NexusHarnessGraph Naming system

Status: draft · 1.26.0 · JSON

Identifies AI-harness graphs acting as Device actors in the EMR (Task.requester, Task.owner on assignment-to-agent, Provenance.agent.who). References under this system are LOGICAL: Reference.identifier carries the graph name and Reference.type carries 'Device'; no Device resource need exist on the FHIR server -- the identifier points into the harness's graph registry. Reference.display SHOULD carry the human-readable graph name. Granularity is deliberately the graph/feature level (what an auditor or clinician can recognize), never per-agent or per-model. WHERE THIS URI IS ALREADY WRITTEN, AND WHERE IT IS NOT, DIFFERS BY ELEMENT. As the actor -- Reference.identifier.system on Task.requester, Provenance.agent.who and Composition.author -- Nexus EMR writes the URI declared here today, so a consumer reading the actor resolves one system and needs no second address. The same graph name also travels as run lineage, as the system of a Coding in meta.tag beside the run tag under NamingSystem/nexus-harness-run, and THAT writer is still on the pre-migration address https://nexus.awaremd.com/summary-id/graph, which this specification does not publish; moving it is tracked as FHIR-377. Write the URI declared here on both. Reading meta.tag, expect the pre-migration address on everything stamped before that writer moves and the URI declared here after; reading a logical Device reference, expect the URI declared here alone.

Type	Value	Preferred
uri	../NamingSystem/nexus-harness-graph	true
uri	https://nexus.awaremd.com/summary-id/graph	false