Eight spatial models, one package
scVIVA-Tools packages scVIVA, ResolVI, DestVI, DiagVI, Harreman, gimVI, Stereoscope, and Tangram — deconvolution, denoising, niche-aware differential expression, gene imputation, and cell-cell communication — behind a single, consistent API.
Unified AnnData-first API
Every model follows the same setup/train/extract pattern, so switching from deconvolution with DestVI to niche-aware modeling with scVIVA means changing the class name, not relearning an API.
One install, shared infrastructure
A single pip install scviva-tools replaces juggling separate packages, docs, and release cadences for each spatial method, with optional SpatialData/squidpy interop via the spatial extra.
GPU-accelerated where it counts
Neighbor-graph-heavy steps (e.g. ResolVI, scVIVA) can use GPU acceleration via the optional rapids extra (cuML/cuPy/cuGraph), so analyses scale with dataset size.