It fetches data from the public APIs of openlibrary.org and inventaire.io. From my point of view, inventaire is really great as it synchronizes with wikidata and is far better at not duplicating works or authors.