mag · log odds

// emitting a match score.
//
// A learned artifact is stored as interpretable numeric coefficients (aedt_model_coefficients) plus a monotone
// calibration map (aedt_calibration_bins), and served in Go into shopspring/decimal, so serving needs no
// second datastore, no cloud ML service, and no Python at request time. ExecuteModelIntent gathers the model's
// FEATURES by running each declared FeatureIntent through the EXISTING guarded ExecuteIntent path (so RLS, the
// read allowlist, ForbidEEO, and suppression all still hold), reads the champion coefficients, computes
// log odds = intercept + SUM(coefficient_j * feature_j) in decimal, applies the logistic (a single isolated
// float conversion, re quantized to decimal immediately, so every SURFACED number is decimal), applies the
// calibration map, and returns the calibrated probability plus a per feature additive contribution trace
// shaped exactly like engine.ComponentScore. That trace is the citations the existing faithfulness gate grounds
// the narration against.
//
// The first MAG intent is opportunity_win_probability. At v0 it is PRIOR ONLY: the intercept is the logit of a
// defensible federal recompete base rate and every learned coefficient is 0.0, so the model returns the
// calibrated market base rate the day it ships and its learned coefficients light up automatically as decided
// bid labels accrue. It emits opportunity_score, never a candidate score.

log odds = intercept + SUM(coefficient_j * feature_j) in decimal, applies the logistic (a single isolated