model-on-demand-packaging

Keep the app lightweight by downloading ML models only when users enable a feature, never bundling large models into the installer. Use when adding model-backed features, download flows, model storage, caching, or packaging exclusions.