OSZK Webarchívum

Load it

import { createArchive, providers } from "@agntn/archives";

const archive = createArchive(providers.oszk());
const response = await archive.snapshots("oszk.hu");

What it lists

snapshots asks webadmin.oszk.hu/pywb/cdx and gets NDJSON back, one capture per line. A bare domain is a prefix, so oszk.hu lists every page under it, not just the front door. Rows carry status, mime, digest and length in _meta. The provider asks for those six fields through fl and nothing else.

Yes, the public archive lives on a host called webadmin. It's the link the library itself puts on webarchivum.oszk.hu, not a back door.

Is it just another Wayback?

Pretty much, and that's the nice part. It's plain pywb, so limit, from, to and sort=reverse do what they say. No streaming tricks, no exact URL rule. One catch: limit=0 means "send everything" to pywb. The provider never sends it and answers with an empty list instead.

Reading bodies

Captures replay raw under /pywb/<timestamp>id_/<url>. A read without an exact timestamp takes the newest capture that answered 2xx. Revisit rows (warc/revisit, status 0) only win when nothing better exists.

Ask for the 301 from oszk.hu to www.oszk.hu and pywb quietly hands you the www page instead. Same URL key, so it treats the redirect as noise. No redirect, no new URL. Want the 301 itself? This archive won't give it to you. Only Memento-Datetime says which capture came back, and that's the one _meta.timestamp and snapshot report.

Gotchas

• Coverage is Hungarian. The library's own site goes back to February 2019, younger than you'd expect from a national archive.
• What's online is the public part. The rest sits on reading room computers in the library, in Buda Castle. Hence the icon.
• The index canonicalizes www and the scheme, so https://oszk.hu/ and http://www.oszk.hu/ share rows.

Where it lives

src/providers/oszk.ts.