev);\n\n // when VAD fires END_OF_SPEECH, it already waited for the silence_duration\n this.speaking = false;\n\n if (\n this.vadBaseTurnDetection ||\n (this.turnDetectionMode === 'stt' && this.userTurnCommitted)\n ) {\n const chatCtx = this.hooks.retrieveChatCtx();\n this.runEOUDetection(chatCtx);\n }\n break;\n }\n }\n } catch (e) {\n this.logger.error(e, 'Error in VAD task');\n } finally {\n this.logger.debug('VAD task closed');\n }\n }\n\n setInputAudioStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputAudioStream() {\n this.deferredInputStream.detachSource();\n }\n\n clearUserTurn() {\n this.audioTranscript = '';\n this.audioInterimTranscript = '';\n this.audioPreflightTranscript = '';\n this.finalTranscriptConfidence = [];\n this.userTurnCommitted = false;\n\n this.sttTask?.cancelAndWait().finally(() => {\n this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n this.sttTask.result.catch((err) => {\n this.logger.error(`Error running STT task: ${err}`);\n });\n });\n }\n\n commitUserTurn(audioDetached: boolean) {\n const commitUserTurnTask =\n (delayDuration: number = 500) =>\n async (controller: AbortController) => {\n if (Date.now() - this.lastFinalTranscriptTime > delayDuration) {\n // flush the stt by pushing silence\n if (audioDetached && this.sampleRate !== undefined) {\n const numSamples = Math.floor(this.sampleRate * 0.5);\n const silence = new Int16Array(numSamples * 2);\n const silenceFrame = new AudioFrame(silence, this.sampleRate, 1, numSamples);\n this.silenceAudioWriter.write(silenceFrame);\n }\n\n // wait for the final transcript to be available\n await delay(delayDuration, { signal: controller.signal });\n }\n\n if (this.audioInterimTranscript) {\n // append interim transcript in case the final transcript is not ready\n this.audioTranscript = `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n }\n this.audioInterimTranscript = '';\n\n const chatCtx = this.hooks.retrieveChatCtx();\n this.logger.debug('running EOU detection on commitUserTurn');\n this.runEOUDetection(chatCtx);\n this.userTurnCommitted = true;\n };\n\n // cancel any existing commit user turn task\n this.commitUserTurnTask?.cancel();\n this.commitUserTurnTask = Task.from(commitUserTurnTask());\n\n this.commitUserTurnTask.result\n .then(() => {\n this.logger.debug('User turn committed');\n })\n .catch((err: unknown) => {\n this.logger.error(err, 'Error in user