Skip to main content
The FHIR source reads clinical resources from any FHIR R4 server: patients, observations, conditions, encounters, medication requests, allergy intolerances, immunizations, and procedures. It is built with an HTTP manifest and is alpha. It targets FHIR R4.

​
Configuration

​
Resources

Each resource runs a FHIR search (GET /<ResourceType>) and returns one row per Bundle entry. Every row carries id (the resource ID), last_updated (meta.lastUpdated, when the server provides it), and the full resource as JSON so no clinical detail is lost in projection. The picker defaults to patients, observations, conditions, encounters, and medication requests. The other three resources are opt-in.

​
Modes

All resources support full reads and incremental reads. Incremental sync tracks last_updated (meta.lastUpdated) per resource and re-requests with _lastUpdated=ge<timestamp> — the ge prefix carries “greater or equal” semantics; a bare timestamp would mean “equal” and silently return the wrong result set. A 5-minute overlap window re-fetches recently updated records so retroactive updates are not missed. If interrupted, a full read of a top-level resource resumes after its last completed page.

​
Behavior

  • Auth: none. v1 targets unauthenticated servers such as the public HAPI test server. Authenticated servers (OAuth2 client credentials, SMART Backend Services) are a planned follow-up.
  • Pagination: each resource walks every page (_count=100 per request) by following the Bundle link entry whose relation is "next". The walk ends on the page with no next link.
  • Rate limiting: five requests per second, a polite default for the shared public HAPI server. Lower it further for smoke tests; do not backfill against the public server.
  • Response shape: FHIR search returns a Bundle; rows come from Bundle.entry[].resource. Empty result sets emit no rows.
  • The public HAPI server holds synthetic test data only. Point base_url at your own FHIR server for real records.