Cookbook

Task-oriented recipes. Each page is self-contained and runnable end to end — copy the code block, uv run python it (with triwarp installed), and it prints real diagnostics from a real, intentionally imperfect input.

Recipe	What it covers
Cleaning and remeshing a rough mesh	repair.make_solid → remesh.isotropic_remesh on a mesh with a hole and stray debris
Point cloud to watertight surface	reconstruction.screened_poisson and reconstruction.ball_pivoting, side by side
Geodesic distance fields	heat.heat_geodesic, precomputed operators, and where to go for vector heat / log maps
Aligning two scans	registration.procrustes for a coarse fit, registration.icp to refine it

Every recipe here is also a starting point rather than a finished pipeline — each links back into the relevant API reference page for the full set of keyword arguments a real use case will eventually need (a target edge length that varies over the surface, a feature-angle threshold, a solver tolerance). See also Concepts for the design principles these recipes all lean on, and Migrating from another library if you're translating an existing pipeline built on trimesh, Open3D, libigl, MeshLab, potpourri3d, or PyTorch3D.