Forked from unclecode/crawl4ai (Apache-2.0), full history preserved, upstream
kept as a remote. We were running ghcr.io/hanzoai/crawl as a verbatim skopeo
mirror of docker.io/unclecode/crawl4ai with no source repo -- so it was
amd64-only, a digest behind, and unbuildable by us. This makes it ours.
Package crawl4ai -> crawl. The upstream name is a compound marketing name;
crawl is the noun, and it already matches the image and crawl.hanzo.svc.
Hosts are mapped explicitly rather than by the identifier rename, because
crawl4ai -> crawl would produce crawl.com -- a real domain we do not own, which
every example and live test would then hit. They point at crawl.hanzo.ai.
The suffix-confusion fixture crawl4ai.com.attacker.net maps to
crawl.hanzo.ai.attacker.net, preserving exactly what that test asserts.
LICENSE, CONTRIBUTORS, SECURITY-CREDITS and SPONSORS are untouched, and NOTICE
records the lineage -- Apache-2.0 s4 requires it, and there is no GitHub fork
badge to carry it since the fork API refused.
.github/workflows is dropped (no GitHub builders); CI lands in .hanzo/workflows.
docker-compose.yml -> compose.yml.
AsyncWebCrawler and friends are NOT renamed: they are descriptive API names, not
brand, and churning the primary public API in the same commit would make both
unreviewable.
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
zooqueenandhanzo-dev
d19bf412f7