Read data directly from wz file. This would be the most convenient for the user, but existing libs might be difficult to port to .NET core, contain proprietary data and appear to have some "staticness" to them (which might make them more error prone in my app and might require rebuilding).
Read data from nx files. Should be easier to port to .NET Core than the wz solution, less likely to contain proprietary data, but the user will have to do the conversion personally.
Build converter to my own format based on the WzDumper (or ReWz library). This should be relatively easy and I don't even need to port anything to .NET Core (the converter can stay in the .NET Framework realm). The format appears organically as I code (sanitized version of wz data). The only problem is storing this data efficiently (dumped v83 data takes a more than 20GB). Maybe compressed JSON or SQLite.
Dump actually takes only about 5GB, which means that sticking all files together with ductape should be enough (the problem is the number of files, not their size).
Options:
dumped v83 data takes a more than 20GB). Maybe compressed JSON or SQLite.