scvi-tools

scvi-tools is an agent skill for fitting probabilistic models to single-cell omics data with scvi-tools 1.5.1, released on 10 September 2026. It covers scVI for batch integration, scANVI for cell-type annotation, totalVI for CITE-seq data, MultiVI for joint RNA and ATAC integration, and further models for methylation, peaks and Poisson counts. The skill helps you choose the model and check that its input matches the model's expectations. The examples are illustrative. The repository tests run short CPU fits on small synthetic data for the main models, along with count-preserving preprocessing, the output contracts for differential expression, saving, loading and minification. The tested stack is Python 3.13, Scanpy 1.12.4, AnnData 0.13.4, MuData 0.4.1, PyTorch 2.14.1 and Lightning 2.6.6. Long training runs, hardware-specific paths and specialised workflows are not validated, and the skill says these checks do not establish convergence or biological validity. Use it when a dataset has batch effects, needs annotation transfer, or has paired modalities. It is part of the k-dense-ai scientific agent skills repository on GitHub, which has about 49,600 stars.…