Module promotion 

Signal aggregation + Bayesian champion-vs-challenger promotion math.

Pure functions, no I/O. Callers (CLI, adapters) translate evolve_storage::signals::Signal rows into SignalInput before calling into this module.

Structs

AggregationConfig

Per-kind weights used by the aggregator.

PromotionConfig

Configuration for promotion_decision.

SignalInput

One normalized fitness signal feeding into aggregation.

Enums

Decision

Outcome of a promotion evaluation.

SignalKind

Whether a signal was contributed explicitly by the user (evolve good/bad/thumbs) or inferred implicitly by an adapter from the session log.

Functions

aggregate

Collapse a session’s signals into a single fitness score in [0.0, 1.0].

posterior_probability

Monte Carlo estimate of P(challenger > champion) under beta-binomial posteriors: Beta(1 + wins, 1 + losses) per arm (uniform Jeffreys-like prior).

promotion_decision

Evaluate whether the challenger should be promoted, held, or needs more data.