Network observability lab

I built a passive traffic dashboard and classifier for my own test devices. It uses visible protocol information, packet timing, sizes, and flow duration to explain a classification, with reasons and confidence labels rather than a bare guess.

There is also an optional Ollama summary path in the code. Traffic patterns don’t reveal encrypted content or prove what someone is doing, and no captures or device identities are published here. The current runtime topology still needs a fresh check before I document it as an architecture.