Cowl: Radio-Sight and Vision HUD

What it is

A personal sensory-augmentation system. An RTL-SDR dongle decodes aircraft and 315/433 MHz devices, the Mac's Bluetooth finds named tags, and the phone camera feeds Claude for a plain-English read. An installable phone web app shows it all, answers spoken questions about the live sensor picture and pushes alerts.

The problem

Radio signals and nearby objects are invisible and unreadable without specialist tools; this turns them into a glanceable, conversational feed.

How it works

1. A radio supervisor time-slices the dongle between dump1090-fa (aircraft) and rtl_433 (ISM devices) and ingests into SQLite in WAL mode
2. Classification, alert rules, a briefing timer and an away mode that raises vigilance run alongside ingestion
3. A FastAPI service serves the HUD and streaming vision and ask endpoints backed by the Anthropic API
4. Two launchd-managed processes on an always-on Mac mini, exposed to the phone over Tailscale serve with real TLS; pure logic is covered by a pytest suite

Stack

Disciplines

For your business

Live sensor and camera pipelines with an LLM layer that explains what is happening and alerts on anomalies.