oppiOS / Confidence Engine
Before OPPI acts, it scores the call.
Act, or ask you.
Two of six areas are in service.
This page says which.
Release one floor staff member at 13:30 and hold the second bar shift start until 17:00.
Open a decision to see its signal, diagnosis and confidence.
Example venue. Illustrative data.
Scoring
Predictive confidence asks how good the forecast is here. Intervention confidence asks how well this kind of action has gone before, from similar past episodes and their outcomes. Both feed a composed decision tier, up to and including act on its own.
The gate
In the ASTRA protocol, step four is the gate. The Confidence Engine's composed score meets Clementine's floor there. Above it, resolve. Below it, stage for a person. The execution trace shows the score, the threshold, how many similar episodes it drew on, and how those ended.
In build
The design has six domains: predictive, intervention, epistemic, perceptual, self-model and autonomy. Predictive and intervention are scoring live decisions today. Epistemic scores are stored but not yet consumed. Perceptual is implicit. Self-model and autonomy composition are being built. We would rather list that than round it up.
What it does
How good the forecast is for this venue, now.
How similar past interventions ended.
A tier from ask to act, from the domain scores.
Step four of ASTRA, against the confidence floor.
Score, threshold, similar episodes and their outcomes, on record.
Each episode's result feeds the next score.
Twelve similar episodes, most of them resolved well. The score clears the floor; OPPI resolves.
Two similar episodes, one negative. Below the floor; staged for a person.
Confidence you can read, and a gate you can set.
Two domains in service, four in build, all of it on the record.