How to resample intranet time data using dateOffset in Pandas / Numpy?

I am dealing with futures where the current day starts before 00:00:00. I need to do resampling of 1 minute data by 1 hour of data taking into account the date offset. Let's consider an example:

df1 - 1 minute of data

                        Open     High      Low    Close
2005-09-06 17:27:00  1234.75  1234.75  1234.75  1234.75
2005-09-06 17:28:00  1234.75  1234.75  1234.75  1234.75
2005-09-06 17:29:00  1234.75  1234.75  1234.75  1234.75
2005-09-06 17:30:00  1234.75  1234.75  1234.50  1234.50
2005-09-06 18:01:00  1234.50  1234.50  1234.25  1234.50
2005-09-06 18:02:00  1234.50  1234.50  1234.50  1234.50
2005-09-06 18:03:00  1234.50  1234.50  1234.50  1234.50
2005-09-06 18:04:00  1234.50  1234.50  1234.50  1234.50
2005-09-06 18:05:00  1234.50  1234.50  1234.25  1234.25

      

        
        
        
      

    

This is what happens with regular re-rendering:

conversion = {'Open': 'first', 'High': 'max', 'Low': 'min', 'Close': 'last'}

In [77]: normal_resample = df1.resample(rule='60Min', how=conversion)
In [78]: normal_resample
Out[79]: 
                        Open     High      Low    Close
2005-09-06 17:00:00  1234.75  1234.75  1234.50  1234.50
2005-09-06 18:00:00  1234.50  1234.50  1234.25  1234.25

      

        
        
        
      

    

This is the required output for the offset:

conversion = {'Open': 'first', 'High': 'max', 'Low': 'min', 'Close': 'last'}

In [77]: offset_resample = df1.resample(rule='60Min', how=conversion) + offset:18:00
In [78]: offset_resample
Out[79]: 
                        Open     High      Low    Close
2005-09-06 17:00:00  1234.75  1234.75  1234.50  1234.50
2005-09-07 18:00:00  1234.50  1234.50  1234.25  1234.25

      

        
        
        
      

    

Note that I want to indicate that the new day starts at '18: 00: 00 ', not midnight.

What I did: I have a formula that includes offset, resampling for daily data, but I don't know how to adapt it for intraday resampling.

def resample_day(df):
    df.index = pd.DatetimeIndex(df.index)
    df = df.between_time('18:00', '16:00', include_start=True, include_end=True)
    proxy = df.index + pd.DateOffset(hours=6)  # This is the piece that does the trick
    result = df.groupby(proxy.date).agg(
        {'Open': 'first', 'High': 'max', 'Low': 'min', 'Close': 'last'})
    result = result.reindex(columns=['Open', 'High', 'Low', 'Close'])
    return result

      

        
        
        
      

    

Other than what is suggested here doesn't work, at least for me.

Thanks for the input.