feat: add transcription participants stats at the end of the meeting (#249)

Author: rpurdel

skynet/modules/stt/streaming_whisper/connection_manager.py

@@ -64,10 +64,19 @@ async def send(self, connection: MeetingConnection, results: list[utils.Transcri
                     log.error(f'Meeting {connection.meeting_id}: exception while sending transcription results {ex}')
 
     async def disconnect(self, connection: MeetingConnection, already_closed=False):
+        # Build participant audio distribution string
+        participant_stats = []
+        for participant_id, state in connection.participants.items():
+            participant_stats.append(f"{participant_id}: {state.total_audio_received_s:.1f}s")
+
+        participants_str = " | ".join(participant_stats) if participant_stats else "none"
+
         log.info(
             f"Closed {connection.meeting_id} | Audio: {connection.total_audio_received_s}s | "
             + f"Interims: {connection.total_interims} | Finals: {connection.total_finals}"
         )
+        log.info(f"Participant audio distribution: {participants_str}")
+
         try:
             self.connections.remove(connection)
         except ValueError:

skynet/modules/stt/streaming_whisper/meeting_connection.py

@@ -84,6 +84,7 @@ async def force_transcription(self, participant_id: str):
         if participant_id in self.participants:
             payloads = await self.participants[participant_id].force_transcription(self.previous_transcription_tokens)
             if payloads:
+                await self.update_connection_summary_stats(payloads)
                 await self.update_initial_prompt(payloads)
             return payloads
         return None

skynet/modules/stt/streaming_whisper/state.py

@@ -22,6 +22,7 @@ class State:
     working_audio_starts_at: int
     last_received_chunk: int
     last_speech_timestamp: float
+    total_audio_received_s: float
 
     def __init__(
         self,
@@ -40,6 +41,7 @@ def __init__(
         self.last_received_chunk = utils.now()
         self.is_transcribing = False
         self.last_speech_timestamp = 0.0
+        self.total_audio_received_s = 0.0
 
     def _extract_transcriptions(
         self, last_pause: utils.CutMark, ts_result: utils.WhisperResult
@@ -140,6 +142,7 @@ async def process(self, chunk: Chunk, previous_tokens: list[int]) -> List[utils.
     async def add_to_store(self, chunk: Chunk, tmp_working_audio: bytes = b''):
         now_millis = utils.now()
         self.chunk_count += 1
+        self.total_audio_received_s += chunk.duration
         # if the working audio is empty, set the start timestamp
         if not self.working_audio:
             self.working_audio_starts_at = chunk.timestamp - int(chunk.duration * 1000)