QGIS 3.8 model to Python script, not working in 3.4

Question

I am using QGIS 3.4 but it does not have the option of saving a model to a Python script. So I installed QGIS 3.8, which has that option, loaded my model and saved it as a Python script. I loaded the script into the Python editor, clicked "run" and expected the script to work in Version 3.8, but it didn't. The Python console just gives back

exec(open('P:/.../QGIS_Import.py'.encode('utf-8')).read())

With loading it as a script it worked in QGIS 3.8: The model and the python script do not generate the same results, however, that is not the real problem. When I run the python script in QGIS 3.4 it gives this error:

Traceback (most recent call last):
File "C:\Users\heubacha\AppData\Roaming\QGIS\QGIS3\profiles\default\processing\scripts\__QGIS_3.8_Import_Original.py", line 73, in processAlgorithm
'OUTPUT': QgsProcessing.TEMPORARY_OUTPUT
AttributeError: type object 'QgsProcessing' has no attribute 'TEMPORARY_OUTPUT'

The code might not be downwards compatible...what is the correct coding in QGIS 3.4? This is the code snippet where the error occurs:

alg_params = {
            'DISCARD_NONMATCHING': True,
            'FIELD': 'STP',
            'FIELDS_TO_COPY': None,
            'FIELD_2': 'OZ2',
            'INPUT': outputs['FbRefactorFields']['OUTPUT'],
            'INPUT_2': parameters['formstp'],
            'METHOD': 0,
            'PREFIX': '',
            'OUTPUT': QgsProcessing.TEMPORARY_OUTPUT
        }
        outputs['FbfsJoinAttributesByFieldValue'] = processing.run('native:joinattributestable', alg_params, context=context, feedback=feedback, is_child_algorithm=True)

        feedback.setCurrentStep(50)
        if feedback.isCanceled():
            return {}

Answer 1

The Export to script tool (from a model) does not create a "runnable" script. Instead, it creates a script that can be imported as a tool into Processing. At first sight, it seems quite useless, after all, you get the exact same tool, but since it's now in python, you can tweak it directly on the conde. For example, you can enter conditions and loops, something you can do in model builder.