I don't have a very large collection of connectors, because I don't really use too many different ones myself and wouldn't know where to start. I'm happy to take requests, though. In the beginning while the library is still somewhat small, I'd like to put a "request box" on the web site, and just sort components by popularity and add them. This is meant to replace KiCad's standard library, so in terms of "completeness", it's supposed to have a usefully complete selection of a broad variety of components. I'm not too interested, for instance, in collecting the entire Atmel repertoire, but I've got pretty much every basic part, from resistor networks to P-channel JFETs to the 7400 series.
On the to-do list is a program to render schematic symbols and footprints to PNG (for browsing the library online) and the PDF, as you mention. One of the places where I'm putting the most emphasis is on scripting everything that I can, to ensure consistency, because consistency is one thing that is horribly lacking in KiCad. I'm not just going to throw the footprints onto a PDF myself, that leads to the PDF becoming outdated. That'll probably take a few days to do correctly.
All parts do have proper footprint filters. Like I said, I'd like to eliminate the CvPcb step to the point where you can just click "Perform automatic footprint association". This means using separate schematic parts for different footprints, so each part has exactly one footprint listed, more like every other EDA package out there rather than like "the KiCad way". These are all generated automatically from a template part, so if you like the KiCad way you can just use the template parts. Template parts also have filters, I haven't neglected that. There's no SMD/PTH "default", you can either place SMD or PTH parts. I'm trying to include as much as possible, but there's definitely a bias towards SMD because that's what I use almost exclusively.
I am trying to keep naming schemes as consistent as possible. I haven't actually written rules for this yet, but I plan to.