A lot of its reconciliation of old datasets witch might have links to MediaWiki categories or Wikipedia links, as well as a lot of internal Wikis, one could often use Wikidata to find these but a major benefit of doing it this way is that it takes advantage of the fulltext search of MediaWiki.
No not just yet, it's incredibly dependent on our CDN provider, but one of the goals with this attempt is to make it run in additional(and open) runtimes! I'm not sure it make total sense for them to reuse our setup given their stack and our odd set of reconciliation extensions but maybe there is something useful in there.
- do you intend to support data extension or auto-completion ("suggest services"), and if not, why?
Short answer; no because we don't use those. Longer answer; maybe the auto-completion, but not data extension as I don't think data extension should be dependent on a such a specification, instead, on our end data extension is supported through GREL over any HTTP resource as long as it responds with either HTML, RDF, XML, or JSON.
That said I'm very open to add support for things that might only be beneficial for the community, but I would much rather try to align our OpenRefine client and server with upstream.