I am experimenting with a lightweight solution, but am working out which compromises are reasonable to take so that it is worthwhile but not overwhelming of the machine it runs on. Still have to give it a real test with a large number of files as well.
For the hardware, I figure I could potentially use some old stuff I have (thinking like a Core2 Quad with 16gb RAM and a large hard drive would be fine). I could probably stuff it into an old half-depth 1u server case. The problem now is finding the time to build it...
One criterion I had for a lightweight solution was to not require Java. No problem with Java, just that it is a big dependency and my perception is that it is not a common install on the laptop or desktop of people reading pdfs, at least out side of the STEM stream.
Ideally, it would be nice if the app could do a search across a drive (or NAS, or whatever) for PDFs, pull out a summary and title, and then use that for naming/search/etc. Maybe your app already does this?
To be honest - what I wish I had was a personal Google Search appliance spidering all of my data on my NAS, which was also linked to normal Google, with priority of search results given to local information.
Maybe something like that already exists - I've found open-source solutions that come close, but all for the search/spidering typically required a machine waaaay better than my desktop...