scvi-framework

To route your single-cell analysis, this framework directs your data to the appropriate scvi-* child model based on your specific dataset and analysis goals. It automates model selection by enforcing a shared training contract across all available scvi-tools variants.

The standard data contract for scvi-tools interoperability requires child models to implement a strict setup_anndata, train, get_latent_representation, and save/load sequence. This shared contract standardizes workflows and ensures seamless compatibility across multiple single-cell omics models.

To get latent representations from anndata, the framework enforces a standardized workflow: run setup_anndata, train the selected scvi-* model, and call get_latent_representation. This sequence ensures consistent latent space extraction across all routed single-cell models.

Yes, you can add custom scvi-* model variants to the workflow. The framework is extensible, allowing new child skills to override local concerns while reusing shared components and adhering to the standard setup_anndata, train, and save/load contract for full interoperability.

Yes, you need anndata, numpy, and scipy installed to use this framework. These dependencies are required to manage single-cell omics data structures and execute the shared model training contract across the routed scvi-tools variants.

For scvi-tools model selection, the framework applies pre-train checks and validation utilities to ensure your anndata input meets the shared data contract. These cross-cutting references and verification steps confirm compatibility before routing to a child model for training.