@livekit/protocol 1.50.2 → 1.50.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livekit/protocol",
-  "version": "1.50.2",
+  "version": "1.50.4",
   "description": "",
   "type": "module",
   "require": "dist/index.cjs",

package/src/gen/livekit_agent_simulation_pb.d.ts

@@ -452,11 +452,15 @@ export declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobM
    *   accuracy   = task_completion x3 + stt.keyterm_recall + (1 - stt.wer) + tts.enunciation_score
    *   experience = conversation.turn_taking_score x2 + tts.naturalness_score + llm.conciseness_score
    *
+   * absent when the simulator spoiled the call
+   *
    * @generated from field: optional float accuracy_score = 1;
    */
   accuracyScore?: number;
 
   /**
+   * absent when the simulator spoiled the call
+   *
    * @generated from field: optional float experience_score = 2;
    */
   experienceScore?: number;
@@ -464,6 +468,8 @@ export declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobM
   /**
    * The accuracy anchor: the scenario verdict as 1/0, or the fraction of the declared target state the call reached.
    *
+   * absent when the simulator spoiled the call
+   *
    * @generated from field: optional float task_completion = 3;
    */
   taskCompletion?: number;
@@ -522,7 +528,7 @@ export declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobM
   hasRemoteSession: boolean;
 
   /**
-   * call start (first speech edge); turn times relative to it
+   * audio only; call start (first speech edge); turn times relative to it
    *
    * @generated from field: google.protobuf.Timestamp t0 = 13;
    */
@@ -544,66 +550,76 @@ export declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobM
 }
 
 /**
- * The agent's perception of the caller.
+ * The agent's perception of the caller. Audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.STT
  */
 export declare class SimulationRun_JobMetrics_STT extends Message<SimulationRun_JobMetrics_STT> {
   /**
-   * word error rate, pooled: word_errors / words
+   * needs the RemoteSession; word error rate, pooled: word_errors / words
    *
    * @generated from field: optional float wer = 1;
    */
   wer?: number;
 
   /**
-   * pooling stats: run WER = sum errors / sum words
+   * needs the RemoteSession; pooling stats: run WER = sum errors / sum words
    *
    * @generated from field: optional uint32 words = 2;
    */
   words?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 word_errors = 3;
    */
   wordErrors?: number;
 
   /**
-   * character error rate, pooled
+   * needs the RemoteSession; character error rate, pooled
    *
    * @generated from field: optional float cer = 4;
    */
   cer?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 chars = 9;
    */
   chars?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 char_errors = 10;
    */
   charErrors?: number;
 
   /**
-   * recall of uttered key entities (names, IDs, amounts)
+   * needs the RemoteSession; recall of uttered key entities (names, IDs, amounts)
    *
    * @generated from field: optional float keyterm_recall = 5;
    */
   keytermRecall?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 keyterms_uttered = 6;
    */
   keytermsUttered?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 keyterms_recognized = 7;
    */
   keytermsRecognized?: number;
 
   /**
-   * agent-reported STT/endpointing delay, mean
+   * needs the RemoteSession; agent-reported STT/endpointing delay, mean
    *
    * @generated from field: optional uint32 transcription_latency_ms = 8;
    */
@@ -629,28 +645,28 @@ export declare class SimulationRun_JobMetrics_STT extends Message<SimulationRun_
  */
 export declare class SimulationRun_JobMetrics_LLM extends Message<SimulationRun_JobMetrics_LLM> {
   /**
-   * time to first token, mean (per-turn in turns)
+   * needs the RemoteSession; time to first token, mean (per-turn in turns)
    *
    * @generated from field: optional uint32 ttft_ms = 1;
    */
   ttftMs?: number;
 
   /**
-   * time to first sentence — the smallest speakable unit
+   * needs the RemoteSession; time to first sentence (the smallest speakable unit)
    *
    * @generated from field: optional uint32 ttfs_ms = 2;
    */
   ttfsMs?: number;
 
