Anomaly Detection: Does it identify only spikes?


I have an index in which I log all calls to my system.
The index has a historical depth of 15 days and looking at the graph of calls (count(fn) over time), I can see that there is regularity
I need that when there is a drop in calls, an alert notifies me.

I have defined an anomaly detector that counts the number of calls but I see that the tool only detects me peaks of calls, not decrements:

Can someone help me to understand if the algorithm could be able to detect the decrement of call ?
Do I have to modify something in the detector settings or model configuration?