I see that you are progressing well with the "technology" side of csv data storage but I wonder (because I am interested in alternative methods) how you are coping with the complexity of the information itself.
I have been always wary of csv because I associate it with spreadsheets and though I have tried very hard from the early days of Lotus 123 I have not found a way to use them to "turn" data to cope with its 3D nature.
What I mean by turn is illustrated by the diagram to the left which suggests it is natural to use a cross section to review the number of floors in a building but plan section to review each floor's spaces. I think this is sufficient to demonstrate data is 3d in nature (or perhaps more accurately access to it). The animation I have demonstrated before allows me to cope with it in a 2D UI.
Information that is often presented as schedules - rebar, door, sanitaryware, ironmongery(hardware) and so on - can be dealt with in the same way, but the properties of the real (as opposed to code) objects require a data or backup display so that you can see collections of key/value pairs for editting.
So assuming you agree basically with what I describe, it would be would interesting to see both a UI design and the structure of the csv or hash solution to compare with the UI and Javascript object notation (json) that I currently use.
Cheers!