Reject duplicate resource ids in the FHIR Bundle assembler

[maziyarpanahi]

Summary

In openmed/clinical/exporters/fhir/bundle.py, to_bundle builds reference_map keyed by 'ResourceType/id'. When two input resources share the same resourceType and id, the second silently overwrites the first in reference_map, so internal references resolve to only one of them (whichever appears last). The assembler claims to produce no dangling internal references, but duplicate ids produce silently wrong cross-references.

Scope

• Detect duplicate 'ResourceType/id' keys while building reference_map and raise a clear ValueError naming the colliding resourceType/id
• Keep the existing behavior for resources without an id (they remain unreferenceable, which is valid)
• Document the duplicate-id contract in the to_bundle docstring

Acceptance criteria

• to_bundle raises a clear ValueError when two resources share the same resourceType and id
• A Bundle with unique ids still assembles and rewrites references correctly
• A regression test asserts the duplicate-id error
• .venv/bin/python -m pytest tests/ -q

Out of scope

• FHIR profile validation
• Synthesizing or deduplicating resources

Files

• openmed/clinical/exporters/fhir/bundle.py
• tests/unit/clinical/test_fhir_bundle.py

---

Task: OM-340 · Milestone: v1.7 · Priority: P2 · Size: S
Depends on: — · Blocks: —
Roadmap: v1.7 expansion (post-1.6.0)
Spec: PLANS/V2/EXECUTION/tasks/OM-340.md

[gitcommit90]

I'd like to take this one — opened #557 with a fix: the FHIR Bundle assembler now rejects duplicate ResourceType/id keys with a clear ValueError instead of silently overwriting cross-references, plus 5 regression tests. Happy to adjust based on review.