Say I have a domain object which has a filename property; this points at a PDF.

Can I structure the setup so that Hibernate.Search will automatically index the associated file and when I query I get the domain object back rather than the PDF itself?

yes, by writing a custom FieldBridge, you basically can do everything in between the object model and the index structure, including streaming the data from a URL.
Interesting use case

Ok, so first I have to finish porting Hibernate.Search to NHibernate; then I'll write a custom bridge :-)

Would have thought it fairly common problem. We have a web site with articles/brochures on it, managed by a CMS - so you want to go to the page of owning entity rather than directly to the conent.

This is a common problem, but how you extract data from your document is fairly business specific, hence no built-in solution.

BTW, check this out, it will probably be useful for you.

Thanks that's most useful.