   /**
-   * decode rate, mean
+   * needs the RemoteSession; decode rate, mean (per-turn in turns.llm)
    *
    * @generated from field: optional float tokens_per_second = 3;
    */
   tokensPerSecond?: number;
 
   /**
-   * judged 0-1: brief enough for voice
+   * audio only; needs the audio judge; judged 0-1: brief enough for voice
    *
    * @generated from field: optional float conciseness_score = 4;
    */
@@ -672,57 +688,63 @@ export declare class SimulationRun_JobMetrics_LLM extends Message<SimulationRun_
 }
 
 /**
- * The agent's voice.
+ * The agent's voice. audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.TTS
  */
 export declare class SimulationRun_JobMetrics_TTS extends Message<SimulationRun_JobMetrics_TTS> {
   /**
-   * Time to first AUDIO — the user-relevant signal. Some providers stream
-   * bytes before actual speech, so byte-level TTFB alone over-promises;
-   * it is kept below for disambiguation.
+   * needs the RemoteSession
    *
    * @generated from field: optional uint32 ttfa_ms = 1;
    */
   ttfaMs?: number;
 
   /**
-   * provider byte-level TTFB, agent-reported
+   * needs the RemoteSession; provider byte-level TTFB
    *
    * @generated from field: optional uint32 ttfb_ms = 2;
    */
   ttfbMs?: number;
 
   /**
-   * word error rate, pooled: word_errors / words
+   * needs the RemoteSession; word error rate, pooled: word_errors / words
    *
    * @generated from field: optional float wer = 3;
    */
   wer?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 words = 7;
    */
   words?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 word_errors = 8;
    */
   wordErrors?: number;
 
   /**
-   * character error rate, pooled
+   * needs the RemoteSession; character error rate, pooled
    *
    * @generated from field: optional float cer = 9;
    */
   cer?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 chars = 10;
    */
   chars?: number;
 
   /**
+   * needs the RemoteSession
+   *
    * @generated from field: optional uint32 char_errors = 11;
    */
   charErrors?: number;
@@ -735,14 +757,28 @@ export declare class SimulationRun_JobMetrics_TTS extends Message<SimulationRun_
   speechRateWpm?: number;
 
   /**
-   * judged 0-1: prosody / expressiveness
+   * pooling stats for speech_rate_wpm: raw heard words (not WER tokens)
+   *
+   * @generated from field: optional uint32 heard_words = 12;
+   */
+  heardWords?: number;
+
+  /**
+   * agent voiced time, sum of raw VAD segments
+   *
+   * @generated from field: optional uint32 speech_ms = 13;
+   */
+  speechMs?: number;
+
+  /**
+   * needs the audio judge; judged 0-1: prosody / expressiveness
    *
    * @generated from field: optional float naturalness_score = 5;
    */
   naturalnessScore?: number;
 
   /**
-   * judged 0-1: key entities audibly intact
+   * needs the audio judge; judged 0-1: key entities audibly intact
    *
    * @generated from field: optional float enunciation_score = 6;
    */
@@ -764,8 +800,7 @@ export declare class SimulationRun_JobMetrics_TTS extends Message<SimulationRun_
 }
 
 /**
- * End-to-end floor dynamics, measured off the audio timeline — the ground
- * truth of what the caller hears, independent of agent self-reports.
+ * Audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.Conversation
  */
@@ -795,14 +830,21 @@ export declare class SimulationRun_JobMetrics_Conversation extends Message<Simul
   responseLatencyP99Ms?: number;
 
   /**
-   * how long the agent talks after a barge-in, mean
+   * one turn's floor-transfer offset, cut-in if negative; job/run serve the percentiles
+   *
+   * @generated from field: optional int32 response_latency_ms = 16;
+   */
+  responseLatencyMs?: number;
+
+  /**
+   * agent audio continuing past a barge-in — per turn: on the barging persona row (presence = barge-in); job: mean
    *
    * @generated from field: optional uint32 agent_yield_latency_ms = 5;
    */
   agentYieldLatencyMs?: number;
 
