Module entity_resolve 

v0.19: bundled entity resolution.

Hardcoded lookup table for common entities a curator hand-adds via vela finding add --entities. Maps (normalized_name, entity_type) to a canonical ResolvedId (UniProt for proteins, MeSH for diseases/anatomy, ChEBI for compounds, NCBI Taxonomy for organisms). When a match lands, the entity’s canonical_id is populated, resolution_method = Manual, resolution_confidence = 0.95, needs_review = false.

This is deliberately not exhaustive. The ~50-entry table covers the Alzheimer’s drug-target landscape’s natural vocabulary plus the cross- domain v0.10 enums (particle, instrument). Real research will need either a much larger bundled table or live ontology API integration — but a manual-curation user gets meaningful resolution today, and needs_human_review strict-blockers drop for matched entities.

Structs

FindingResolutionReport

Per-finding outcome.

ResolveReport

Whole-frontier outcome.

Functions

bundled_entry_count

Convenience: how many entries are bundled. Used by tests + the CLI’s --list flag.

iter_bundled

Iterate the bundle without leaking the internal struct shape.

resolve_frontier

Walk every entity on every finding and apply the bundled lookup. Already-resolved entities (canonical_id is Some, regardless of source) are skipped — caller can pass force to re-resolve.