Changelog

[Unreleased]

Default AI Model Changed: The default AI model has been switched from gpt-4o to gemini-3-flash-preview to make the CLI more accessible out of the box (Google provides a free tier for Gemini API keys). If you previously relied on the default gpt-4o and only had an OPENAI_API_KEY set, you may need to pass --ai-model gpt-4o explicitly.

Improved handleCookieConsent() precision to avoid accidentally hiding non-banner layout elements.
Fixed an issue where the Adblocker plugin would register multiples times during BrowserEngine initialization.

Robust handling for Next.js image optimizer:
- Skip direct downloads of /_next/image to avoid HTTP 402.
- Parse and download the original image from the url= parameter.
- Alias /_next/image URLs to the same local file as the original.
Runtime asset rewriter:
- Rewrites src, href, poster, style background-image,
- Rewrites srcset and imagesrcset (critical for responsive images),
- Observes DOM mutations to rewrite dynamic nodes (e.g., hover/popover content).
- Optional debug logging with [MW rewrite] lines.
Microlink support:
- Captures Microlink images and follows JSON to the actual screenshot/image URL.
- Downloader resolves Microlink indirection and saves the final bytes offline.
Hover simulation and computed asset capture:
- Triggers common hover/popover patterns so previews render during mirroring.
- Surfaces computed background and pseudo-element images into the DOM for capture.

The rewriter is automatically injected for Next.js/React pages.
Use --debug to view [MW rewrite] logs and verify asset rewriting at runtime.