   /**
-   * agent started before the caller's turn ended
+   * agent started before the caller's turn ended; 0/1 per turn
    *
    * @generated from field: optional uint32 eot_misprediction_count = 6;
    */
@@ -835,35 +877,35 @@ export declare class SimulationRun_JobMetrics_Conversation extends Message<Simul
   silenceTotalMs?: number;
 
   /**
-   * gaps past the natural-pause threshold
+   * gaps past the natural-pause threshold; 0/1 per turn
    *
    * @generated from field: optional uint32 awkward_silence_count = 11;
    */
   awkwardSilenceCount?: number;
 
   /**
-   * the simulated party spoke, the agent never responded
+   * the simulated party spoke, the agent never responded; 0/1 per turn
    *
    * @generated from field: optional uint32 unanswered_persona_turns = 12;
    */
   unansweredPersonaTurns?: number;
 
   /**
-   * agent paused for a non-interruption
+   * needs the RemoteSession; agent paused for a non-interruption
    *
    * @generated from field: optional uint32 false_interruption_count = 13;
    */
   falseInterruptionCount?: number;
 
   /**
-   * of those, never resumed
+   * needs the RemoteSession; of those, never resumed
    *
    * @generated from field: optional uint32 false_interruption_unrecovered_count = 14;
    */
   falseInterruptionUnrecoveredCount?: number;
 
   /**
-   * the agent's own claim, mean
+   * needs the RemoteSession; the agent's own claim, mean
    *
    * @generated from field: optional uint32 agent_reported_e2e_latency_ms = 15;
    */
@@ -922,6 +964,8 @@ export declare class SimulationRun_JobMetrics_Simulator extends Message<Simulati
 }
 
 /**
+ * Rows exist for text and audio;
+ *
  * @generated from message livekit.SimulationRun.JobMetrics.Turn
  */
 export declare class SimulationRun_JobMetrics_Turn extends Message<SimulationRun_JobMetrics_Turn> {
@@ -940,78 +984,48 @@ export declare class SimulationRun_JobMetrics_Turn extends Message<SimulationRun
   role: ChatRole;
 
   /**
-   * relative to t0; unset if the turn could not be aligned to the audio
+   * audio only; relative to t0; unset if the turn could not be aligned to the audio
    *
    * @generated from field: optional uint32 start_ms = 3;
    */
   startMs?: number;
 
   /**
+   * audio only
+   *
    * @generated from field: optional uint32 end_ms = 4;
    */
   endMs?: number;
 
   /**
-   * Agent-turn measurements; a negative response latency is a cut-in (the
-   * agent started before the caller's turn ended).
+   * Job groups hold means/pools of these; a field with no per-turn measurement stays unset here.
    *
-   * @generated from field: optional int32 response_latency_ms = 5;
-   */
-  responseLatencyMs?: number;
-
-  /**
-   * @generated from field: optional uint32 transcription_delay_ms = 6;
-   */
-  transcriptionDelayMs?: number;
-
-  /**
-   * @generated from field: optional uint32 llm_ttft_ms = 7;
-   */
-  llmTtftMs?: number;
-
-  /**
-   * @generated from field: optional uint32 llm_ttfs_ms = 8;
-   */
-  llmTtfsMs?: number;
-
-  /**
-   * @generated from field: optional uint32 tts_ttfa_ms = 9;
-   */
-  ttsTtfaMs?: number;
-
-  /**
-   * @generated from field: optional uint32 tts_ttfb_ms = 10;
-   */
-  ttsTtfbMs?: number;
-
-  /**
-   * @generated from field: optional uint32 agent_reported_e2e_latency_ms = 11;
-   */
-  agentReportedE2eLatencyMs?: number;
-
-  /**
-   * @generated from field: optional float conciseness_score = 12;
+   * transcription_latency_ms
+   *
+   * @generated from field: livekit.SimulationRun.JobMetrics.STT stt = 5;
    */
-  concisenessScore?: number;
+  stt?: SimulationRun_JobMetrics_STT;
 
