FHIR Bulk Data checkpoints¶

OpenMed provides a small local-only manifest for resuming a FHIR Bulk Data export. It records the resource type, digests for the opaque page token, privacy policy, and endpoint scope, plus aggregate page and resource counts. It never stores a page token, resource payload, or policy configuration in the manifest.

Create and validate a checkpoint¶

from openmed.interop.fhir.bulk_checkpoint import (
    create_checkpoint,
    validate_resume,
)

checkpoint = create_checkpoint(
    "Patient",
    page_token="synthetic-next-page-token",
    policy={"name": "local-safe-policy", "date_shift_days": 7},
    endpoint_scope={
        "base": "https://synthetic.example/fhir",
        "export": "group-synthetic",
    },
    pages_processed=3,
    resources_processed=42,
)
checkpoint.write("bulk-checkpoint.json")

# Raises BulkCheckpointCompatibilityError if any resume identity changed.
validate_resume(
    checkpoint,
    resource_type="Patient",
    page_token="synthetic-next-page-token",
    policy={"name": "local-safe-policy", "date_shift_days": 7},
    endpoint_scope={
        "base": "https://synthetic.example/fhir",
        "export": "group-synthetic",
    },
)

Policy and endpoint values may be mappings or other JSON-compatible values; mapping keys are sorted before hashing, so equivalent configurations produce the same fingerprint. None is supported for the final page marker. Resume validation compares the manifest version, resource type, page-token digest, policy fingerprint, and endpoint-scope digest. A mismatch fails closed and the exception names only the mismatched field, never its value.

The manifest is deterministic JSON and performs no network call. Filesystem I/O happens only when write, write_checkpoint, or load_checkpoint is explicitly called. The checkpoint is progress metadata, not a clinical decision or a substitute for a FHIR server's export guarantees.