Tell us what you want — we route you to the source that actually knows. Software questions go to GitHub. Music goes to MusicBrainz. Shopping goes to Amazon. Plus our own 4.7M-entity knowledge graph and a 32M-page web index.
A keyword-rules layer catches the obvious shapes (“react hooks”, “X lyrics”), and a Qwen3 embedding model handles the long tail — the same “vertical selection” pattern big search engines use.
~2,000 hand-curated + Wikidata-derived routes. Each intent maps to the sources that genuinely own that vertical — ranked, with the query pre-filled.
If your query is an entity we know, you get the canonical 4ort.xyz page with cited facts — plus page-level hits from a Common Crawl index ranked by domain authority.