Take a look at Swish-e. It's a search engine written in Perl and it can be configured to index PDF files using filters that translate PDFs into text or HTML. Once you've got that setup, check out CGI::Application::Search, a very easy to use front-end for Swish-e.