You can do what you want with the API. MERL is for turning models into text, so read-only; the API is for manipulating models, so includes create, read, write and update operations. You can of course combine the two: MERL can call a program you have written that uses the API, or you can use the API to call MERL generators.

 

For the dialog, my recommendation would be to collect all the input fields that you want to enable for manual values into their own object type. You can then have in your main object a property that is initially empty but can contain an object of that type. When writing a code generator or generators for the object symbol you can check if the object is there, and if so use its manual values; otherwise, your generator would produce the automatic values.