I've temporarily pruned my game project down to a game library project, which I hope to have in a working state in time for next year's 7DRL. If it is, I'll do a 7DRL based on it as a sort of demo for the library.
The library is going to be called Graphlike, and will provide capabilities for procedural level generation based on a higher-level abstract representation of the level as a planar graph. If I can ever get my algorithms sorted (ha), I believe this will represent a useful advancement in roguelike science, and provide a useful starting point for much experimentation.
I've been cracking at this off and on for a bit, and several approaches have gone out the window, but I'm currently going slow-and-steady on one I think will work fine once I've gotten through the slog of it. Guess we'll see!