feat: Celery setup and transcription task — Whisper + FFmpeg pipeline

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/tests/test_transcription_service.py

@@ -0,0 +1,41 @@
+import os
+import pytest
+from unittest.mock import patch, MagicMock
+
+from app.services.transcription import extract_audio, transcribe_audio
+
+
+@pytest.mark.asyncio
+async def test_extract_audio_calls_ffmpeg():
+    with patch("app.services.transcription.subprocess.run") as mock_run:
+        mock_run.return_value = MagicMock(returncode=0)
+        result = extract_audio("/tmp/test.mp4")
+        assert result.endswith(".wav")
+        mock_run.assert_called_once()
+        args = mock_run.call_args
+        assert "ffmpeg" in args[0][0]
+
+
+@pytest.mark.asyncio
+async def test_transcribe_audio_returns_expected_structure():
+    mock_segment = MagicMock()
+    mock_segment.start = 0.0
+    mock_segment.end = 5.0
+    mock_segment.text = " Hello world "
+
+    mock_info = MagicMock()
+    mock_info.language = "en"
+
+    with patch("app.services.transcription._get_model") as mock_model:
+        mock_model.return_value.transcribe.return_value = (
+            iter([mock_segment]),
+            mock_info,
+        )
+        result = transcribe_audio("/tmp/test.wav")
+
+    assert "text" in result
+    assert result["text"] == "Hello world"
+    assert len(result["segments"]) == 1
+    assert result["segments"][0]["start"] == 0.0
+    assert result["language"] == "en"
+    assert "processing_time" in result