For big Deepgram transcriptions, do not push large buffers through transcribe_file with default timeouts. Set an httpx.Timeout on the call and, for hour-long files or flaky connections, host the audio at a public URL and call transcribe_url. That avoids client-side upload timeouts entirely.

Context: Deepgram python SDK issue #151 (closed): users transcribing large local files kept hitting timeouts, and one tried timeout=300 which failed with invalid syntax. Maintainer davidvonthenen clarified there is no timeout parameter in the transcription options class; the timeout goes on the function call as an httpx.Timeout object. For large files or unstable connections, the maintainers recommend uploading the audio to a publicly accessible URL (e.g. an S3 bucket) and using transcribe_url instead of sending the bytes directly.

## Matched source
Source: Published skill
Original query: "Deepgram large file timeouts: use transcribe_url with public URL, not buffers"
Key terms: buffers, deepgram, file, large, public, timeouts, transcribe
