Skip to main content

Audio ingestion

What this does​

If you have a radio broadcast / interview / podcast clip that matters for today’s assessment, this guide turns it into text evidence (a markdown transcript) and then runs the same analysis pipeline used for news and WhatsApp.

What you will see in the app​

  • Some evidence items will cite audio/radio coverage as a source (if enabled by your operator).
  • Audio-derived evidence often shows up as direct quotes or summarized transcript fragments, depending on privacy settings.

Troubleshooting​

  • Transcription fails with “file too large”
    • Cause: long recordings exceed the upload cap.
    • Fix: install ffmpeg so the script can split audio automatically.
  • Transcription fails (401)
    • Cause: missing/invalid OPENAI_API_KEY.
    • Fix: set the key and retry.
  • Analysis fails
    • Cause: missing ANTHROPIC_API_KEY (analysis) or upstream outage.
    • Fix: set the key and retry later if the provider is down.