What's Changed
- Fix AssemblyAI transcription (#277, #280): AssemblyAI's API now requires the
speech_modelsparameter in transcript creation requests. Without it, all transcription requests were rejected with HTTP 400. This release addsuniversal-3-pro(withuniversal-2fallback) as the default model chain, restoring AssemblyAI transcription for all users. - Better error diagnostics: Added logging for 4xx errors from AssemblyAI's transcript creation endpoint to aid future debugging.
Thanks to @gthay for reporting the issue and providing detailed logs and reproduction steps.