Running on AMD

Speech-to-text runs on the edge gateway (Raspberry Pi); classification and alert explanations run on AMD Token Factory (Qwen3.8-27B).

Token Factory serves language models but not audio, so Whisper transcribes voice notes on the Pi next to the line. Every call the backend makes is logged below.