NewThe examples make the boundary between classification and action concrete: Jev supplies typed probabilities, while ordinary application code applies thresholds, confidence checks, and human-review branches.
The playground sends a user-editable text or JSON state and one or more typed Noul, Choice, or Score questions through its API proxy. It displays Jev's answer and probabilities, then runs example-specific application code over the typed answers to show what a consumer might do. The page is an interactive teaching demonstration; its configured sample states and outcomes are not records of observed production calls.
Is this email spam or phishing?
The pinned email-spam example includes sender, subject, body and a verification URL; it also defines coworker and newsletter alternatives.
Which folder should receive the email?
The pinned email-spam example includes sender, subject, body and a verification URL; it also defines coworker and newsletter alternatives.
The user selects an example and can change its sample state or questions in the playground.
The browser posts model, parsed text/JSON state, and named typed questions to the local API route; the route proxies to the selected provider.
The workbench displays response details and passes typed answers into the example's decide function for its illustrative downstream result.
The examples make the boundary between classification and action concrete: Jev supplies typed probabilities, while ordinary application code applies thresholds, confidence checks, and human-review branches. One request can ask several distinct questions over the same state.
