repaire sentences

sentence_api/tests/test_api.py

@@ -1029,3 +1029,139 @@ def test_admin_can_update_course_title(tmp_path):
         json={"title": "Missing course"},
     )
     assert missing_response.status_code == 404
+
+
+
+def test_admin_can_repair_sentence_gaps(tmp_path, monkeypatch):
+    settings = replace(Settings.from_env(), data_dir=tmp_path / "data")
+    settings.ensure_directories()
+    repository = VideoRepository(settings.database_path)
+    repository.upsert_upload(
+        video_hash=VIDEO_HASH,
+        title="Repair gaps lesson",
+        filename="lesson.mp4",
+        stored_filename=f"{VIDEO_HASH}.mp4",
+        content_type="video/mp4",
+        size_bytes=5,
+        language="en",
+    )
+    (settings.videos_dir / f"{VIDEO_HASH}.mp4").write_bytes(b"fake-video")
+    repository.save_processing_result(
+        SentenceBoundaryDocument(
+            video_hash=VIDEO_HASH,
+            duration_ms=18000,
+            algorithm_version="test",
+            sentences=[
+                SentenceBoundary(
+                    index=0,
+                    start_ms=0,
+                    end_ms=2000,
+                    text="Hello.",
+                    language="en",
+                    reference_speech_duration_ms=1800,
+                ),
+                SentenceBoundary(
+                    index=1,
+                    start_ms=15000,
+                    end_ms=18000,
+                    text="Goodbye.",
+                    language="en",
+                    reference_speech_duration_ms=2500,
+                ),
+            ],
+        ),
+        "Hello. Goodbye.",
+    )
+    repository.save_sentence_translations(VIDEO_HASH, {0: "你好。", 1: "再见。"})
+
+    from sentence_api.transcription import Transcript, TranscriptionSegment, _wav_bytes
+    import numpy as np
+
+    class GapTranscriber:
+        available = True
+
+        def transcribe(self, audio_path, language=None):
+            return Transcript(
+                text="The missing sentence.",
+                segments=[
+                    TranscriptionSegment(
+                        start_seconds=1.0,
+                        end_seconds=5.0,
+                        text="The missing sentence.",
+                    )
+                ],
+            )
+
+    def fake_extract(media_path, output_path, *, start_ms, duration_ms):
+        sample_rate = 16000
+        t = np.arange(int(max(1.0, duration_ms / 1000) * sample_rate)) / sample_rate
+        samples = (0.25 * np.sin(2 * np.pi * 220 * t)).astype(np.float32)
+        output_path.write_bytes(_wav_bytes(samples, sample_rate))
+
+    monkeypatch.setattr("sentence_api.processing.extract_audio_segment", fake_extract)
+    client = TestClient(
+        create_app(
+            BoundaryStore(tmp_path / "boundaries.json"),
+            settings=settings,
+            repository=repository,
+            transcriber=GapTranscriber(),
+        )
+    )
+
+    response = client.post(f"/api/v1/admin/videos/{VIDEO_HASH}/repair-gaps")
+
+    assert response.status_code == 200, response.text
+    payload = response.json()
+    assert payload["gap_count"] == 1
+    assert payload["inserted_count"] == 1
+    assert payload["sentence_count"] == 3
+    document = repository.get_document(VIDEO_HASH)
+    assert [item.text for item in document.sentences] == [
+        "Hello.",
+        "The missing sentence.",
+        "Goodbye.",
+    ]
+    assert [item.zh_text for item in document.sentences] == ["你好。", None, "再见。"]
+
+
+def test_admin_gap_repair_requires_transcriber(tmp_path):
+    settings = replace(Settings.from_env(), data_dir=tmp_path / "data")
+    settings.ensure_directories()
+    repository = VideoRepository(settings.database_path)
+    repository.upsert_upload(
+        video_hash=VIDEO_HASH,
+        title="Repair gaps lesson",
+        filename="lesson.mp4",
+        stored_filename=f"{VIDEO_HASH}.mp4",
+        content_type="video/mp4",
+        size_bytes=5,
+        language="en",
+    )
+    (settings.videos_dir / f"{VIDEO_HASH}.mp4").write_bytes(b"fake-video")
+    repository.save_processing_result(
+        SentenceBoundaryDocument(
+            video_hash=VIDEO_HASH,
+            duration_ms=18000,
+            algorithm_version="test",
+            sentences=[
+                SentenceBoundary(index=0, start_ms=0, end_ms=2000, text="Hello."),
+                SentenceBoundary(index=1, start_ms=15000, end_ms=18000, text="Goodbye."),
+            ],
+        ),
+        None,
+    )
+
+    class UnavailableTranscriber:
+        available = False
+
+    client = TestClient(
+        create_app(
+            BoundaryStore(tmp_path / "boundaries.json"),
+            settings=settings,
+            repository=repository,
+            transcriber=UnavailableTranscriber(),
+        )
+    )
+
+    response = client.post(f"/api/v1/admin/videos/{VIDEO_HASH}/repair-gaps")
+    assert response.status_code == 503