continue fixing

sentence_api/tests/test_assessment_api.py

@@ -1,5 +1,8 @@
 import io
 import json
+from pathlib import Path
+import subprocess
+from unittest import mock
 import wave
 from dataclasses import replace
 
@@ -20,6 +23,9 @@ VIDEO_HASH = "c" * 64
 class FakeTranscriber:
     available = True
 
+    def __init__(self, segments=None):
+        self.segments = segments or []
+
     def transcribe(self, audio_path, language=None):
         return Transcript(text="The meeting starts at nine.", segments=[])
 
@@ -37,7 +43,7 @@ def make_wav() -> bytes:
     return output.getvalue()
 
 
-def make_client(tmp_path, client_api_key=""):
+def make_client(tmp_path, client_api_key="", transcriber=None):
     legacy_path = tmp_path / "legacy.json"
     legacy_path.write_text(json.dumps({"videos": {}}), encoding="utf-8")
     settings = replace(
@@ -82,7 +88,7 @@ def make_client(tmp_path, client_api_key=""):
         BoundaryStore(legacy_path),
         settings=settings,
         repository=repository,
-        transcriber=FakeTranscriber(),
+        transcriber=transcriber or FakeTranscriber(),
     )
     return TestClient(app)
 
@@ -148,6 +154,54 @@ def test_get_dub_share_includes_video_hash(tmp_path):
     assert response.json()["video_hash"] == VIDEO_HASH
 
 
+def test_create_dub_share_aligns_audio_with_whisper_boundaries(tmp_path):
+    from sentence_api.transcription import Transcript, TranscriptionSegment
+    from types import SimpleNamespace
+
+    aligned = tmp_path / "aligned.wav"
+    aligned.write_bytes(b"aligned")
+    calls = []
+
+    class AlignedFakeTranscriber(FakeTranscriber):
+        available = True
+
+        def transcribe(self, audio_path, language=None):
+            calls.append(audio_path)
+            return Transcript(
+                text="The meeting starts at nine.",
+                segments=[
+                    TranscriptionSegment(
+                        start_seconds=0.4,
+                        end_seconds=1.6,
+                        text="The meeting starts at nine.",
+                    )
+                ],
+            )
+
+    client = make_client(tmp_path, transcriber=AlignedFakeTranscriber())
+    original_run = subprocess.run
+
+    def fake_run(command, **kwargs):
+        Path(command[-1]).write_bytes(b"aligned-audio")
+        return SimpleNamespace(returncode=0)
+
+    with mock.patch("sentence_api.main.subprocess.run", side_effect=fake_run), mock.patch(
+        "sentence_api.main._media_duration_seconds",
+        return_value=2.0,
+    ):
+        response = client.post(
+            "/api/v1/dub-shares",
+            data={
+                "video_hash": VIDEO_HASH,
+                "segments": json.dumps([{"sentence_index": 0}]),
+            },
+            files={"files": ("dub.wav", make_wav(), "audio/wav")},
+        )
+
+    assert response.status_code == 201, response.text
+    assert calls[0].read_bytes() == b"aligned-audio"
+
+
 def test_assessment_client_key_is_enforced_when_configured(tmp_path):
     client = make_client(tmp_path, client_api_key="tablet-key")
     endpoint = f"/api/v1/videos/{VIDEO_HASH}/sentences/0/assessments"