input.json

{
  "page": 1,
  "pages": 2270,
  "limit": 10,
  "total": 22693,
  "items": [
    {
      "address": {
        "city": "cityname first dataset",
        "company_name": "companyname first dataset"
      },
      "amount": 998,
      "items": [
        {
          "description": "first part of first dataset",
          "number": "part number of first part of first dataset"
        }
      ],
      "number": "number of first dataset",
      "service_date": {
        "type": "DEFAULT",
        "date": "2015-11-18"
      },
      "vat_option": null
    },
    {
      "address": {
        "city": "cityname second dataset",
        "company_name": "companyname second dataset"
      },
      "amount": 998,
      "items": [
        {
          "description": "first part of second dataset",
          "number": "part number of first part of second dataset"
        },
        {
          "description": "second part of second dataset",
          "number": "part number of second part of second dataset"
        }
      ],
      "number": "number of second dataset",
      "service_date": {
        "type": "DEFAULT",
        "date": "2015-11-18"
      },
      "vat_option": null
    }
  ]
}

output.csv

StackOverflow 46724816.ipynb

Whoooho... think I got it: http://pandas.pydata.org/pandas-docs/stable/io.html#normalization
Now adapting it to my data set... keep you posted 💃

Okay, so far so good - in general, it is working, currently I have a slight problem with a "distinguishing prefix" - i think because I am in conflict with 'numbers' - seems to be used in pandas itself?
json_normalize is a great way to "massage" the json data - great tip using pandas!

https://stackoverflow.com/questions/46731658/json-normalize-delivers-valueerror-need-of-distinguishing-prefix

Now also with working code example... :D