Auto-Repair (self-healing pipeline)

Auto-Repair is the closed loop that runs after Drift Detection fires inside Concord by IaxaI. The repair worker pulls a sample of the new-shape events, calls into the Universal Adapter's mapping inference path with the previous mapping as context, and produces a proposed mapping. The proposal is calibrated against held-out samples to score how well it preserves the canonical OCSF semantics. High-confidence repairs run in shadow mode in parallel with the existing mapping for a configurable window. Output divergence is measured event-by-event. Low-divergence shadow runs promote to active and the previous version retires reversibly. Low-confidence repairs and high-divergence shadow runs escalate to an analyst with the diff pre-rendered. Every state transition (proposed, calibrated, shadow-started, promoted, retired, rolled back, escalated) writes a signed receipt to the audit ledger. Mapping changes become reversible, defensible, and replayable. Patent candidate, in development.