OpenClinic

Provider-facing clinical workspace prototype for Apple platforms. It combines SMART on FHIR import, SwiftData charting, dermatology photo workflows, and on-device clinical intelligence built on Apple Foundation Models.

Provider Workspace, Not a Patient App

A provider-side clinical workspace. It imports records over SMART on FHIR, charts them on the device, and answers questions about them without sending anything out. A prototype, not something to put in front of a patient.

Prototype · not on the App Store

OpenClinic is a SwiftUI and SwiftData clinical workstation built for provider-side workflows. It imports records through SMART on FHIR OAuth, preserves provenance on imported records, stores local chart state on device, and uses Apple Foundation Models plus a clinic-specific retrieval stack derived from OpenIntelligence internals for chart Q&A and note drafting.

Prototype status: OpenClinic is an architectural prototype. It is not approved for live clinical deployment, does not replace an EHR, and should not be represented as a production medical system.

Capabilities

What Makes OpenClinic Different

Provider-first workflow

Daily schedules, patient routing, visit states, and chart review live in a native multi-pane workspace designed for clinicians, not consumer health tracking.

SMART on FHIR import

Uses SMART on FHIR discovery plus ASWebAuthenticationSession to import Patient, Condition, MedicationRequest, and Appointment resources from supported R4 sandboxes.

Local charting and provenance

SwiftData persists notes, appointments, medications, and photos locally while preserving source-system metadata and sync lineage for imported records.

Dermatology-friendly imaging

Region-tagged clinical photos, body-map placement, and lesion timelines support dermatology-style follow-up workflows.

On-device note drafting

Structured encounter dictation and local summarization use Apple Foundation Models to assist with chart creation without sending PHI to third-party model providers.

Clinical retrieval with verification

Core ML embeddings, SQLite FTS5, reciprocal-rank fusion, and a 9-gate verification layer are used to keep chart answers scoped, grounded, and patient-isolated.

Under the hood

Technical Profile

Core APIs

  • UI: SwiftUI, SwiftData
  • Models: Apple Foundation Models, Core ML
  • Interoperability: SMART on FHIR, ASWebAuthenticationSession
  • Retrieval: SQLite FTS5, local vector store

Data boundary

  • Local: Charts, notes, photos, vector indexes, provenance metadata
  • Remote: OAuth and FHIR resource fetches from configured servers
  • Not present: HealthKit sync as an active product feature
  • Status: Prototype only, no writeback and no deployment certification

Workflow

How OpenClinic Runs

1

Authenticate and import

The app discovers SMART endpoints, performs OAuth through ASWebAuthenticationSession, and imports patient-context data into local models.

2

Persist local chart state

SwiftData stores appointments, medications, local clinical records, and image metadata while Keychain protects connection tokens.

3

Index and retrieve context

Clinical records are chunked, embedded locally with Core ML, indexed into SQLite FTS5 plus a local vector store, and fused at query time for patient-scoped retrieval.

4

Synthesize and verify

Apple Foundation Models generate chart assistance on device, and a 9-gate clinical verification pass checks grounding, numeric sanity, contradictions, and patient isolation before display.

Support

Support and Privacy Routes