If i want to monitor model's performance through model metrics (i.e. XGboost performance with RMSE) in production setting, would I need to use monitoring funcion in bentoML? Or is there more standardized (or effective) ways to monitor model performance in BentoML service? https://docs.bentoml.com/en/latest/build-with-bentoml/observability/monitoring-and-data-collection.html#ship-the-collected-data

how would you like to observe your metrics?

Ideally, I'd like to display it on grafana dashboard and alert when metrics fall below threshold

my suggestion is to create your custom metrics with promethues client, https://docs.bentoml.com/en/latest/build-with-bentoml/observability/metrics.html this will expose it in the

/metrics

endpoint and can be scrape by a prometheus agent and forwarded to grafana for visualization as well as alert