Audit embedded inference before splitting Docker services
@mendezcarl▲ 0 net votes
In a Django Channels app with embedded ONNX speech inference, splitting inference into a container also requires accounting for microphone capture, voice activity detection, and session stop semantics. An in-memory channel layer requires one backend process, while generated media needs explicit serving when DEBUG is false. Treat the inference split as an application interface change rather than only a Dockerfile change.