Expanding path for ffpmeg

Author: MrDesjardins

services/transcription.py

@@ -36,8 +36,11 @@ def compress_audio_for_whisper(audio_path: str) -> str:
     Raises:
         Exception: If compression fails
     """
+    # Expand ~ in path for ffmpeg (ffmpeg doesn't handle ~ paths)
+    expanded_audio_path = str(Path(audio_path).expanduser().resolve())
+
     logger.info(
-        f"Compressing audio file {audio_path} for Whisper (1.5x speed, saves API costs)"
+        f"Compressing audio file {expanded_audio_path} for Whisper (1.5x speed, saves API costs)"
     )
 
     # Create temporary file for compressed audio
@@ -51,7 +54,7 @@ def compress_audio_for_whisper(audio_path: str) -> str:
         compress_cmd = [
             "ffmpeg",
             "-i",
-            audio_path,
+            expanded_audio_path,
             "-filter:a",
             "atempo=1.5",  # Speed up by 1.5x (33% file size reduction)
             "-map",
@@ -80,7 +83,7 @@ def compress_audio_for_whisper(audio_path: str) -> str:
 
         temp_path_path = Path(temp_path)
         compressed_size = temp_path_path.stat().st_size
-        original_size = Path(audio_path).expanduser().resolve().stat().st_size
+        original_size = Path(expanded_audio_path).stat().st_size
 
         logger.info(
             f"Compressed audio from {original_size / 1024 / 1024:.2f}MB "
@@ -218,19 +221,22 @@ def transcribe_audio_gemini(audio_path: str, retries: int = 3) -> str:
     if not config.gemini_api_key:
         raise ValueError("Gemini API key not configured")
 
+    # Expand ~ in path (Python's open() doesn't handle ~ paths)
+    expanded_audio_path = str(Path(audio_path).expanduser().resolve())
+
     last_error = None
 
     for attempt in range(retries):
         try:
             logger.info(
-                f"Transcribing audio file with Gemini: {audio_path} (attempt {attempt + 1}/{retries})"
+                f"Transcribing audio file with Gemini: {expanded_audio_path} (attempt {attempt + 1}/{retries})"
             )
 
             # Create Gemini client
             client = genai.Client(api_key=config.gemini_api_key)
 
             # Upload the audio file
-            with open(audio_path, "rb") as audio_file:
+            with open(expanded_audio_path, "rb") as audio_file:
                 audio_data = audio_file.read()
 
             # Use Gemini to transcribe

tests/services/test_transcription.py

@@ -30,21 +30,27 @@ def test_compress_audio_success(
         compressed_stat.st_size = 3 * 1024 * 1024
         mock_temp_path_instance.stat.return_value = compressed_stat
 
-        # Second call: Path(audio_path).expanduser().resolve()
-        original_stat = Mock()
-        original_stat.st_size = 10 * 1024 * 1024
-
-        # Create the chain: instance.expanduser().resolve().stat() returns original_stat
-        mock_resolved_instance = Mock()
-        mock_resolved_instance.stat.return_value = original_stat
-
+        # Second call: Path(audio_path).expanduser().resolve() - for path expansion
         mock_expanded_instance = Mock()
-        mock_expanded_instance.resolve.return_value = mock_resolved_instance
+        mock_expanded_instance.resolve.return_value = Mock(
+            __str__=lambda x: "/expanded/path/audio.mp3"
+        )
 
         mock_audio_path_instance = Mock()
         mock_audio_path_instance.expanduser.return_value = mock_expanded_instance
 
-        mock_path.side_effect = [mock_temp_path_instance, mock_audio_path_instance]
+        # Third call: Path(expanded_audio_path).stat() - for original file size
+        original_stat = Mock()
+        original_stat.st_size = 10 * 1024 * 1024
+
+        mock_original_path_instance = Mock()
+        mock_original_path_instance.stat.return_value = original_stat
+
+        mock_path.side_effect = [
+            mock_audio_path_instance,
+            mock_temp_path_instance,
+            mock_original_path_instance,
+        ]
 
         # Mock file sizes
         mock_getsize.side_effect = [10 * 1024 * 1024, 3 * 1024 * 1024]  # 10MB -> 3MB
@@ -110,34 +116,37 @@ def test_compress_audio_still_too_large(
         """Test compression when result is still too large."""
         mock_mkstemp.return_value = (123, "/tmp/test.mp3")
 
-        # Mock Path with .expanduser().resolve() chain
-        # First call: Path(temp_path) - no expanduser/resolve
+        # Mock Path calls in order:
+        # 1. Path(audio_path).expanduser().resolve() - for path expansion
+        mock_expanded_instance = Mock()
+        mock_expanded_instance.resolve.return_value = Mock(
+            __str__=lambda x: "/expanded/path/audio.mp3"
+        )
+
+        mock_audio_path_instance = Mock()
+        mock_audio_path_instance.expanduser.return_value = mock_expanded_instance
+
+        # 2. Path(temp_path).stat() - for compressed file size
         mock_temp_path_instance = Mock()
         compressed_stat = Mock()
         compressed_stat.st_size = 30 * 1024 * 1024  # 30MB (over 25MB limit)
         mock_temp_path_instance.stat.return_value = compressed_stat
 
-        # Second call: Path(audio_path).expanduser().resolve()
+        # 3. Path(expanded_audio_path).stat() - for original file size
         original_stat = Mock()
         original_stat.st_size = 50 * 1024 * 1024  # 50MB original
 
-        # Create the chain: instance.expanduser().resolve().stat() returns original_stat
-        mock_resolved_instance = Mock()
-        mock_resolved_instance.stat.return_value = original_stat
-
-        mock_expanded_instance = Mock()
-        mock_expanded_instance.resolve.return_value = mock_resolved_instance
-
-        mock_audio_path_instance = Mock()
-        mock_audio_path_instance.expanduser.return_value = mock_expanded_instance
+        mock_original_path_instance = Mock()
+        mock_original_path_instance.stat.return_value = original_stat
 
-        # Need a third instance for cleanup in except block
+        # 4. Path(temp_path) - for cleanup in except block
         mock_cleanup_path_instance = Mock()
         mock_cleanup_path_instance.exists.return_value = True
 
         mock_path.side_effect = [
-            mock_temp_path_instance,
             mock_audio_path_instance,
+            mock_temp_path_instance,
+            mock_original_path_instance,
             mock_cleanup_path_instance,
         ]
         mock_exists.return_value = True