Since MLflow installations are only supported within AIPCC base images, bundling the UI in the Python wheel is redundant, as the MLflow server Konflux container build handles its own UI build. While the original goal was to maintain UI consistency for local use by data scientists, we can simplify things now.

I apologize for the extra work caused by this oversight. Please go ahead and remove the yarn build step. I’ll ensure future sdists exclude those Yarn vendored dependencies.