Skip to content
Field readings (edge)

Field readings (edge)

What it is: the field-reading narrative, which writes up a walk with a handheld reader such as the Flipper edge reader.

What it does

It turns what a handheld reader picked up on a site walk into one plain paragraph that says what was found and, just as clearly, what the walk does not establish. Its jobs:

  • Phrase the walk. Writes one paragraph over readings that have already been decided, without changing any of them.
  • Keep the distinctions. Keeps apart:
    • a reading with no value and a reading of zero;
    • “absent” (the sensor was driven and nothing answered) and “unstable” (something is there, but it isn’t a measurement);
    • readings that support the walk’s claim and ambient readings taken at the same time;
    • a gap (nobody was standing there) and an outage, which a gap never is.
  • Say what it doesn’t prove. Includes the walk’s “what this does not establish” sentence every time.

What it reads

The frozen readings of one walk as the reader’s connector reported them, such as NFC, 125 kHz RFID, infrared, sub-GHz, 1-Wire and Wi-Fi sightings, with each reading’s role and quality, and the walk’s own statement of what it doesn’t establish.

What it produces

One short paragraph, with no headings, bullets, recommendations or next steps. When the paragraph can’t be written from the input alone, it says exactly what is missing instead.

Human oversight

  • The paragraph is a report. It proposes no action. Any action on a site walk is a person’s decision.
  • It changes no figure. Every number in it is given to it and frozen.

Model

The deployment’s general reasoning tier.

The narrative records which model wrote it. The readings and their classification are set before the model is called.

Where it runs and data handling

On the deployment’s own model plane: Runink’s shared machines on a Trial, your own cloud on Cloud and your own servers on On-premises. No outside AI service is called on any plan. Labels from a field reader are treated as untrusted, because anyone nearby can influence what a radio reader hears. See Facts shared by every card.

Guardrails

  • Never computes, rounds, corrects or reconciles a figure. If two figures look inconsistent, it names both and leaves them unresolved.
  • Never writes a reading with no value as a measurement.
  • Never presents an ambient reading as support for the walk’s claim.
  • An enclosure or device temperature measures the equipment, never the cargo, and the paragraph keeps that qualifier.

How to read its results

  • A sighting says a thing was heard. It isn’t a location, a count or an inventory.
  • Radio occupancy says a band was busy, not what was transmitting.
  • It doesn’t identify people or say whether anyone was authorised to be there.