add Moss result to db

sentence_api/tests/test_processing.py

@@ -2,10 +2,12 @@ import numpy as np
 
 from sentence_api.processing import (
     MOSS_ALGORITHM_VERSION,
+    VideoProcessor,
     document_from_transcript,
     split_sentence_text,
 )
-from sentence_api.models import SentenceBoundary
+from sentence_api.models import SentenceBoundary, SentenceBoundaryDocument
+from sentence_api.repository import VideoRepository
 from sentence_api.transcription import (
     Transcript,
     TranscriptionSegment,
@@ -158,3 +160,166 @@ def test_split_sentence_text_uses_whisper_word_timestamps(tmp_path):
         ("Alpha beta", 1000, 3000),
         ("Gamma delta", 3000, 5000),
     ]
+
+
+def test_processor_persists_original_moss_segments(tmp_path, monkeypatch):
+    from sentence_api.config import Settings
+    from dataclasses import replace
+
+    settings = replace(Settings.from_env(), data_dir=tmp_path / "data")
+    settings.ensure_directories()
+    media_path = settings.videos_dir / f"{VIDEO_HASH}.mp4"
+    media_path.write_bytes(b"video")
+    repository = VideoRepository(settings.database_path)
+    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",
+    )
+    transcript = Transcript(
+        text="Hello world. Next.",
+        segments=[
+            TranscriptionSegment(
+                start_seconds=0.0,
+                end_seconds=1.2,
+                text="Hello world.",
+                speaker="S01",
+                words=[WordTimestamp(0.0, 0.5, "Hello"), WordTimestamp(0.5, 1.2, "world.")],
+            ),
+            TranscriptionSegment(
+                start_seconds=1.2,
+                end_seconds=2.4,
+                text="Next.",
+                speaker="S01",
+            ),
+        ],
+    )
+
+    class AvailableTranscriber:
+        available = True
+
+        def transcribe(self, audio_path, language=None):
+            return transcript
+
+    monkeypatch.setattr("sentence_api.processing._ensure_webm_cues_front", lambda path: None)
+    monkeypatch.setattr(
+        "sentence_api.processing.extract_audio",
+        lambda src, dst: dst.write_bytes(b"RIFF"),
+    )
+    monkeypatch.setattr("sentence_api.processing._media_duration_ms", lambda path: 2400)
+    monkeypatch.setattr(
+        "sentence_api.processing.document_from_transcript",
+        lambda **kwargs: SentenceBoundaryDocument(
+            video_hash=kwargs["video_hash"],
+            duration_ms=2400,
+            algorithm_version=MOSS_ALGORITHM_VERSION,
+            sentences=[
+                SentenceBoundary(
+                    index=0,
+                    start_ms=0,
+                    end_ms=1200,
+                    text="Hello world.",
+                    language="en",
+                    reference_speech_duration_ms=1100,
+                ),
+                SentenceBoundary(
+                    index=1,
+                    start_ms=1200,
+                    end_ms=2400,
+                    text="Next.",
+                    language="en",
+                    reference_speech_duration_ms=1100,
+                ),
+            ],
+        ),
+    )
+
+    VideoProcessor(settings, repository, AvailableTranscriber()).process(VIDEO_HASH)
+
+    segments = repository.get_moss_segments(VIDEO_HASH)
+    assert [(item["start_ms"], item["end_ms"], item["text"], item["speaker"]) for item in segments] == [
+        (0, 1200, "Hello world.", "S01"),
+        (1200, 2400, "Next.", "S01"),
+    ]
+    assert segments[0]["words"] == [
+        {"start_seconds": 0.0, "end_seconds": 0.5, "text": "Hello"},
+        {"start_seconds": 0.5, "end_seconds": 1.2, "text": "world."},
+    ]
+    assert segments[1]["words"] is None
+
+
+def test_reprocess_without_moss_clears_original_segments(tmp_path, monkeypatch):
+    from sentence_api.config import Settings
+    from dataclasses import replace
+
+    settings = replace(Settings.from_env(), data_dir=tmp_path / "data")
+    settings.ensure_directories()
+    media_path = settings.videos_dir / f"{VIDEO_HASH}.mp4"
+    media_path.write_bytes(b"video")
+    repository = VideoRepository(settings.database_path)
+    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=2400,
+                    text="Hello.",
+                    language="en",
+                    reference_speech_duration_ms=2000,
+                )
+            ],
+        ),
+        "Hello.",
+        moss_segments=[
+            {
+                "start_seconds": 0.0,
+                "end_seconds": 2.4,
+                "text": "Hello.",
+                "speaker": "S01",
+            }
+        ],
+    )
+    assert repository.get_moss_segments(VIDEO_HASH)
+
+    class UnavailableTranscriber:
+        available = False
+
+    monkeypatch.setattr("sentence_api.processing._ensure_webm_cues_front", lambda path: None)
+    monkeypatch.setattr(
+        "sentence_api.processing.make_entry",
+        lambda *args, **kwargs: (
+            {
+                "duration_ms": 2400,
+                "sentences": [
+                    SentenceBoundary(
+                        index=0,
+                        start_ms=0,
+                        end_ms=2400,
+                        text=None,
+                        language="en",
+                    )
+                ],
+            },
+            {},
+        ),
+    )
+    VideoProcessor(settings, repository, UnavailableTranscriber()).process(VIDEO_HASH)
+    assert repository.get_moss_segments(VIDEO_HASH) == []
+