Make every version immutable
Each build receives the next sequential version and is assembled in a private staging directory. Publication happens through a final rename only after every required artifact exists.
Why this project exists
Small data projects often start as a notebook or one-off script. As soon as the output matters, basic operational questions become difficult to answer: which inputs were used, what failed validation, what changed between builds, and which schema produced the artifact?
Dataset Foundry explores how much production-minded behavior can fit into a focused local tool. The goal is not to imitate a distributed platform. It is to make one complete dataset workflow deterministic, inspectable, and safe before adding more infrastructure.