Evidence

Evidence is treated as a first-class operational input.

Current repositories include evidence extraction, evidence control, evidence graph, evidence quality, evidence synthesis, policy proof, and appeal-assembly domains.

Evidence classes

For governed reasoning, the current EHR/denial boundary distinguishes:

Class	Meaning
<code>source_fact</code>	Directly present in an authoritative source record
<code>derived_fact</code>	Deterministically computed from source facts
<code>model_inference</code>	Produced by an AI/model process
<code>recommendation</code>	A proposed action or next step
<code>unknown</code>	Required information is unavailable or not established

The important distinction is that an inference or recommendation does not rewrite the source record.

Provenance

Evidence and canonical EHR objects carry source references and lineage where supported by the current implementation.