Voice Ingest

Take an audio file → transcript → classification → proposed route. Output lands in inbox/ until the user confirms a route via route-capture.

Inputs

$ARGUMENTS:

• Required: <audio-path> — path to an audio file (opus, mp3, m4a, wav, ogg).
• Optional: --preset=<preset> — override transcription preset (default: read from companion).
• Optional: --engine=<assemblyai|gemini> — override engine (default: companion config; gemini-cleaned recommended for noisy memos).

Examples

/career:capture ~/voice-memos/2026-04-28-1430.opus
/career:capture ~/Downloads/note.m4a --engine=gemini

Procedure