fix: make speaker and confidence optional in TranscriptUtterance

- server.py: use result.get('confidence') instead of result['confidence']
  to prevent KeyError when word-level data is unavailable
- models.py: make TranscriptUtterance.speaker and .confidence optional
  since format_result() only includes them conditionally (speaker requires
  diarization, confidence requires word-level data)
- test_models.py: add tests for utterances without speaker/confidence

Author: Odrec

src/murmurai_server/models.py

@@ -105,11 +105,11 @@ class TranscriptWord(BaseModel):
 class TranscriptUtterance(BaseModel):
     """Speaker utterance (segment) data."""
 
-    speaker: str
+    speaker: str | None = None
     text: str
     start: int  # milliseconds
     end: int  # milliseconds
-    confidence: float
+    confidence: float | None = None
     words: list[TranscriptWord] | None = None
 
 

src/murmurai_server/server.py

@@ -187,7 +187,7 @@ def sync_progress_callback(progress: float) -> None:
                 text=result["text"],
                 words=result["words"],
                 utterances=result["utterances"],
-                confidence=result["confidence"],
+                confidence=result.get("confidence"),
                 audio_duration=result["audio_duration"],
                 language_code=result["language_code"],
                 progress=1.0,

tests/test_models.py

@@ -165,6 +165,39 @@ def test_utterance_without_words(self):
         )
         assert utterance.words is None
 
+    def test_utterance_without_speaker(self):
+        """Test utterance without speaker (no diarization)."""
+        utterance = TranscriptUtterance(
+            text="Hello World",
+            start=0,
+            end=1000,
+            confidence=0.96,
+        )
+        assert utterance.speaker is None
+        assert utterance.text == "Hello World"
+
+    def test_utterance_without_confidence(self):
+        """Test utterance without confidence (no word-level data)."""
+        utterance = TranscriptUtterance(
+            text="Hello World",
+            start=0,
+            end=1000,
+        )
+        assert utterance.confidence is None
+        assert utterance.speaker is None
+
+    def test_utterance_without_speaker_and_confidence(self):
+        """Test utterance with only required fields (no diarization, no word-level data)."""
+        utterance = TranscriptUtterance(
+            text="Test segment",
+            start=0,
+            end=500,
+        )
+        assert utterance.speaker is None
+        assert utterance.confidence is None
+        assert utterance.words is None
+        assert utterance.text == "Test segment"
+
 
 class TestTranscript:
     """Tests for Transcript model."""