Trace the evidence behind each step A result should be inspectable. This engineering demo connects a visual path to recorded inputs, explanations and source references. Separate inputs, explanations and sources Select an evidence layer. Inspect the recorded inputs, read the explanation, and follow the source identifiers. Each layer stays connected to the same case. A missing record is not a negative finding An absent record must not become a negative finding. Here, a missing note remains distinct from a recorded zero. The explanation preserves what is unknown. Keep unresolved conflicts visible Conflicting evidence stays visible for review. Inspect the explanation and both source identifiers, rather than silently collapsing different records into one answer. Move from a summary back to its evidence Filter the synthetic cohort, then inspect the case behind its state. The table and chart describe the same prepared records. Explicit evidence, reviewable software This supports a trustworthy, knowledge-driven architecture. Neural interpretation is a future integration layer. Start with one evidence inspection module. These are prepared states, not a clinical inference run.