and while not documented fully in DocBook the usage of %NOPARALLEL can be used to tell the optimizer not to use parallel processing.

If we look to what InterSystems provides as an application like Ensemble(Interoperability) or HealthShare we will find the class defintions exist in the ENSLIB or HSLIB namespace.  Then via Package mapping you will find that Interoperability enabled namespaces or HS enabled namespaces map the package definitions.  For example in the namespace HSEDGE1 we see

which says the classes from these packages are visible to the HSEDGE1 namespace.  

The data for the extent of these classes by default would live in the associated db for the namespace HSEDGE1.

In most cases the data for the extent is not mapped so as to live in a single univeral database but generally speaking it could be, although in the case iof the Ensemble/Interoperability and HealthShare data you shoud not do this.

The error you are seeing is a compile error, its not a run time error so the DC AI Bot isnt telling the correct answer.  The field in the RecordMap UI is a bit wonky.  If you hover over the label you get a tooltip which indicates you need to separate the parameter values by ;

So something like this works

MAXLEN=80;PATTERN=14N

which gets stored in the RecordMap definition class as

  <Field name="NPI" required="0" ignored="0" datatype="%String" index="1" params="MAXLEN=80;PATTERN=14N" sqlColumn="2" repeating="0">
  </Field>
and the generated RecordMap class now has

Property NPI As %String(MAXLEN = 80, PATTERN = "14N") [ SqlColumnNumber = 2 ];

so 

1. parameters should be separated by ;

2. for pattern you do not need to surround the pattern value by "s

3. If something like ResearchId is actually a pure numeric and number it might be better defined as %Integer

4. For something like FirstName you might want PATTERN=1.A  which means 1 to N alpha characters