retraining-worker

What problem does it solve?

This Skill explains the retraining worker that reacts to drift signals, rebuilds anomaly detection models from recent data, and promotes a better challenger without manual intervention.

Core Features & Use Cases

• Kafka-triggered retraining: Shows how retraining requests are consumed from the pipeline and dispatched safely to background workers.
• Celery execution model: Documents the subprocess worker pattern, task retries, and non-blocking dispatch used for model training jobs.
• Shadow testing and promotion: Covers how the new model is evaluated against the current one, then written to Redis when it performs at least as well.
• Operational insight: Includes health, metrics, and implementation gotchas such as missing scored events, bootstrap promotion behavior, and process-level metric limitations.

Quick Start

Use the retraining-worker skill to explain how this service receives retraining triggers, trains a challenger model, shadow tests it, and promotes it when it wins.