My goal is exactly that, usability without me present :-).

The tools for processing PDFs into searchable text have a lot of warts. For a while IBM was offering a free Watson service to do this (now its part of Watson Discovery) which has some warts. I did manage a set of perl scripts that would post process the statements that I downloaded from the bank into CSV files, but I would still like to pull tabular data out of PDF book scans to make the data they provide more useful.

I have a simple frontend based on the perl Mojolicious module which Blekko had developed as part of another project but my indexing tools are still quite primitive. Simple bi-gram and tri-grams, and a growing synonym index. I don't give it enough queries to use my own traffic for ranking feedback. So basically everything is nearly equal rank. Basically I am about to the AltaVista level of search capability :-).

The vision is it just runs as a server and anyone on the same network can access it like a web service an pull up documents (and in the future media) of interest.

For my eBooks, I use a java app called DocFetcher. It's indexing and word search capabilities are pretty nice, and I've had great luck with it, especially as I can set it up portably.

Did you get very far with other enterprise stuff like Sharepoint?

I didn't try Sharepoint, although I looked at Elastic Search on AWS briefly. The goal though was to have all of this stuff on premises both for latency reasons and to maintain a credible defense should someone come after me for copyright infringement.

A bit of a tangent, but if you use a Mac, I've found that Dash is really great for getting things done without internet access. It's main purpose is to pull in local copies of docsets and search those, but it can also pull in sections of stack overflow (such as all Python questions, or all Pandas questions), and it integrates with things like Alfred, so you can do scoped searches without any GUI interaction.