Module ripple 

Ripple propagation of confidence penalties along dependency edges.

When an entity becomes stale or its confidence drops, dependent entities receive an attenuated penalty via BFS up to MAX_DEPTH = 2 hops. Every change is appended to kg_confidence_log.

Functions

add_dependency

Insert a dependency edge: source_id depends on target_id.

propagate

Propagate a confidence penalty from origin_id to all entities that transitively depend on it, up to MAX_DEPTH hops.