Events & confidence · the output contract

Confidence that controls action.

Tollscopic DOT# separates what the system believes from what the deployment is allowed to do with that belief. The output is a structured event with evidence and a clear usage contract — not a number to interpret.

01 · Event family

Two events. One contract.

A passage event records that an eligible commercial vehicle passed. An identity event records the carrier result or the refusal state. Both point back to the same evidence package and the same versions.

01 · passage_event
A commercial vehicle was here.

Site, lane, timestamp, vehicle-class hint, track reference. Emitted whenever an eligible commercial vehicle passes through the capture zone — even if the identity event below ends in refusal.

02 · identity_event
What the system can say about who.

Carrier result (or refusal class), confidence, evidence references, catalog snapshot, model and scorer versions, and the identity_usage_policy that governs what downstream systems are allowed to do.

No silent drops. Every eligible track produces a passage event. Every passage event produces an identity event — including refusal classes. The downstream system always knows the track existed.
02 · Confidence classes

Six outcomes. Each one is a designed product output.

Refusal classes are not embarrassing exceptions. They are part of the contract — a serious buyer wants a system that can say 'I do not know' without hallucinating or silently dropping the pass.

Class
Plain-English meaning
Typical public action
verified
One carrier is strongly supported.
Correlation and review · stronger actions only if deployment policy permits.
probable
One carrier is likely but not final.
Correlation and review.
ambiguous
Multiple carriers remain plausible.
Review.
unreadable
Evidence was insufficient.
Measure · audit · improve.
no_panel
No readable panel was found in the selected frames.
Measure · review patterns.
catalog_miss
Evidence was readable but unsupported by the catalog.
Review · catalog / process analysis.
03 · Identity usage policy

The event does not just say carrier X with confidence Y.

It also carries what the deployment is allowed to do with that result. Confidence and downstream action are separate decisions. The same identity event can be safe for correlation, useful for review, eligible for alerts, or blocked from billing — depending on deployment policy.

identity_event
Northridge Freight Co
VERIFIED · 0.94 · gap 0.73
identity_usage_policy
✓ correlation join to existing violation records
✓ review route to human reviewer
· alert gated by deployment policy
· billing gated by deployment policy
Same event. Same confidence. Different deployment can permit different downstream actions. Most early deployments leave billing off while still producing useful correlation and audit outputs.
04 · Evidence and replay

A reviewer should be able to ask: why?

Every identity event points back to the evidence that produced it, the catalog snapshot it was scored against, and the versions of the models, prompts, and scorer that decided. Months later, the same inputs and the same versions produce the same answer.

01
Selected frames

The evidence frames the cloud reasoned over, by reference.

02
Extracted fields

Per-frame and consensus observations, with per-field confidence.

03
Catalog snapshot

The FMCSA snapshot version used at decision time.

04
Decision versions

Model · scorer · config · prompt versions that produced the result.

An event you can act on, review, or refuse.

Tollscopic DOT# exposes what it knows, what it suspects, and what it refuses to claim. Confidence is a contract, not a mood.

Talk to Tollscopic →