Standalone Sister App + Trained Model Surface

Polyfolds

This app will expose a trained neural-network workflow for classifying and repairing polyhedral nets. Data generation and model training stay offline in the sibling polyfolds development folder.

Planned User Flow

Input a net, classify it as valid or invalid, and return either the same structure or a corrected one.

Offline Dev Workspace

/srv/polyfolds

Use this folder for one-time data generation, manifest building, training, and evaluation.

Cloud Storage Target

gs://aix-labs-polyfolds

Once the initial repo is committed, this app can be wired to a dedicated bucket and deployed as its own service.