Design principles¶

Preserve input fidelity through documented codecs. Keep one normalized Storyline model and one owner for storage and query semantics. Pin Source versions for each operation, retain provenance for transformed output, and bound parsing, I/O, and query work. Failed publication must leave the previous committed data readable. Keep datasets portable across files, object storage, CLI, and Web consumers.