Mini JevLOCAL DECISION WORKBENCH
Checking model API…

A frozen model. A finite set of answers.

One message.
Three kinds of decision.

Route a support request, test its urgency, and estimate its priority. Inspect the candidate distribution behind each typed value.

Decision inputs

Shared state

SUPPORT TRIAGE

Edit the message or load another example. Each question reads this state independently.

Questions & candidates

EDITABLE

Start the model API to run a live request.

Typed readout

NO REQUEST YET
Live inference onlyOne shared state → three questions
Choicesemantic key

Which queue?

One candidate key, with a probability for every option.

Noulnumber · 0–1

Urgent or not?

The true candidate’s probability, rather than a forced yes/no.

Scoreexpected stage

What priority?

A probability-weighted stage, with the stage order you define.

Inspect the API exchangeRequest · validated response

The request can also be sent to POST /v1/systemone with the Python SDK or curl. Response metadata describes the last completed request.

Current request

Last response
No response yet. Run the decisions to inspect live model output.