Refactor timed objects and data structures

whisperlivekit/metrics.py

@@ -6,7 +6,7 @@ text normalization, and word-level timestamp accuracy metrics with greedy alignm
 
 import re
 import unicodedata
-from typing import Dict, List, Optional
+from typing import Dict, List
 
 
 def normalize_text(text: str) -> str:

whisperlivekit/metrics_collector.py

@@ -78,7 +78,6 @@ class SessionMetrics:
 
     def log_summary(self) -> None:
         """Emit a structured log line summarising the session."""
-        self.total_processing_time_s = sum(self.transcription_durations)
         d = self.to_dict()
         d["session_elapsed_s"] = round(time.time() - self.session_start, 3) if self.session_start else 0
         logger.info(f"SESSION_METRICS {d}")

whisperlivekit/thread_safety.py

@@ -20,8 +20,8 @@ Usage:
     export WHISPERLIVEKIT_LOCK_TIMEOUT=60
 """
 
-import os
 import logging
+import os
 import threading
 
 logger = logging.getLogger(__name__)

whisperlivekit/timed_objects.py

@@ -1,12 +1,18 @@
 from dataclasses import dataclass, field
-from datetime import timedelta
 from typing import Any, Dict, List, Optional, Union
 
 PUNCTUATION_MARKS = {'.', '!', '?', '。', '！', '？'}
 
 def format_time(seconds: float) -> str:
-    """Format seconds as HH:MM:SS."""
+    """Format seconds as H:MM:SS.cc (centisecond precision)."""
-    return str(timedelta(seconds=int(seconds)))
+    total_cs = int(round(seconds * 100))
+    cs = total_cs % 100
+    total_s = total_cs // 100
+    s = total_s % 60
+    total_m = total_s // 60
+    m = total_m % 60
+    h = total_m // 60
+    return f"{h}:{m:02d}:{s:02d}.{cs:02d}"
 
 @dataclass
 class Timed:
@@ -227,4 +233,4 @@ class State():
     new_translation: List[Any] = field(default_factory=list)
     new_diarization: List[Any] = field(default_factory=list)
     new_tokens_buffer: List[Any] = field(default_factory=list)  # only when local agreement
-    new_translation_buffer= TimedText()
+    new_translation_buffer: TimedText = field(default_factory=TimedText)