Rolling standard deviation excluding current rows in python

Question

I'm trying calculate the rolling standard deviation for a groupby object which contains Invoice Date, Vendor Name and Agency Name, while excluding the current row in the calculation. So the rolling std() for each Vendor-Agency pair basically. It’s my first time asking a question here, so please let me know if additional information is required.

I already managed to do something similar with the rolling mean with the codes below.

last_3_invoices = dft_gb_comb.set_index("INVOICE_DATE").groupby(["VENDOR_NAME","OCP_AGNCY_NAME"])["daily_avg_amount"].rolling(3,min_periods = 1, closed="left").mean().fillna(0).reset_index()
last_30_df= dft_gb_comb.set_index("INVOICE_DATE").groupby(["VENDOR_NAME","OCP_AGNCY_NAME"])["daily_avg_amount"].rolling("30d",closed="left").mean().fillna(0).reset_index()

The problem is, if I try to replicate this code with .std() instead of .mean() I get no values. If i omit the "closed ="left"" statement I get values, but those include the current row, so that wouldn't work for me. I tried several ways to work with the .shift() statement instead

This could would work for some cases but it does not respect the ["VENDOR_NAME","OCP_AGNCY_NAME"] boundaries.

dft_gb_comb.set_index("INVOICE_DATE").groupby(["VENDOR_NAME","OCP_AGNCY_NAME"])["daily_avg_amount"].rolling(3,min_periods=1).std().shift().reset_index()

If I use shift() directly after the groupby like below it only returns the Invoice date with the values while dropping ["VENDOR_NAME","OCP_AGNCY_NAME"] from the groupby object...

dft_gb_comb.set_index("INVOICE_DATE").groupby(["VENDOR_NAME","OCP_AGNCY_NAME"])["daily_avg_amount"].shift().rolling(3,min_periods=1).std().reset_index()

dft_gb_comb.set_index("INVOICE_DATE").groupby(["VENDOR_NAME","OCP_AGNCY_NAME"])["daily_avg_amount"].apply(lambda x : x.shift(1).rolling(3,min_periods=1).std()).reset_index()

	INVOICE_DATE	daily_avg_amount
0	2018-04-19	NaN
1	2018-12-10	NaN
408184	2019-06-05	139.02567
408185	2019-06-19	124.92972
408186	2018-05-16	34.92400
408187	2019-04-19	135.41095
408188	2018-08-06	NaN

instead of something like this, which would be the desired end result.

	VENDOR_NAME	OCP_AGNCY_NAME	INVOICE_DATE	daily_avg_amount
0	AED PROFESSIONALS N. E.	OKLA. A & M COLLEGE	2018-04-19	NaN
2	E3 DIAGNOSTIC,INC. DBA E3 GORDON STOWE/	OKLAHOMA STATE UNIVERSITY	2018-01-17	215.66757

I need to join the groupby object with the rolling values back to the main dataframe in the end, so I can't allow "VENDOR_NAME","OCP_AGNCY_NAME" to be dropped.

current python version would be

python : 3.8.8.final.0

pandas : 1.2.4

numpy : 1.20.1

I would be thankful for any kind of suggestions. I'm working with vendor data from here