Closed-Loop Fine-Tuning using Phoenix Metrics • How can we trigger model retraining or fine-tuning workflows based on specific performance thresholds flagged by Phoenix(e.g., LLM hallucination rate)? • Can Phoenix push alerts to a CI/CD or retraining pipeline (e.g., via webhook or integration with Airflow/SageMaker/etc.)