   /**
-   * @generated from field: optional float naturalness_score = 13;
+   * ttft_ms, ttfs_ms, tokens_per_second, conciseness_score
+   *
+   * @generated from field: livekit.SimulationRun.JobMetrics.LLM llm = 6;
    */
-  naturalnessScore?: number;
+  llm?: SimulationRun_JobMetrics_LLM;
 
   /**
-   * judged 0-1: this turn's key entities audibly intact
+   * ttfa_ms, ttfb_ms, naturalness_score, enunciation_score
    *
-   * @generated from field: optional float enunciation_score = 14;
+   * @generated from field: livekit.SimulationRun.JobMetrics.TTS tts = 7;
    */
-  enunciationScore?: number;
+  tts?: SimulationRun_JobMetrics_TTS;
 
   /**
-   * judge failure tags ("verbosity_or_filler", ...)
+   * response_latency_ms, agent_yield_latency_ms, agent_reported_e2e_latency_ms, 0/1 counts
    *
-   * @generated from field: repeated string flags = 15;
+   * @generated from field: livekit.SimulationRun.JobMetrics.Conversation conversation = 8;
    */
-  flags: string[];
+  conversation?: SimulationRun_JobMetrics_Conversation;
 
   constructor(data?: PartialMessage<SimulationRun_JobMetrics_Turn>);
 

package/src/gen/livekit_agent_simulation_pb.js

@@ -181,7 +181,7 @@ export const SimulationRun_JobMetrics = /*@__PURE__*/ proto3.makeMessageType(
 );
 
 /**
- * The agent's perception of the caller.
+ * The agent's perception of the caller. Audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.STT
  */
@@ -217,7 +217,7 @@ export const SimulationRun_JobMetrics_LLM = /*@__PURE__*/ proto3.makeMessageType
 );
 
 /**
- * The agent's voice.
+ * The agent's voice. audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.TTS
  */
@@ -233,6 +233,8 @@ export const SimulationRun_JobMetrics_TTS = /*@__PURE__*/ proto3.makeMessageType
     { no: 10, name: "chars", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
     { no: 11, name: "char_errors", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
     { no: 4, name: "speech_rate_wpm", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
+    { no: 12, name: "heard_words", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
+    { no: 13, name: "speech_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
     { no: 5, name: "naturalness_score", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
     { no: 6, name: "enunciation_score", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
   ],
@@ -240,8 +242,7 @@ export const SimulationRun_JobMetrics_TTS = /*@__PURE__*/ proto3.makeMessageType
 );
 
 /**
- * End-to-end floor dynamics, measured off the audio timeline — the ground
- * truth of what the caller hears, independent of agent self-reports.
+ * Audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.Conversation
  */
@@ -252,6 +253,7 @@ export const SimulationRun_JobMetrics_Conversation = /*@__PURE__*/ proto3.makeMe
     { no: 2, name: "response_latency_p50_ms", kind: "scalar", T: 5 /* ScalarType.INT32 */, opt: true },
     { no: 3, name: "response_latency_p95_ms", kind: "scalar", T: 5 /* ScalarType.INT32 */, opt: true },
     { no: 4, name: "response_latency_p99_ms", kind: "scalar", T: 5 /* ScalarType.INT32 */, opt: true },
+    { no: 16, name: "response_latency_ms", kind: "scalar", T: 5 /* ScalarType.INT32 */, opt: true },
     { no: 5, name: "agent_yield_latency_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
     { no: 6, name: "eot_misprediction_count", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
     { no: 7, name: "overlap_ratio", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
@@ -284,6 +286,8 @@ export const SimulationRun_JobMetrics_Simulator = /*@__PURE__*/ proto3.makeMessa
 );
 
