fixed a bug

sentence_api/tests/test_processing.py

@@ -323,3 +323,61 @@ def test_reprocess_without_moss_clears_original_segments(tmp_path, monkeypatch):
     VideoProcessor(settings, repository, UnavailableTranscriber()).process(VIDEO_HASH)
     assert repository.get_moss_segments(VIDEO_HASH) == []
 
+
+def test_mark_processing_clears_previous_sentence_data(tmp_path):
+    repository = VideoRepository(tmp_path / "oral.sqlite3")
+    repository.upsert_upload(
+        video_hash=VIDEO_HASH,
+        title="MOSS lesson",
+        filename="lesson.mp4",
+        stored_filename=f"{VIDEO_HASH}.mp4",
+        content_type="video/mp4",
+        size_bytes=5,
+        language="en",
+    )
+    repository.save_processing_result(
+        SentenceBoundaryDocument(
+            video_hash=VIDEO_HASH,
+            duration_ms=2400,
+            algorithm_version=MOSS_ALGORITHM_VERSION,
+            sentences=[
+                SentenceBoundary(
+                    index=0,
+                    start_ms=0,
+                    end_ms=1200,
+                    text="Old first.",
+                    language="en",
+                    reference_speech_duration_ms=1000,
+                ),
+                SentenceBoundary(
+                    index=1,
+                    start_ms=1200,
+                    end_ms=2400,
+                    text="Old second.",
+                    language="en",
+                    reference_speech_duration_ms=1000,
+                ),
+            ],
+        ),
+        "Old first. Old second.",
+        moss_segments=[
+            {
+                "start_seconds": 0.0,
+                "end_seconds": 1.2,
+                "text": "Old first.",
+                "speaker": "S01",
+            }
+        ],
+    )
+
+    repository.mark_processing(VIDEO_HASH)
+
+    video = repository.get_video(VIDEO_HASH)
+    document = repository.get_document(VIDEO_HASH)
+    assert video["status"] == "processing"
+    assert video["transcription"] is None
+    assert video["sentence_count"] == 0
+    assert document is not None
+    assert document.sentences == []
+    assert repository.get_moss_segments(VIDEO_HASH) == []
+