Writing System Selector: The font viewer's preview now offers a writing system dropdown listing only the scripts the loaded font actually covers (detected via the existing canvas coverage probe — no font parsing or extra network). Picking a script fills the preview with a native sample sentence for it. Covers Latin, Greek, Cyrillic, Armenian, Hebrew, Arabic, Devanagari, Thai, Ethiopic, Japanese, Chinese, and Korean. The dropdown is hidden when a single script is detected.
Download Button: A toolbar button downloads the formatted source you're viewing. The filename follows the server's Content-Disposition, the URL's own filename, or a download.<ext> default. A downloaded HTML page gets a <base> pointing at the original URL so it still renders (styles, images, scripts) when opened locally, and a downloaded CSS file has its relative url(...)/@import references rewritten to absolute so it keeps finding its fonts and images. (Cross-origin webfonts that the server doesn't expose via CORS won't render from a local file — a fully self-contained download is planned for later.)
Search in Source: Cmd/Ctrl+F (or the toolbar search button) now searches the whole source, not just the part rendered in the viewport. Matches are highlighted and scrolled into view as you type, with next/previous, match-case, whole-word, and regular-expression options. The search bar is localized in all languages.

Locale-Aware Preview Text: The initial preview text now follows the browser's UI language, derived from its ISO 15924 script (e.g. Japanese, Chinese, or Cyrillic instead of a Latin pangram). This replaces the previous per-locale sample string with a single source of truth shared with the writing-system selector, so the initial text and the "Latin" menu entry no longer diverge.
Direct Source Links: Links to CSS/JS/JSON/XML files in the source now point at the real file URL (still shown in place by the extension when navigated to) instead of an encoded viewer.html link, so the address bar keeps a clean, copy-able URL. HTML links still open through the viewer to show their source.

Source Text Encoding: Pages that declare their charset only in an HTML <meta> tag or via a malformed Content-Type header were decoded as UTF-8 and showed replacement characters (�) on accented or non-Latin text (e.g. développeurs). The source is now decoded using the charset from the HTTP header (parsed leniently), falling back to the <meta> declaration, then UTF-8. The page-weight status also reflects the true transferred size.