Currently, only applicaton/json is supported as content-type for deserialization in the preprocess step. In case the content-type is absent or is not equal to application/json the body is passed as a binary to the callback function.
A nice solution would be to allow custom parsers (as MFAs or HOFs) per content-type in erf's configuration.