freedium.cfd no longer resolves, and freedium-mirror.cfd has been
rewritten as a SvelteKit app that renders the post inside <main> rather
than the old <div id="main-content">. The content check therefore
rejected every mirror response, so Medium articles always fell through
to the archive fallbacks.
Accept either marker, and reject the mirror's error pages explicitly:
they return HTTP 200 with a streamed error payload and ship an empty
<article> skeleton, so <article> alone does not prove a post rendered.
Also fix an unrelated crash in process_html_document that blocked this
path (and every page with an inline script): get_text() has no "default"
keyword, so it raised TypeError.
Closes#49
Change the /status endpoint to accept the target URL via a query parameter instead of a path component. Update client-side EventSource to call /status?url=... and import urllib.parse.unquote to decode the incoming URL. Return a 400 error when the url param is missing. This avoids routing/encoding issues with encoded URLs.
Replaces the previous synchronous form-POST flow (which left users
staring at a white screen for the duration of the upstream fetch)
with an SSE-driven status page that shows real-time progress through
each stage of the bypass attempt.
Bypass pipeline:
- Direct fetch with Googlebot UA (unchanged for non-Medium URLs)
- Detect Cloudflare-style anti-bot challenge pages in the response
- Fallback 1: Freedium (freedium-mirror.cfd, then freedium.cfd) for
medium.com hosts — short-circuited before direct fetch since
Cloudflare consistently blocks the Googlebot UA from non-Google IPs
- Fallback 2: archive.org Wayback Machine snapshot
- Fallback 3: archive.today / archive.ph / archive.is / archive.li
mirrors (all four tried in sequence)
- Surface a clear error if every source fails
UI:
- Animated step list with spinner / checkmark per stage
- Progress bar and elapsed-time counter
- Honest error box with retry button instead of a blank page
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Wasi Master