An open-source sandbox for report-pipeline integration patterns: FHIR and HL7 ingestion, audit-grade import runs, site-scoped access, and explainable downstream triage — deployable locally in minutes.
Research-use workflow software for evaluation, demonstration, and pilot-style exploration under your institution's governance. Not a production clinical system and not regulatory-cleared.
- FHIR
DiagnosticReportingestion including inlineReference.identifiermetadata preservation and text-likepresentedFormattachment decoding (plain text and XHTML narratives) - HL7 v2 ORU ingestion with base64
EDdecode, customMSH-2separator handling, escape-sequence normalization, repeatedOBX-5merging, and subcomponent-aware metadata extraction - CSV / JSON / JSONL report imports for simpler feeds
- Persisted import-run audit records with stable failure buckets (
parse_error,unsupported_payload,validation_error,site_scope_rejection) and cross-actor visibility rules you can verify yourself - Pluggable pilot auth modes: mock auth for local work, trusted-proxy and header-based identity forwarding overlays for controlled demo environments
- A smoke matrix that proves it: ~25 make targets covering success, failure, shared-visibility, and site-rejection paths against real built images, plus hosted CI smoke runs
- Every import outcome is auditable: who imported, what failed, which bucket, what's visible to whom
- Deterministic triage downstream means integration testing has stable expected outputs
- Apache-2.0 — adapt the adapters, keep your changes, or contribute them back
- Honest boundary-setting: docs distinguish what's automated, what's manual, and what's out of scope
git clone https://github.com/alex-varga14/pancreatic-signal
cd pancreatic-signal
make pilot-proxy-demo-up && make pilot-proxy-demo-fhir-smokeFull overlay and smoke matrix: DEPLOYMENT.md. Interop gaps you'd need filled (other resource types, other HL7 segments, IHE profiles)? Open a use-case interest issue.