turn commit task:');\n });\n }\n\n async close() {\n this.detachInputAudioStream();\n await this.commitUserTurnTask?.cancelAndWait();\n await this.sttTask?.cancelAndWait();\n await this.vadTask?.cancelAndWait();\n await this.bounceEOUTask?.cancelAndWait();\n }\n\n private get vadBaseTurnDetection() {\n return ['vad', undefined].includes(this.turnDetectionMode);\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAA2B;AAE3B,iBAA+B;AAC/B,0BAAiC;AACjC,iBAAoB;AACpB,6BAAmE;AACnE,gCAAkC;AAClC,oCAAqC;AACrC,iBAAkD;AAClD,mBAA4B;AAC5B,iBAAsD;AA8C/C,MAAM,iBAAiB;AAAA,EACpB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA,aAAS,gBAAI;AAAA,EACb,0BAA0B;AAAA,EAC1B,kBAAkB;AAAA,EAClB,yBAAyB;AAAA,EACzB,2BAA2B;AAAA,EAC3B,4BAAsC,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA,oBAAoB;AAAA,EACpB,WAAW;AAAA,EACX;AAAA,EAEA;AAAA,EACA;AAAA,EACA,wBAAwB,IAAI,4CAA8B;AAAA,EAC1D;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,MAA+B;AACzC,SAAK,QAAQ,KAAK;AAClB,SAAK,MAAM,KAAK;AAChB,SAAK,MAAM,KAAK;AAChB,SAAK,eAAe,KAAK;AACzB,SAAK,oBAAoB,KAAK;AAC9B,SAAK,sBAAsB,KAAK;AAChC,SAAK,sBAAsB,KAAK;AAChC,SAAK,eAAe;AAEpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,UAAM,CAAC,gBAAgB,cAAc,IAAI,KAAK,oBAAoB,OAAO,IAAI;AAC7E,SAAK,iBAAiB;AACtB,SAAK,qBAAiB,oDAAqB,gBAAgB,KAAK,sBAAsB,QAAQ;AAC9F,SAAK,qBAAqB,KAAK,sBAAsB,SAAS,UAAU;AAAA,EAC1E;AAAA;AAAA;AAAA;AAAA,EAKA,IAAI,oBAA4B;AAC9B,QAAI,KAAK,wBAAwB;AAC/B,aAAO,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,IACvE;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAED,SAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,WAAW,IAAiB;AApI5C;AAqII,QACE,KAAK,sBAAsB,YAC3B,KAAK,sBACJ,KAAK,kBAAkB,UACtB,KAAK,cAAc,QACnB,GAAG,QAAQ,2BAAgB,qBAC7B;AAGA,WAAK,OAAO;AAAA,QACV;AAAA,UACE,mBAAmB,KAAK;AAAA,UACxB,cAAa,UAAK,kBAAL,mBAAoB;AAAA,UACjC,QAAQ,GAAG;AAAA,UACX,mBAAmB,KAAK;AAAA,QAC1B;AAAA,QACA;AAAA,MACF;AACA;AAAA,IACF;AAEA,YAAQ,GAAG,MAAM;AAAA,MACf,KAAK,2BAAgB;AACnB,aAAK,MAAM,kBAAkB,EAAE;AAC/B,cAAM,cAAa,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AACzC,cAAM,eAAa,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,eAAc;AACvD,aAAK,gBAAe,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AAE1C,YAAI,CAAC,YAAY;AAEf;AAAA,QACF;AAEA,aAAK,OAAO;AAAA,UACV;AAAA,YACE,iBAAiB;AAAA,YACjB,UAAU,KAAK;AAAA,UACjB;AAAA,UACA;AAAA,QACF;AAEA,aAAK,0BAA0B,KAAK,IAAI;AACxC,aAAK,mBAAmB,IAAI,UAAU;AACtC,aAAK,kBAAkB,KAAK,gBAAgB,UAAU;AACtD,aAAK,0BAA0B,KAAK,UAAU;AAC9C,cAAM,oBAAoB,KAAK,oBAAoB,KAAK;AACxD,aAAK,yBAAyB;AAC9B,aAAK,2BAA2B;AAEhC,YAAI,CAAC,KAAK,OAAO,KAAK,qBAAqB,QAAW;AAMpD,eAAK,mBAAmB,KAAK,IAAI;AAAA,QACnC;AAEA,YAAI,KAAK,wBAAwB,KAAK,mBAAmB;AACvD,cAAI,mBAAmB;AACrB,iBAAK,OAAO;AAAA,cACV,EAAE,YAAY,KAAK,gBAAgB;AAAA,cACnC;AAAA,YACF;AACA,iBAAK,MAAM,uBAAuB;AAAA,cAChC,eAAe,KAAK;AAAA,cACpB,sBACE,KAAK,0BAA0B,SAAS,IACpC,KAAK,0BAA0B,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IACxD,KAAK,0BAA0B,SAC/B;AAAA,YACR,CAAC;AAAA,UACH;AAEA,cAAI,CAAC,KAAK,UAAU;AAClB,kBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,iBAAK,OAAO,MAAM,+CAA+C;AACjE,iBAAK,gBAAgB,OAAO;AAAA,UAC9B;AAAA,QACF;AACA;AAAA,MACF,KAAK,2BAAgB;AACnB,aAAK,MAAM,oBAAoB,EAAE;AACjC,cAAM,wBAAsB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,SAAQ;AAC1D,cAAM,wBAAsB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,eAAc;AAChE,cAAM,qBAAoB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AAEhD,cAAM,gCAAgC;AACtC,YACE,CAAC,KAAK,gBACL,qBAAqB,oBAAoB,SAAS,+BACnD;AACA,eAAK,eAAe;AAAA,QACtB;AAEA,YAAI,CAAC,qBAAqB;AACxB;AAAA,QACF;AAEA,aAAK,OAAO;AAAA,UACV;AAAA,YACE,iBAAiB;AAAA,YACjB,UAAU,KAAK;AAAA,UACjB;AAAA,UACA;AAAA,QACF;AAGA,aAAK,0BAA0B,KAAK,IAAI;AAExC,aAAK,2BACH,GAAG,KAAK,eAAe,IAAI,mBAAmB,GAAG,UAAU;AAC7D,aAAK,yBAAyB;AAE9B,YAAI,CAAC,KAAK,OAAO,KAAK,qBAAqB,QAAW;AAEpD,eAAK,mBAAmB,KAAK,IAAI;AAAA,QACnC;AAEA,YAAI,KAAK,sBAAsB,YAAY,KAAK,mBAAmB;AACjE,gBAAM,iBAAiB,CAAC,GAAG,KAAK,2BAA2B,mBAAmB;AAC9E,eAAK,OAAO;AAAA,YACV;AAAA,cACE,YACE,KAAK,yBAAyB,SAAS,MACnC,KAAK,yBAAyB,MAAM,GAAG,GAAG,IAAI,QAC9C,KAAK;AAAA,YACb;AAAA,YACA;AAAA,UACF;AACA,eAAK,MAAM,uBAAuB;AAAA,YAChC,eAAe,KAAK;AAAA,YACpB,sBACE,eAAe,SAAS,IACpB,eAAe,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,eAAe,SAC3D;AAAA,UACR,CAAC;AAAA,QACH;AACA;AAAA,MACF,KAAK,2BAAgB;AACnB,aAAK,OAAO,MAAM,EAAE,aAAY,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,KAAK,GAAG,oBAAoB;AAClF,aAAK,MAAM,oBAAoB,EAAE;AACjC,aAAK,2BAAyB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,SAAQ;AAC5D;AAAA,MACF,KAAK,2BAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC,aAAK,MAAM,gBAAgB;AAAA,UACzB,MAAM,wBAAa;AAAA,UACnB,cAAc;AAAA,UACd,WAAW,KAAK,IAAI;AAAA,UACpB,gBAAgB;AAAA,UAChB,iBAAiB;AAAA,UACjB,QAAQ,CAAC;AAAA,UACT,aAAa;AAAA,UACb,mBAAmB;AAAA,UACnB,UAAU;AAAA,UACV,uBAAuB;AAAA,UACvB,sBAAsB;AAAA,QACxB,CAAC;AACD,aAAK,WAAW;AAChB,aAAK,mBAAmB,KAAK,IAAI;AAEjC,mBAAK,kBAAL,mBAAoB;AACpB;AAAA,MACF,KAAK,2BAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC,aAAK,MAAM,cAAc;AAAA,UACvB,MAAM,wBAAa;AAAA,UACnB,cAAc;AAAA,UACd,WAAW,KAAK,IAAI;AAAA,UACpB,gBAAgB;AAAA,UAChB,iBAAiB;AAAA,UACjB,QAAQ,CAAC;AAAA,UACT,aAAa;AAAA,UACb,mBAAmB;AAAA,UACnB,UAAU;AAAA,UACV,uBAAuB;AAAA,UACvB,sBAAsB;AAAA,QACxB,CAAC;AACD,aAAK,WAAW;AAChB,aAAK,oBAAoB;AACzB,aAAK,mBAAmB,KAAK,IAAI;AAEjC,YAAI,CAAC,KAAK,UAAU;AAClB,gBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,eAAK,OAAO,MAAM,4CAA4C;AAC9D,eAAK,gBAAgB,OAAO;AAAA,QAC9B;AAAA,IACJ;AAAA,EACF;AAAA,EAEQ,gBAAgB,SAAsB;AApUhD;AAqUI,SAAK,OAAO;AAAA,MACV;AAAA,QACE,KAAK,KAAK;AAAA,QACV,iBAAiB,KAAK;AAAA,QACtB,mBAAmB,KAAK;AAAA,MAC1B;AAAA,MACA;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,CAAC,KAAK,mBAAmB,KAAK,sBAAsB,UAAU;AAE5E,WAAK,OAAO,MAAM,wBAAwB;AAC1C;AAAA,IACF;AAEA,cAAU,QAAQ,KAAK;AACvB,YAAQ,WAAW,EAAE,MAAM,QAAQ,SAAS,KAAK,gBAAgB,CAAC;AAElE,UAAM;AAAA;AAAA,MAEJ,KAAK,mBAAmB,KAAK,sBAAsB,WAAW,KAAK,eAAe;AAAA;AAEpF,UAAM,gBACJ,CACE,kBACA,yBACA,oBAEF,OAAO,eAAgC;AACrC,UAAI,mBAAmB,KAAK;AAE5B,UAAI,cAAc;AAChB,aAAK,OAAO,MAAM,6BAA6B;AAC/C,YAAI,CAAC,aAAa,iBAAiB,KAAK,YAAY,GAAG;AACrD,eAAK,OAAO,MAAM,2CAA2C,KAAK,YAAY,EAAE;AAAA,QAClF,OAAO;AACL,gBAAM,uBAAuB,MAAM,aAAa,iBAAiB,OAAO;AACxE,eAAK,OAAO;AAAA,YACV,EAAE,sBAAsB,UAAU,KAAK,aAAa;AAAA,YACpD;AAAA,UACF;AAEA,gBAAM,oBAAoB,MAAM,aAAa,kBAAkB,KAAK,YAAY;AAChF,eAAK,OAAO;AAAA,YACV;AAAA,cACE;AAAA,cACA;AAAA,cACA,UAAU,KAAK;AAAA,cACf,YAAY,KAAK;AAAA,YACnB;AAAA,YACA;AAAA,UACF;AAEA,cAAI,qBAAqB,uBAAuB,mBAAmB;AACjE,+BAAmB,KAAK;AAAA,UAC1B;AAAA,QACF;AAAA,MACF;AAEA,UAAI,aAAa;AACjB,UAAI,qBAAqB,QAAW;AAClC,sBAAc,mBAAmB,KAAK,IAAI;AAAA,MAC5C;AAEA,UAAI,aAAa,GAAG;AAElB,kBAAM,oBAAM,KAAK,IAAI,YAAY,CAAC,GAAG,EAAE,QAAQ,WAAW,OAAO,CAAC;AAAA,MACpE;AAEA,WAAK,OAAO,MAAM,EAAE,YAAY,KAAK,gBAAgB,GAAG,kBAAkB;AAE1E,YAAM,gBACJ,KAAK,0BAA0B,SAAS,IACpC,KAAK,0BAA0B,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IACxD,KAAK,0BAA0B,SAC/B;AAEN,UAAI;AACJ,UAAI;AACJ,UAAI;AACJ,UAAI;AAIJ,UACE,4BAA4B,KAC5B,qBAAqB,UACrB,oBAAoB,QACpB;AACA,4BAAoB;AACpB,4BAAoB;AACpB,6BAAqB,KAAK,IAAI,0BAA0B,kBAAkB,CAAC;AAC3E,8BAAsB,KAAK,IAAI,IAAI;AAAA,MACrC;AAEA,YAAM,YAAY,MAAM,KAAK,MAAM,YAAY;AAAA,QAC7C,eAAe,KAAK;AAAA,QACpB,sBAAsB;AAAA,QACtB,oBAAoB,sBAAsB;AAAA,QAC1C,qBAAqB,uBAAuB;AAAA,QAC5C;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,WAAW;AAEb,aAAK,kBAAkB;AACvB,aAAK,4BAA4B,CAAC;AAClC,aAAK,mBAAmB;AACxB,aAAK,0BAA0B;AAC/B,aAAK,kBAAkB;AAAA,MACzB;AAEA,WAAK,oBAAoB;AAAA,IAC3B;AAGF,eAAK,kBAAL,mBAAoB;AAEpB,SAAK,gBAAgB,kBAAK;AAAA,MACxB,cAAc,KAAK,kBAAkB,KAAK,yBAAyB,KAAK,eAAe;AAAA,IACzF;AAEA,SAAK,cAAc,OAChB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,8BAA8B;AAAA,IAClD,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,UAAI,eAAe,SAAS,IAAI,QAAQ,SAAS,4BAA4B,GAAG;AAE9E;AAAA,MACF;AACA,WAAK,OAAO,MAAM,KAAK,8BAA8B;AAAA,IACvD,CAAC;AAAA,EACL;AAAA,EAEA,MAAc,cAAc,KAA0B,QAAqB;AACzE,QAAI,CAAC,IAAK;AAEV,SAAK,OAAO,MAAM,gDAAgD;AAElE,UAAM,YAAY,MAAM,IAAI,KAAK,gBAAgB,CAAC,CAAC;AAEnD,QAAI,OAAO,WAAW,cAAc,KAAM;AAE1C,QAAI,qBAAqB,2BAAgB;AACvC,YAAM,SAAS,UAAU,UAAU;AAEnC,aAAO,iBAAiB,SAAS,YAAY;AAC3C,YAAI;AACF,iBAAO,YAAY;AACnB,iBAAM,uCAAW;AAAA,QACnB,SAAS,GAAG;AACV,eAAK,OAAO,MAAM,8CAA8C,CAAC;AAAA,QACnE;AAAA,MACF,CAAC;AAED,UAAI;AACF,eAAO,MAAM;AACX,cAAI,OAAO,QAAS;AAEpB,gBAAM,EAAE,MAAM,OAAO,GAAG,IAAI,MAAM,OAAO,KAAK;AAC9C,cAAI,KAAM;AAEV,cAAI,OAAO,OAAO,UAAU;AAC1B,kBAAM,IAAI,MAAM,iCAAiC;AAAA,UACnD,OAAO;AACL,kBAAM,KAAK,WAAW,EAAE;AAAA,UAC1B;AAAA,QACF;AAAA,MACF,SAAS,GAAG;AACV,gBAAI,mDAA2B,CAAC,GAAG;AACjC;AAAA,QACF;AACA,aAAK,OAAO,MAAM,EAAE,OAAO,EAAE,GAAG,wCAAwC;AAAA,MAC1E,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,UAAU,OAAO;AAAA,QACzB,SAAS,GAAG;AACV,eAAK,OAAO;AAAA,YACV;AAAA,YACA;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,cAAc,KAAsB,QAAqB;AAlgBzE;AAmgBI,QAAI,CAAC,IAAK;AAEV,UAAM,YAAY,IAAI,OAAO;AAC7B,cAAU,kBAAkB,KAAK,cAAc;AAE/C,UAAM,eAAe,MAAM;AACzB,gBAAU,kBAAkB;AAC5B,gBAAU,MAAM;AAChB,aAAO,oBAAoB,SAAS,YAAY;AAAA,IAClD;AACA,WAAO,iBAAiB,SAAS,YAAY;AAE7C,QAAI;AACF,uBAAiB,MAAM,WAAW;AAChC,YAAI,OAAO,QAAS;AAEpB,gBAAQ,GAAG,MAAM;AAAA,UACf,KAAK,wBAAa;AAChB,iBAAK,OAAO,MAAM,2BAA2B;AAC7C,iBAAK,MAAM,gBAAgB,EAAE;AAC7B,iBAAK,WAAW;AAGhB,gBAAI,GAAG,OAAO,SAAS,KAAK,GAAG,OAAO,CAAC,GAAG;AACxC,mBAAK,aAAa,GAAG,OAAO,CAAC,EAAE;AAAA,YACjC;AAEA,uBAAK,kBAAL,mBAAoB;AACpB;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,MAAM,mBAAmB,EAAE;AAEhC,gBAAI,GAAG,uBAAuB,GAAK;AACjC,mBAAK,mBAAmB,KAAK,IAAI;AAEjC,kBAAI,KAAK,oBAAoB,QAAW;AACtC,qBAAK,kBAAkB,KAAK,IAAI;AAAA,cAClC;AAAA,YACF;AACA;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,OAAO,MAAM,yBAAyB;AAC3C,iBAAK,MAAM,cAAc,EAAE;AAG3B,iBAAK,WAAW;AAEhB,gBACE,KAAK,wBACJ,KAAK,sBAAsB,SAAS,KAAK,mBAC1C;AACA,oBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,mBAAK,gBAAgB,OAAO;AAAA,YAC9B;AACA;AAAA,QACJ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,GAAG,mBAAmB;AAAA,IAC1C,UAAE;AACA,WAAK,OAAO,MAAM,iBAAiB;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,oBAAoB,aAAyC;AAC3D,