Overview
Jev Pages creates a landing page while the user types a few lines about a product. Rather than generating free-form copy, Jev classifies the notes and selects from fixed options for content direction and each layer of the visual design; application code then assembles the page from those choices.
The project can also run without a Jev key by using a simple offline preview. It is distributed as an MIT-licensed repository and requires only Python 3 to run locally.
Design Process
Jev first interprets each note as a headline, point, or fact. It then identifies the product name, offer, page purpose, opening message, subjects worth photographing, product color, and facts that deserve visual emphasis.
The design pass selects the headline and subheadline, overall look, lighting, depth, typography, motion, color rhythm, opening composition, point layout, image band, ending, and photos. A final decision determines which headline word receives emphasis.
Interactive Editing
Every AI question is answered with a fixed option or a yes-or-no choice. Users can click any choice to lock or change that direction, after which Jev recalculates the remaining design decisions around it in about 0.2 seconds.
This approach keeps the supplied product notes separate from the design system: Jev chooses among defined possibilities, while the app's layout, styles, and logic build the result.
Local Setup and Structure
Running the included Python server opens the app locally, with a macOS command file provided as an alternative launcher. A Jev key can be added from the interface; it stays in the browser and is passed to the Jev API through the local server.
The project stores its photo library, visual looks, product colors, and example pages in separate data files. This structure allows photos to be added with a single line and keeps palettes, fonts, shapes, and examples organized independently from the main application.