On my end, I don't find the conversion to datetime mentioned by @piRSquared to be necessary. You can just do:
df[<column_name>] = df[<column_name>].astype(str) df.to_dict('records')