rultra adds a range finder and a note on repeatable versus correct

What rultra is

A small hardware project kept deliberately honest: every sensor claim has an ADR, the console shows connection loss instead of stale data, and a deploy is not considered done until the service is proven to have restarted.

What changed

The newest commit adds an HC-SR04 ultrasonic range finder and ADR-0006, which separates a repeatable reading from a correct one. A sensor that returns the same wrong number consistently still looks healthy on a dashboard.

Nearby commits made the UI detect connection loss rather than display stale values as live, and made deployment actually restart the service and prove it.

Get started

Hardware plus a checkout. There is no package install for this one.

git clone https://github.com/ruvnet/rultra.git

Expected result: the firmware, console and ADR history. The hardware table in the README records what works and what merely acknowledges — an MCP23008 LCD currently acknowledges on I²C without displaying anything.

No documented MCP server and no npm package. This is a hardware repository.

Use it today

Practical case: a sensing prototype where you need to trust the numbers. Input is a wired sensor; the workflow is read, then verify against a tape measure; the output is a calibration you wrote down.

Acceptance test: measure a known distance ten times. Tight spread proves repeatability; matching the tape measure proves correctness. You need both.

Push it further

Experimental commentary. Combining cheap ultrasonic ranging with WiFi sensing gives two independent views of the same room, which is how you catch a sensor that is confidently wrong.

Falsifiable test: disagree the two sensors deliberately by occluding one. A fused system should flag the disagreement rather than average it away.