fix: UnboundLocalError in transcribe when cloud + include_text=false

[javidjamae]

Summary

process_transcribe_media in services/v1/media/media_transcribe.py raises UnboundLocalError whenever a caller asks for cloud-mode output without text.

The else branch at line 137 assigns text_file = None (wrong name), but line 153 returns text_filename. In the common case of include_text=True the bug is hidden because the if branch assigns text_filename for real. The moment you disable include_text in cloud mode, Python bails out with:

cannot access local variable 'text_filename' where it is not associated with a value

Reproduction

POST /v1/media/transcribe
{
  "media_url": "https://.../speech.mp3",
  "response_type": "cloud",
  "include_text": false,
  "include_srt": true
}

The job is queued, Whisper transcribes, then the upload step explodes with the UnboundLocalError above and the job is marked failed.

Fix

One-line typo fix — bind text_filename = None in the else branch so the return statement sees a defined name.

             else:
-                text_file = None
+                text_filename = None

Test plan

• Reproduced the error against upstream/main (d9bb567) using SRT-only cloud mode.
• After the fix, the same request completes, the SRT is uploaded, and the webhook returns { srt_url, text_url: null, segments_url: null } as expected.
• include_text=True path still behaves identically (the if branch was already correct).

🤖 Generated with Claude Code