Merge 24797cd278 into c5da69cd13

litellm/litellm_core_utils/audio_utils/utils.py

@@ -311,7 +311,9 @@ def calculate_request_duration(file: FileTypes) -> float | None:
                 content: Final = file[1]
                 if isinstance(content, bytes):
                     file_content = content
-                elif hasattr(content, "read") and not isinstance(content, (str, os.PathLike)):
+                elif isinstance(content, os.PathLike):
+                    return calculate_request_duration(content)
+                elif hasattr(content, "read") and not isinstance(content, str):
                     # File-like object in tuple
                     current_pos: Final = getattr(content, "tell", lambda: None)()
                     # Seek to start to ensure we read the entire content

tests/unit/litellm_core_utils/test_audio_utils.py

@@ -6,6 +6,7 @@ import base64
 import io
 import os
 import tempfile
+from pathlib import Path
 from typing import Final
 from unittest.mock import mock_open, patch
 
@@ -329,16 +330,12 @@ class TestCalculateRequestDuration:
         duration = calculate_request_duration(file_obj)
 
         # Verify it succeeded (returns a duration, not None)
-        assert (
-            duration is not None
-        ), "Duration should be calculated even when BytesIO is at end"
+        assert duration is not None, "Duration should be calculated even when BytesIO is at end"
         assert isinstance(duration, float), "Duration should be a float"
         assert duration > 0, "Duration should be positive"
 
         # Verify the file position was restored
-        assert file_obj.tell() == len(
-            wav_header
-        ), "File position should be restored to original position"
+        assert file_obj.tell() == len(wav_header), "File position should be restored to original position"
 
 
 class TestGetAudioFileContentHash:
@@ -364,9 +361,7 @@ class TestGetAudioFileContentHash:
         hash1 = get_audio_file_content_hash((filename1, content))
         hash2 = get_audio_file_content_hash((filename2, content))
 
-        assert (
-            hash1 == hash2
-        ), "Same content should produce same hash regardless of filename"
+        assert hash1 == hash2, "Same content should produce same hash regardless of filename"
 
     def test_bytes_input(self):
         """Test that raw bytes input works"""
@@ -473,3 +468,17 @@ class TestSpeechMediaTypeFromAudioBytes:
         from litellm.litellm_core_utils.audio_utils.utils import speech_media_type_from_audio_bytes
 
         assert speech_media_type_from_audio_bytes(audio) == expected
+
+
+@pytest.mark.parametrize("include_content_type", [False, True])
+def test_duration_reads_pathlike_content_in_multipart_tuple(tmp_path: Path, include_content_type: bool) -> None:
+    audio_path: Final = tmp_path / "sample.wav"
+    audio_path.write_bytes(base64.b64decode(_encoded_audio(2, 8000, 1, "WAV")))
+    upload: Final = ("sample.wav", audio_path, "audio/wav") if include_content_type else ("sample.wav", audio_path)
+    assert calculate_request_duration(upload) == 2.0
+
+
+def test_duration_does_not_open_bare_string_tuple_content(tmp_path: Path) -> None:
+    audio_path: Final = tmp_path / "sample.wav"
+    audio_path.write_bytes(base64.b64decode(_encoded_audio(2, 8000, 1, "WAV")))
+    assert calculate_request_duration(("sample.wav", str(audio_path))) is None