Instinct UI Selector
Natural-language requests select and configure real UI components.
Added to Jevfast
How it uses Jev
From the user’s description, code proposes bounded entity, field, and icon candidates. Jev selects a component family and leaf, scores and types candidate fields, and chooses style and layout values. Application adapters turn the typed results into real shadcn component props.
What Jev decides
Decide whether a full page is neededExample answers · not a recorded Jev response · Source ↗
Question 1 · scope
YOUR APP
INSTRUCTION
Does this case need a complete page section, such as a dashboard with navigation or a full login/signup page, rather than a single UI piece?
STATE
The user case text, after deterministic extraction of candidate entities and field spans.
JEV · NOUL
YesNo
App workflow
- Extract candidates
Pattern-based code proposes entity and field spans from the user’s case.
- Choose component and properties
Jev evaluates component families, candidate fields, types, and style settings in bounded questions.
- Refine uncertain picks
The pipeline reranks family members and can issue a head-to-head question when the leading choice is ambiguous.
- Render the interface
Adapters assemble the selected component’s props and the app renders the real UI.
Why it is interesting
The pipeline turns a free-text request into real UI components without having Jev generate code or copy; candidate extraction and component rendering remain ordinary application code.