URL fetch for convert-html-to-pdf now rewrites CSS and image references to absolute http(s) URLs instead of inlining file bodies.

Date: 2026-09-06 10:48 Summary: URL fetch for convert-html-to-pdf now rewrites CSS and image references to absolute http(s) URLs instead of inlining file bodies. Visibility: people

[question] css和图片返回完全链接引用即可

Tool links (English)

https://onlinefreetools.org/tools/convert-html-to-pdf Chinese: https://onlinefreetools.org/zh/tools/convert-html-to-pdf

[try to solve]

Context

The previous direction was to fetch CSS/JS and embed their bodies in the returned HTML. The user then asked to keep CSS and images as complete URL references only—no inlining.

Process

Removed the extra Worker fetches that pulled stylesheet and script bytes. The HTML API still: (1) GET the page with SSRF guards, (2) inject <base href> as a fallback, (3) walk the HTML with HTMLRewriter and rewrite link[href], img[src]/srcset, source, video[poster], SVG image, input[type=image], inline style url(), and <style> url()/@import to absolute http(s) URLs resolved against the final page URL. data:, blob:, hash, and blocked hosts are left unchanged.

Root cause / analysis

Inlining CSS/JS costs Worker subrequests and still does not make html2canvas see cross-origin fonts/images. Absolute URLs are enough for the browser (and <base>) to request assets from the origin site. Scripts remain stripped in the tab before PDF.

Solution

src/tools/convertHtmlToPdf.ts returns one HTML document whose CSS/image refs are absolute. Copy in ten locales states that rewrite. Local wrangler must be restarted so the new Worker handler is live.

Notes / boundaries

External CSS files are not downloaded, so url() inside those files is unchanged (the file is still requested from the original host). Cross-origin images can still fail html2canvas without CORS. JS is not executed.

[actions]