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,yBAAyB;AACvB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA,EAEA,gBAAgB;AA3kBlB;AA4kBI,SAAK,kBAAkB;AACvB,SAAK,yBAAyB;AAC9B,SAAK,2BAA2B;AAChC,SAAK,4BAA4B,CAAC;AAClC,SAAK,oBAAoB;AAEzB,eAAK,YAAL,mBAAc,gBAAgB,QAAQ,MAAM;AAC1C,WAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,WAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,aAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,MACpD,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,eAAe,eAAwB;AA1lBzC;AA2lBI,UAAM,qBACJ,CAAC,gBAAwB,QACzB,OAAO,eAAgC;AACrC,UAAI,KAAK,IAAI,IAAI,KAAK,0BAA0B,eAAe;AAE7D,YAAI,iBAAiB,KAAK,eAAe,QAAW;AAClD,gBAAM,aAAa,KAAK,MAAM,KAAK,aAAa,GAAG;AACnD,gBAAM,UAAU,IAAI,WAAW,aAAa,CAAC;AAC7C,gBAAM,eAAe,IAAI,2BAAW,SAAS,KAAK,YAAY,GAAG,UAAU;AAC3E,eAAK,mBAAmB,MAAM,YAAY;AAAA,QAC5C;AAGA,kBAAM,oBAAM,eAAe,EAAE,QAAQ,WAAW,OAAO,CAAC;AAAA,MAC1D;AAEA,UAAI,KAAK,wBAAwB;AAE/B,aAAK,kBAAkB,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,MACvF;AACA,WAAK,yBAAyB;AAE9B,YAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,WAAK,OAAO,MAAM,yCAAyC;AAC3D,WAAK,gBAAgB,OAAO;AAC5B,WAAK,oBAAoB;AAAA,IAC3B;AAGF,eAAK,uBAAL,mBAAyB;AACzB,SAAK,qBAAqB,kBAAK,KAAK,mBAAmB,CAAC;AAExD,SAAK,mBAAmB,OACrB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,qBAAqB;AAAA,IACzC,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,WAAK,OAAO,MAAM,KAAK,iCAAiC;AAAA,IAC1D,CAAC;AAAA,EACL;AAAA,EAEA,MAAM,QAAQ;AApoBhB;AAqoBI,SAAK,uBAAuB;AAC5B,YAAM,UAAK,uBAAL,mBAAyB;AAC/B,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,kBAAL,mBAAoB;AAAA,EAC5B;AAAA,EAEA,IAAY,uBAAuB;AACjC,WAAO,CAAC,OAAO,MAAS,EAAE,SAAS,KAAK,iBAAiB;AAAA,EAC3D;AACF;","names":[]}
|
|
@@ -8,8 +8,15 @@ import type { TurnDetectionMode } from './agent_session.js';
|
|
|
8
8
|
import type { STTNode } from './io.js';
|
|
9
9
|
export interface EndOfTurnInfo {
|
|
10
10
|
newTranscript: string;
|
|
11
|
+
transcriptConfidence: number;
|
|
11
12
|
transcriptionDelay: number;
|
|
12
13
|
endOfUtteranceDelay: number;
|
|
14
|
+
startedSpeakingAt: number | undefined;
|
|
15
|
+
stoppedSpeakingAt: number | undefined;
|
|
16
|
+
}
|
|
17
|
+
export interface PreemptiveGenerationInfo {
|
|
18
|
+
newTranscript: string;
|
|
19
|
+
transcriptConfidence: number;
|
|
13
20
|
}
|
|
14
21
|
export interface RecognitionHooks {
|
|
15
22
|
onStartOfSpeech: (ev: VADEvent) => void;
|
|
@@ -18,6 +25,7 @@ export interface RecognitionHooks {
|
|
|
18
25
|
