Skip to main content
Hiroshi supports automatic inbound voice note transponding using deep acoustic pipelines.

​
🎙️ 1. Message Ingestion Pass

When a user transmits an audio note containing a supported MIME type (audio/ogg, audio/mp3, audio/wav), the ingestion layer:
  1. Captures the binary stream buffer.
  2. Dispatches it to Whisper/Deepgram endpoints.
  3. Automatically replaces the raw audio file attachment with the generated text transcription before forwarding the message block to the prompt assembly engine.