Are you looking for the frequency to be daily or minutely. That is, should the prices be over 5 and 34 days OR 5 and 34 minutes?

Also, if daily, should the current days prices be included in the averages (this may make it more complicated if so).

I should have specified, sorry. I'd like the frequency to be in minutes. Thanks for your help!

Maybe something like this? Also see the attached notebook.

    # Add 1 to the LONG_SMA to get the current and previous bar values  
    my_data = data.history(context.security, ['high', 'low'], LONG_SMA+1, '1m')  
    high_low_mean_short = (my_data.high.rolling(SHORT_SMA).mean() -  
                           my_data.low.rolling(SHORT_SMA).mean())/2  
    high_low_mean_long = (my_data.high.rolling(LONG_SMA).mean() -  
                          my_data.low.rolling(LONG_SMA).mean())/2  
    ao = high_low_mean_short - high_low_mean_long  

Here is the code in an algorithm...

Here is another version of Bill Williams Awesome Oscillator utilizing talib.MEDPRICE function

import talib  
# -------------------  
stock = symbol('SPY')  
fast, slow = 5, 34  
# -------------------  
def initialize(context):  
    pass

def handle_data(context, data):

    H = data.history(stock, 'high', slow , '1m').bfill()  
    L = data.history(stock, 'low', slow , '1m').bfill()  

    median_prices = talib.MEDPRICE(H, L)  
    mavg_fast = median_prices[-fast:].mean()  
    mavg_slow = median_prices[-slow:].mean()

    AO = mavg_fast - mavg_slow # Awesome_Oscillator  
    print AO  
    record(Awesome_Oscillator = AO)