onInterimTranscript: (ev: SpeechEvent) => void;
|
|
19
26
|
onFinalTranscript: (ev: SpeechEvent) => void;
|
|
20
27
|
onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;
|
|
28
|
+
onPreemptiveGeneration: (info: PreemptiveGenerationInfo) => void;
|
|
21
29
|
retrieveChatCtx: () => ChatContext;
|
|
22
30
|
}
|
|
23
31
|
export interface _TurnDetector {
|
|
@@ -48,7 +56,10 @@ export declare class AudioRecognition {
|
|
|
48
56
|
private lastFinalTranscriptTime;
|
|
49
57
|
private audioTranscript;
|
|
50
58
|
private audioInterimTranscript;
|
|
59
|
+
private audioPreflightTranscript;
|
|
60
|
+
private finalTranscriptConfidence;
|
|
51
61
|
private lastSpeakingTime;
|
|
62
|
+
private speechStartTime;
|
|
52
63
|
private userTurnCommitted;
|
|
53
64
|
private speaking;
|
|
54
65
|
private sampleRate?;
|
|
@@ -8,8 +8,15 @@ import type { TurnDetectionMode } from './agent_session.js';
|
|
|
8
8
|
import type { STTNode } from './io.js';
|
|
9
9
|
export interface EndOfTurnInfo {
|
|
10
10
|
newTranscript: string;
|
|
11
|
+
transcriptConfidence: number;
|
|
11
12
|
transcriptionDelay: number;
|
|
12
13
|
endOfUtteranceDelay: number;
|
|
14
|
+
startedSpeakingAt: number | undefined;
|
|
15
|
+
stoppedSpeakingAt: number | undefined;
|
|
16
|
+
}
|
|
17
|
+
export interface PreemptiveGenerationInfo {
|
|
18
|
+
newTranscript: string;
|
|
19
|
+
transcriptConfidence: number;
|
|
13
20
|
}
|
|
14
21
|
export interface RecognitionHooks {
|
|
15
22
|
onStartOfSpeech: (ev: VADEvent) => void;
|
|
@@ -18,6 +25,7 @@ export interface RecognitionHooks {
|
|
|
18
25
|
onInterimTranscript: (ev: SpeechEvent) => void;
|
|
19
26
|
onFinalTranscript: (ev: SpeechEvent) => void;
|
|
20
27
|
onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;
|
|
28
|
+
onPreemptiveGeneration: (info: PreemptiveGenerationInfo) => void;
|
|
21
29
|
retrieveChatCtx: () => ChatContext;
|
|
22
30
|
}
|
|
23
31
|
export interface _TurnDetector {
|
|
@@ -48,7 +56,10 @@ export declare class AudioRecognition {
|
|
|
48
56
|
private lastFinalTranscriptTime;
|
|
49
57
|
private audioTranscript;
|
|
50
58
|
private audioInterimTranscript;
|
|
59
|
+
private audioPreflightTranscript;
|
|
60
|
+
private finalTranscriptConfidence;
|
|
51
61
|
private lastSpeakingTime;
|
|
62
|
+
private speechStartTime;
|
|
52
63
|
private userTurnCommitted;
|
|
53
64
|
private speaking;
|
|
54
65
|
private sampleRate?;
|