Synapse schema migration

Sage archived schematic in April 2026 with an end-of-year deprecation, pointing DCCs at Synapse-native JSON Schema curation, and shipping a manual authoring path, not a migrator. The gap in a manual migration is fidelity: Synapse's JsonSchema service supports a subset of draft-07, and hand-subsetting a data model means silently losing constraints nobody inventoried. This recipe is the inventory. Bring your data model as JSON Schema (the shape schematic-era pipelines compile to) or LinkML (where DCC modeling is heading), and export the Synapse profile: the draft-07 subset the JsonSchema REST object actually models, with a registered-schema $id composed {org}-{name}-{semver}, and every keyword the subset cannot carry declared in the lossiness ledger, which IS the migration fidelity report. Honest scope: schematic's native CSV/JSON-LD model files are not decoded directly, and schematic-specific validation rules do not translate; unconfirmed Synapse wire behaviors are recorded as lossiness pending Sage fixtures rather than guessed. Companion curation-manifest and dictionary CSVs (value sets included) come from the generate_synapse_manifests tool.