Skip to main content
Termlib

Changelog

What changed in each release

Newest first. Current version: v0.4.0.

  1. v0.4.0Current

    A redesigned canvas, sharper recognition, more reliable pages (v0.4.0)

    Mockup canvas

    • The canvas now holds only page artboards. Select a page, or one block on it, and say what to change in the prompt bar at the bottom.
    • Style is a canvas-wide setting. Changing the primary color, mode, radius, or font size reskins pages instantly, with no regeneration.
    • "New direction" makes two versions at once, shown next to the page so you can compare and adopt one. Deleting a page can be undone.

    Page generation

    • A shared design system, with the app menu, site navigation, and footer rendered by the system, so every page in a set matches.
    • Pages are planned block by block before they're written, and facts that aren't in your brief, like prices or user counts, are no longer made up.
    • Each page is checked for blocks, layout, overflow, overlap, and contrast, and fixed once automatically at no extra cost. Pro plans can polish a page from its screenshot.

    Recognition and glossary

    • The extension's local recognition now combines several signals, so far fewer elements go unrecognized. Results show as sure, maybe, or not recognized, and you can report a wrong match.
    • Running AI recognition on the same element again reuses the cached result at no cost.
    • 40 new component terms (page header, form item, chat bubble, pricing table, FAQ, and more), 225 in total.

    Agent upgrade

    • Replicas use placeholders for text that can't be read instead of inventing numbers and dates. Upgrades and edits are self-checked, and the result page shows what was found.
  2. v0.2.0

    First release (v0.2.0)

    Glossary

    • Terms in six categories: layouts, UI components, interactions, styling and CSS, frontend engineering, and AI coding. Each one has a plain-English definition, a wireframe, and “How to tell AI”.
    • Browse by category and tag, search by name and alias, and save terms once you're logged in.
    • Guides: topics grouped by the kind of page you're building.

    Browser extension

    • Hover inspection: matches components locally by DOM and component library classes, with AI identify as a fallback.
    • One-click full-page analysis, with results in the side panel.
    • Works in Chrome and Edge, and only runs on local development addresses by default.

    Page analysis and suggestions

    • Analyze a whole page from a screenshot or a public URL: overall layout, blocks, components, and content traits.
    • Every block is checked against hand-written rules. Reasons quote the glossary, and you can accept or ignore each suggestion.
    • Before-and-after wireframe previews for a single suggestion.

    Agent upgrade

    • Replicates the page, then applies the suggestions you accepted, with follow-up edits, version rollback, and HTML download.
    • Turns suggestions and edit requests into prompts for tools like Cursor and Claude Code.

    Accounts and credits

    • Email signup and login, shared between the extension and the website.
    • Signup credits, redeem codes, and a credit history.
    • Your own API key: any OpenAI-compatible API, set separately for vision and code tasks, with no platform credits used.