CoreModelsLinkML to Synapse Registered Schema
The LinkML-to-Synapse toolchain every DCC currently hand-builds, with the subset losses declared.
LinkML is where serious DCC data modeling is heading, the NF metadata dictionary proved the pattern, but there is no supported LinkML-to-Synapse toolchain: each team hand-builds its own compile-register pipeline and re-derives Synapse's draft-07-subset quirks alone, a gap made urgent by schematic's 2026 sunset. This recipe is that toolchain, generalized: import a LinkML schema (classes, slots, enums with ontology meanings) as governed Types and Elements, then export the Sage Bionetworks Synapse profile: the draft-07 subset Synapse's JsonSchema REST object actually models, with a registered-schema $id composed as {org}-{name}-{semver}. Constraints Synapse cannot carry are declared in the lossiness ledger instead of quietly evaporating on registration; unconfirmed wire behaviors are recorded as lossiness pending Sage fixtures rather than guessed. synapse is encode-only: its output is plain draft-07 JSON Schema, re-importable via the jsonschema key. Companion curation-manifest and dictionary CSVs (value sets from governed taxonomies) come from the generate_synapse_manifests MCP tool.
14-day free trial. Builder $49/mo after the trial. Cancel any time.