As a somewhat different approach to this, you may wish to consider using Apache Stanbol .
Briefly, Stanbol is a system for building semantic web applications. Among other things, it allows you to import the LOC data as n-triples (either in SKOS or MADS/RDF format) . The data is stored in a Solr index and you can query it over HTTP, so it ends up being really, really fast. Plus, all of the RDF relationships are stored in the index, so you can query the API in a lot of different ways. The iks-project has a demo version of stanbol  with a variety of "locally managed entity hubs" -- they don't have LC subject headings, but they do include a lot of other data collections.
I use Stanbol fairly extensively for a number of these types of applications, though I don't expose the API directly to the web. In our system, we have the LOC subject headings, LOC name authority records and the full geonames database. If you do run this, you will want to put it on a system with a good amount of memory.
Systems Administrator and Programmer
Academic Technology Services, Amherst College
[log in to unmask]<mailto:[log in to unmask]>
On Jun 4, 2013, at 7:31 PM, Joshua Welker <[log in to unmask]<mailto:[log in to unmask]>> wrote:
I am building an auto-suggest feature into our library's search box, and I am wanting to include LOC subject headings in my suggestions list. Does anyone know of any web service that allows for automated harvesting of LOC Subject Headings? I am also looking for name authorities, for that matter. Any format will be acceptable to me: RDF, XML, JSON, HTML, CSV... I have spent a while Googling with no luck, but this seems like the sort of general-purpose thing that a lot of people would be interested in. I feel like I must be missing something. Any help is appreciated.
Electronic/Media Services Librarian
Southwest Baptist University