Test speaker alignment and schema v1.1

tests/test_whisper_service.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 import json
 from pathlib import Path
 
+from app.models.transcript import DiarizationSpan, Transcript, TranscriptSegment, TranscriptWord
 from app.services.whisper_service import WhisperService
 
 
@@ -60,12 +61,6 @@ def test_transcribe_sync_includes_word_timestamps() -> None:
 
 
 def test_word_cues_are_clamped_to_segment() -> None:
-    from app.models.transcript import (
-        Transcript,
-        TranscriptSegment,
-        TranscriptWord,
-    )
-
     transcript = Transcript(
         language="en",
         language_probability=1.0,
@@ -95,6 +90,79 @@ def test_word_caption_formats() -> None:
     assert vtt.startswith("WEBVTT\n")
 
 
+def test_diarization_schema_and_alignment() -> None:
+    transcript = Transcript(
+        language="en",
+        language_probability=1.0,
+        duration=4.0,
+        task="transcribe",
+        model="tiny",
+        text="hello world",
+        segments=[
+            TranscriptSegment(
+                id=0,
+                start=0.0,
+                end=2.0,
+                text=" hello world ",
+                words=[
+                    TranscriptWord(word=" hello", start=0.2, end=0.8),
+                    TranscriptWord(word=" world", start=1.2, end=1.8),
+                ],
+            )
+        ],
+    )
+    aligned = WhisperService.apply_diarization(
+        transcript,
+        [
+            {"speaker": "SPEAKER_00", "start": 0.0, "end": 1.0},
+            {"speaker": "SPEAKER_01", "start": 1.0, "end": 2.0},
+        ],
+    )
+    assert aligned.schema_version == "1.1"
+    assert aligned.speakers == ["SPEAKER_00", "SPEAKER_01"]
+    assert aligned.segments[0].speaker == "SPEAKER_00"
+    assert aligned.segments[0].words[0].speaker == "SPEAKER_00"
+    assert aligned.segments[0].words[1].speaker == "SPEAKER_01"
+
+
+def test_diarization_overlap_prefers_longest_overlap() -> None:
+    transcript = Transcript(
+        language="en",
+        language_probability=1.0,
+        duration=3.0,
+        task="transcribe",
+        model="tiny",
+        text="test",
+        segments=[
+            TranscriptSegment(
+                id=0,
+                start=0.0,
+                end=2.0,
+                text=" test ",
+                words=[TranscriptWord(word=" test", start=0.5, end=1.5)],
+            )
+        ],
+    )
+    aligned = WhisperService.apply_diarization(
+        transcript,
+        [
+            DiarizationSpan(speaker="A", start=0.0, end=0.9),
+            DiarizationSpan(speaker="B", start=0.9, end=1.5),
+        ],
+    )
+    assert aligned.segments[0].speaker == "B"
+    assert aligned.segments[0].words[0].speaker == "B"
+
+
+def test_diarization_rejects_invalid_span() -> None:
+    try:
+        DiarizationSpan(speaker="A", start=2.0, end=2.0)
+    except ValueError as exc:
+        assert "end must be greater" in str(exc)
+    else:
+        raise AssertionError("invalid diarization span was accepted")
+
+
 async def test_word_caption_output(tmp_path, settings) -> None:
     service = WhisperService(settings)
     source = tmp_path / "sample.wav"
@@ -126,7 +194,7 @@ async def test_json_output_preserves_words(tmp_path, settings) -> None:
     payload = json.loads(result.path.read_text())
     assert payload["word_timestamps"] is True
     assert payload["segments"][0]["words"][0]["word"] == " hello"
-    assert payload["schema_version"] == "1.0"
+    assert payload["schema_version"] == "1.1"
 
 
 async def test_all_text_formats_are_derived_from_canonical_transcript(tmp_path, settings) -> None:
@@ -152,7 +220,7 @@ async def test_transcript_metadata_uses_canonical_schema(tmp_path, settings) ->
         source, tmp_path / "out", model_name="tiny", task="translate"
     )
     transcript = result.metadata["transcript"]
-    assert transcript["schema_version"] == "1.0"
+    assert transcript["schema_version"] == "1.1"
     assert transcript["task"] == "translate"
     assert transcript["model"] == "tiny"
     assert transcript["segments"][0]["words"][1]["end"] == 1.988