I am trying to replace a value in a pandas dataframe cell that I am constructing from a JSON file that I get from Airtable API the JSON file looks like this one:

[
    {
        "id": "rec0Y7ydscZnU5NjZ",
        "createdTime": "2023-03-17T16:32:57.000Z",
        "fields": {
            "field_id": [
                "recq6vsFWI2H8j84i"
            ],
            "field_name": "S-166",
            "currency": [
                "Dolares"
            ],
            "state": [
                5
            ],
            "roi": {
                "specialValue": "Infinity"
            },
            "investment_return": {
                "specialValue": "NaN"
            },
            "value_amount": 16559.51178956216,
            "income": [
                31978.5
            ]
        }       
    }
]

Obviously this is a sample and I am replacing the real keys and some values, because they are from a private application, but what I am trying to do once I get the API response and then convert the JSON into a dataframe using the code below, is to replace the values {"specialValue": "Infinity"} from roi key and {"specialValue": "NaN"} from investment_return key with zeros:

#read_json.py
def read_json():
    # Open the JSON file
    with open('json_test.json', 'r') as file:
    # Parse the JSON data
        data = json.load(file)

        # Extract all fields from the "fields" dictionary and create a DataFrame
        df1 = pd.DataFrame([record['id'] for record in data],columns=['id'])
        df2 = pd.DataFrame([record['fields'] for record in data])
        
        df = pd.concat([df1,df2], axis=1)
        
    return df

Then once I have the JSON data converted into a dataframe I am trying to replace the values using this code:

df.replace('{"specialValue": "Infinity"}', np.nan, inplace=True)

For some reason is not working:

Here I am using the read_json() function I described above to generate the dataframe from the json file gotten from API response, and inspecting the resulting dataframe:

df = read_json()
print(df)

                  id             field_id field_name   currency state  
0  rec0Y7ydscZnU5NjZ  [recq6vsFWI2H8j84i]      S-166  [Dolares]   [5]   

                            roi        investment_return  value_amount  
0  {'specialValue': 'Infinity'}  {'specialValue': 'NaN'}   16559.51179   

      income  
0  [31978.5] 

Then once I want to replace() for example the string {"specialValue": "Infinity"} by NaN with numpy:

# replace string value with NaN
df.replace('{"specialValue": "Infinity"}', np.nan, inplace=True)
print(df)

what I would expect is to get for the column roi an nan value, but I am still getting the same string that I receive from thr API:

0  rec0Y7ydscZnU5NjZ  [recq6vsFWI2H8j84i]      S-166  [Dolares]   [5]   

                            roi        investment_return  value_amount  
0  {'specialValue': 'Infinity'}  {'specialValue': 'NaN'}   16559.51179   

      income  
0  [31978.5]  

Can somebody give me a possible approach on how to convert those into nan or zeros?