Reinforcement feedback worker
The operational worker that tracks pending and completed recommendations, computes exponential moving average confidence updates per pattern, and writes updated confidence values to both the serving index and the audit trail.
The reinforcement feedback worker closes the learning loop for operational pattern recommendations. When the system recommends an intervention for a detected pattern, that recommendation is logged as pending. The feedback worker watches for the downstream outcome: did the intervention resolve the situation, partially resolve it, or fail to help? Once an outcome is observed (within a configurable timeout window), the worker retrieves the original recommendation, computes the recommendation quality score, and applies an exponential moving average update to that pattern's confidence value in the pattern library. Exponential moving average smooths over individual noisy outcomes while still allowing confidence to trend clearly in response to consistent success or failure patterns. The updated confidence value is written to both the serving index (so future recommendations from that pattern are appropriately weighted) and the audit trail (so the full confidence history is auditable). Patterns whose confidence falls below a minimum threshold are flagged for human review rather than silently continuing to generate low-quality recommendations.