@actualhuman_012 

That message is the generic model-serving “bad request” wrapper we return when a predict operation fails or the request doesn’t match the model’s expected input. The quickest way forward is to validate your payload against the model signature and inspect the endpoint’s inference logs to pinpoint the failure. 

As you have mentioned that this is an Agent Serving endpoint, if you had served the model via the deploy() function,  inference tables are enabled automatically and store payloads.