change parameter from ns to s

Author: chenghao-mou

livekit-agents/livekit/agents/voice/agent_activity.py

@@ -1215,9 +1215,7 @@ def on_start_of_speech(self, ev: vad.VADEvent | None) -> None:
         speech_start_time = time.time()
         if ev:
             speech_start_time = speech_start_time - ev.speech_duration
-        self._session._update_user_state(
-            "speaking", last_speaking_time=int(speech_start_time * 1_000_000_000)
-        )
+        self._session._update_user_state("speaking", last_speaking_time=speech_start_time)
 
         if self._false_interruption_timer:
             # cancel the timer when user starts speaking but leave the paused state unchanged
@@ -1230,7 +1228,7 @@ def on_end_of_speech(self, ev: vad.VADEvent | None) -> None:
             speech_end_time = speech_end_time - ev.silence_duration
         self._session._update_user_state(
             "listening",
-            last_speaking_time=int(speech_end_time * 1_000_000_000),
+            last_speaking_time=speech_end_time,
         )
 
         if (
@@ -1650,9 +1648,7 @@ def _on_first_frame(fut: asyncio.Future[float] | asyncio.Future[None]) -> None:
             except BaseException:
                 started_speaking_at = time.time()
 
-            self._session._update_agent_state(
-                "speaking", start_time=int(started_speaking_at * 1_000_000_000)
-            )
+            self._session._update_agent_state("speaking", start_time=started_speaking_at)
 
         audio_out: _AudioOutput | None = None
         tts_gen_data: _TTSGenerationData | None = None
@@ -1922,9 +1918,7 @@ def _on_first_frame(fut: asyncio.Future[float] | asyncio.Future[None]) -> None:
                 started_speaking_at = fut.result() or time.time()
             except BaseException:
                 started_speaking_at = time.time()
-            self._session._update_agent_state(
-                "speaking", start_time=int(started_speaking_at * 1_000_000_000)
-            )
+            self._session._update_agent_state("speaking", start_time=started_speaking_at)
 
         audio_out: _AudioOutput | None = None
         if audio_output is not None:
@@ -2273,9 +2267,7 @@ def _on_first_frame(fut: asyncio.Future[float] | asyncio.Future[None]) -> None:
                 started_speaking_at = fut.result() or time.time()
             except BaseException:
                 started_speaking_at = time.time()
-            self._session._update_agent_state(
-                "speaking", start_time=int(started_speaking_at * 1_000_000_000)
-            )
+            self._session._update_agent_state("speaking", start_time=started_speaking_at)
 
         tasks: list[asyncio.Task[Any]] = []
         tees: list[utils.aio.itertools.Tee[Any]] = []
@@ -2677,7 +2669,6 @@ def _on_false_interruption() -> None:
                 self._session._update_agent_state(
                     "speaking",
                     otel_context=self._paused_speech._agent_turn_context,
-                    start_time=time.time_ns(),
                 )
                 audio_output.resume()
                 resumed = True

livekit-agents/livekit/agents/voice/agent_session.py

@@ -1165,18 +1165,20 @@ def _update_agent_state(
         state: AgentState,
         *,
         otel_context: otel_context.Context | None = None,
-        start_time: int | None = None,
+        start_time: float | None = None,
     ) -> None:
         if self._agent_state == state:
             return
 
+        start_time_ns = int(start_time * 1_000_000_000) if start_time else None
+
         if state == "speaking":
             self._llm_error_counts = 0
             self._tts_error_counts = 0
 
             if self._agent_speaking_span is None:
                 self._agent_speaking_span = tracer.start_span(
-                    "agent_speaking", context=otel_context, start_time=start_time
+                    "agent_speaking", context=otel_context, start_time=start_time_ns
                 )
 
                 if self._room_io:
@@ -1202,14 +1204,18 @@ def _update_agent_state(
         )
 
     def _update_user_state(
-        self, state: UserState, *, last_speaking_time: int | None = None
+        self, state: UserState, *, last_speaking_time: float | None = None
     ) -> None:
         if self._user_state == state:
             return
 
+        last_speaking_time_ns = (
+            int(last_speaking_time * 1_000_000_000) if last_speaking_time else None
+        )
+
         if state == "speaking" and self._user_speaking_span is None:
             self._user_speaking_span = tracer.start_span(
-                "user_speaking", start_time=last_speaking_time
+                "user_speaking", start_time=last_speaking_time_ns
             )
 
             if self._room_io and self._room_io.linked_participant:
@@ -1221,7 +1227,7 @@ def _update_user_state(
         elif self._user_speaking_span is not None:
             # end_time = last_speaking_time or time.time()
             # self._user_speaking_span.set_attribute(trace_types.ATTR_END_TIME, end_time)
-            self._user_speaking_span.end(end_time=last_speaking_time)
+            self._user_speaking_span.end(end_time=last_speaking_time_ns)
             self._user_speaking_span = None
 
         if state == "listening" and self._agent_state == "listening":

livekit-agents/livekit/agents/voice/audio_recognition.py

@@ -471,9 +471,7 @@ async def _on_stt_event(self, ev: stt.SpeechEvent) -> None:
     async def _on_vad_event(self, ev: vad.VADEvent) -> None:
         if ev.type == vad.VADEventType.START_OF_SPEECH:
             with trace.use_span(
-                self._ensure_user_turn_span(
-                    start_time=int((time.time() - ev.speech_duration) * 1_000_000_000)
-                )
+                self._ensure_user_turn_span(start_time=time.time() - ev.speech_duration)
             ):
                 self._hooks.on_start_of_speech(ev)
 
@@ -695,11 +693,12 @@ async def _forward() -> None:
             await stream.aclose()
 
     @utils.log_exceptions(logger=logger)
-    def _ensure_user_turn_span(self, start_time: int | None = None) -> trace.Span:
+    def _ensure_user_turn_span(self, start_time: float | None = None) -> trace.Span:
         if self._user_turn_span and self._user_turn_span.is_recording():
             return self._user_turn_span
 
-        self._user_turn_span = tracer.start_span("user_turn", start_time=start_time)
+        start_time_ns = int(start_time * 1_000_000_000) if start_time else None
+        self._user_turn_span = tracer.start_span("user_turn", start_time=start_time_ns)
 
         if (room_io := self._session._room_io) and room_io.linked_participant:
             _set_participant_attributes(self._user_turn_span, room_io.linked_participant)

livekit-agents/livekit/agents/voice/recorder_io/recorder_io.py

@@ -126,7 +126,7 @@ async def _forward_task(self) -> None:
         assert self._in_record is not None
         assert self._out_record is not None
 
-        # Forward the input audio to the encoder every 5s.
+        # Forward the input audio to the encoder every WRITE_INTERVAL seconds.
         while True:
             await asyncio.sleep(WRITE_INTERVAL)
             if self._out_record.has_pending_data: