Migrate File

Store one bulk-imported document for company_id without running the pipeline.

One file per request, mirroring POST /files minus its multi-file form: an import driver wants to know which of ten thousand documents failed, and a partial success inside a batched request cannot say that.

The folder can be given as project_id or, for a driver working from the source system's folder names, as project_name — created for this vendor if no folder of that name is there yet. Giving both is refused rather than resolved.

The rows are stamped with the time of the import, not the document's date in the system being migrated from — created reads as the migration date by design.