Error message "Internal server error, Please contact MSFT" with Anomaly Detection using model='azureai'

[aferdoash]

Description

Unhelpful message when running anomaly detection with exogenous variables (future values known), with Azure AI endpoint

There's nothing useful in the traceback too Same script & data works fine with public API (model='timegpt-1')

Use case

anomalies_df = nixtla_client.detect_anomalies( df=train_df_complete[['unique_id','ds','y','3M','6M']], time_col='ds', target_col='y', model='azureai' )

[aferdoash]

After comparing API response with different versions of the input data and also input data itself, it became apparent that the azureai version of the model expects no missing values in the exogenous variables too. Can you make this distinction between TimeGPT & TimeGen-1 clear on your documentation page?