@livekit/protocol 1.50.2 → 1.50.3

package/dist/index.d.cts

@@ -10364,11 +10364,15 @@ declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobMetrics>
    *   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;
@@ -10376,6 +10380,8 @@ declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobMetrics>
   /**
    * 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;
@@ -10434,7 +10440,7 @@ declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobMetrics>
   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;
    */
@@ -10456,66 +10462,76 @@ declare class SimulationRun_JobMetrics extends Message<SimulationRun_JobMetrics>
 }
 
 /**
- * The agent's perception of the caller.
+ * The agent's perception of the caller. Audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.STT
  */
 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;
    */
@@ -10541,28 +10557,28 @@ declare class SimulationRun_JobMetrics_STT extends Message<SimulationRun_JobMetr
  */
 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;
    */
@@ -10584,57 +10600,63 @@ declare class SimulationRun_JobMetrics_LLM extends Message<SimulationRun_JobMetr
 }
 
 /**
- * The agent's voice.
+ * The agent's voice. audio only.
  *
  * @generated from message livekit.SimulationRun.JobMetrics.TTS
  */
 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;
@@ -10647,14 +10669,28 @@ declare class SimulationRun_JobMetrics_TTS extends Message<SimulationRun_JobMetr
   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;
    */
@@ -10676,8 +10712,7 @@ declare class SimulationRun_JobMetrics_TTS extends Message<SimulationRun_JobMetr
 }
 
 /**
- * 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
  */
@@ -10707,14 +10742,21 @@ declare class SimulationRun_JobMetrics_Conversation extends Message<SimulationRu
   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;
    */
@@ -10747,35 +10789,35 @@ declare class SimulationRun_JobMetrics_Conversation extends Message<SimulationRu
   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;
    */
@@ -10834,6 +10876,8 @@ declare class SimulationRun_JobMetrics_Simulator extends Message<SimulationRun_J
 }
 
 /**
+ * Rows exist for text and audio;
+ *
  * @generated from message livekit.SimulationRun.JobMetrics.Turn
  */
 declare class SimulationRun_JobMetrics_Turn extends Message<SimulationRun_JobMetrics_Turn> {
@@ -10852,78 +10896,48 @@ declare class SimulationRun_JobMetrics_Turn extends Message<SimulationRun_JobMet
   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>);