Roadmap
Tapiti is young. This page says plainly what is unfinished, so you know what to rely on.
Growth layout (prototype)
Besides the default layout (one room per grid block), the generator can grow rooms one from another, off a parent's door, in free space. It gives an organic, non-lattice floor and scales well to large worlds (about 165 ms for 81 rooms on a 16×12 grid), but it is marked as a prototype because:
- It often stops short of the room count you ask for. On the default parameters it reaches roughly 75 to 80% of the rooms the grid layout would give.
- Too many floors are rejected for the stairs being too close to the entrance (5 to 10% fail after all draws on levels 1 and 5, and floors cost two to four draws instead of one). Growth should favour depth, and the interface should check
depthagainstminHops. - Topology profiles mostly do not fit it (
backbone,redundantExitandlongLooprarely validate);hubworks. - Leaf chambers are not generated, although their parameters stay visible.
- It has no automated tests yet, and the old
frontier-growthexperiment still sits in the generator behind a feature flag. - The browser lab has no controls for it.
Other open items
- Junction conductance (
junction) is a parameter of the reference model that the generator does not use yet. - Delaunay proximity is in the model but not in the generator (RNG, Gabriel and Near are).
- Global scoring of topology profiles, mission objectives and multi-floor campaigns are described in the model as proposals.
- The lab as the desktop app. The browser lab is the original page. Compiling the desktop interface (egui) to WebAssembly would unify them, and is postponed.
- Publishing the crates on crates.io and tagged releases with desktop binaries.
Done
- Rust generator, validator (V1 to V5) and graph measures, byte-for-byte deterministic.
- Native desktop app with progressive drawing, English and Portuguese.
- This website, in English and Portuguese.
Found something missing? Open an issue.