fix(cli/audio): file format detection failing for whisper

[thiswillbeyourgithub]

Fix for #333

My initial problem

My case is using openai api audio.transcriptions.create -f audio.mp3. Without the fix the returned value is Error: Error code: 400 - {'error': {'message': "Unrecognized file format. Supported formats: ['flac', 'm4a', 'mp3', 'mp4', 'mpeg', 'mpga', 'oga', 'ogg', 'wav', 'webm']", 'type': 'invalid_request_error', 'param': None, 'code': None}}.

What the cause was

The issue was that using BufferReader to read from the audio file was not keeping the name attribute so the file detection code on the server side was failing.

Quality

My fix is a viable workaround but a more invasive fix would be better. I found numerous other posts with this kind of issue for a while :

• https://community.openai.com/t/has-the-whisper-error-been-solved/433315
• https://community.openai.com/t/whisper-api-cannot-read-files-correctly/93420/54
• https://community.openai.com/t/whisper-api-completely-wrong-for-mp4/289256/6
• https://community.openai.com/t/openai-whisper-send-bytes-python-instead-of-filename/84786/3
• https://community.openai.com/t/whisperai-api-not-recognizing-valid-file-formats/145208
• https://community.openai.com/t/whisper-api-cannot-read-files-correctly/93420/51

so I think this quick non invasive fix would be a good idea.

Edit: confirmed this fix is still relevant for 1.2.0. I'm on linux using python 3.9.0 for this test.

[RobertCraigie]

Thanks for the PR!