 /**
+ * Rows exist for text and audio;
+ *
  * @generated from message livekit.SimulationRun.JobMetrics.Turn
  */
 export const SimulationRun_JobMetrics_Turn = /*@__PURE__*/ proto3.makeMessageType(
@@ -293,17 +297,10 @@ export const SimulationRun_JobMetrics_Turn = /*@__PURE__*/ proto3.makeMessageTyp
     { no: 2, name: "role", kind: "enum", T: proto3.getEnumType(ChatRole) },
     { no: 3, name: "start_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
     { no: 4, name: "end_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 5, name: "response_latency_ms", kind: "scalar", T: 5 /* ScalarType.INT32 */, opt: true },
-    { no: 6, name: "transcription_delay_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 7, name: "llm_ttft_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 8, name: "llm_ttfs_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 9, name: "tts_ttfa_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 10, name: "tts_ttfb_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 11, name: "agent_reported_e2e_latency_ms", kind: "scalar", T: 13 /* ScalarType.UINT32 */, opt: true },
-    { no: 12, name: "conciseness_score", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
-    { no: 13, name: "naturalness_score", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
-    { no: 14, name: "enunciation_score", kind: "scalar", T: 2 /* ScalarType.FLOAT */, opt: true },
-    { no: 15, name: "flags", kind: "scalar", T: 9 /* ScalarType.STRING */, repeated: true },
+    { no: 5, name: "stt", kind: "message", T: SimulationRun_JobMetrics_STT },
+    { no: 6, name: "llm", kind: "message", T: SimulationRun_JobMetrics_LLM },
+    { no: 7, name: "tts", kind: "message", T: SimulationRun_JobMetrics_TTS },
+    { no: 8, name: "conversation", kind: "message", T: SimulationRun_JobMetrics_Conversation },
   ],
   {localName: "SimulationRun_JobMetrics_Turn"},
 );

package/src/gen/livekit_metrics_pb.d.ts

@@ -415,6 +415,20 @@ export declare class MetricsRecordingHeader extends Message<MetricsRecordingHead
    */
   jobId: string;
 
+  /**
+   * session is a simulation; the collector skips PII redaction for it unless redaction_enabled is set
+   *
+   * @generated from field: bool simulated = 9;
+   */
+  simulated: boolean;
+
+  /**
+   * force PII redaction on for this session (only ever enables, never disables)
+   *
+   * @generated from field: bool redaction_enabled = 10;
+   */
+  redactionEnabled: boolean;
+
   constructor(data?: PartialMessage<MetricsRecordingHeader>);
 
   static readonly runtime: typeof proto3;

package/src/gen/livekit_metrics_pb.js

@@ -108,6 +108,8 @@ export const MetricsRecordingHeader = /*@__PURE__*/ proto3.makeMessageType(
     { no: 6, name: "room_name", kind: "scalar", T: 9 /* ScalarType.STRING */ },
     { no: 7, name: "room_start_time", kind: "message", T: Timestamp },
     { no: 8, name: "job_id", kind: "scalar", T: 9 /* ScalarType.STRING */ },
+    { no: 9, name: "simulated", kind: "scalar", T: 8 /* ScalarType.BOOL */ },
+    { no: 10, name: "redaction_enabled", kind: "scalar", T: 8 /* ScalarType.BOOL */ },
   ],
 );
 

package/src/gen/livekit_models_pb.d.ts

@@ -1078,11 +1078,16 @@ export declare enum ParticipantInfo_KindDetail {
   CONNECTOR_TWILIO = 3,
 
   /**
-   * NEXT_ID: 5
-   *
    * @generated from enum value: BRIDGE_RTSP = 4;
    */
   BRIDGE_RTSP = 4,
+
+  /**
+   * NEXT_ID: 6
+   *
+   * @generated from enum value: SIMULATION = 5;
+   */
+  SIMULATION = 5,
 }
 
 /**

package/src/gen/livekit_models_pb.js

@@ -396,6 +396,7 @@ export const ParticipantInfo_KindDetail = /*@__PURE__*/ proto3.makeEnum(
     {no: 2, name: "CONNECTOR_WHATSAPP"},
     {no: 3, name: "CONNECTOR_TWILIO"},
     {no: 4, name: "BRIDGE_RTSP"},
+    {no: 5, name: "SIMULATION"},
   ],
 );
 

package/src/gen/version.js

@@ -1,2 +1,2 @@
 // Generated by genversion.
-export const version = '1.50.2';
+export const version = '1.50.4';