Zipline (the Python library that Quantopian is built on) has support for TA-Lib now. Not sure if the support is in Quantopian yet, but if not it will be soon. Here's an example of using TA-Lib in Zipline:

https://github.com/quantopian/zipline/blob/master/zipline/examples/dual_ema_talib.py

Hi Joao and Ben, TA-Lib support just landed in Quantopian: https://www.quantopian.com/posts/new-feature-ta-lib-support

Hope that helps.

Readable code Joao and your RSI vals close to TradingView.
Thx I appreciate the fact that you posted it.

Hi,
I know this is a bit old or not needed (due to TA-Lib) now, but I think the RSI calculation in the code given at the top is not correct. For anybody who's still interested...

The correct answer to the example in the comment should be:
[ 70.02141328 69.57446809 69.37901499 80.26607539 73.39901478 59.90338164 62.61261261 60. 48.47775176 53.87840671
48.95238095 43.86281588 37.6744186 32.20910624 32.66331658
38.0794702 31.70391061 25.066313 30.17902813]

And the modified RSI function to produce above result is:

def rsi(prices, period=14):

num_prices = len(prices)

if num_prices < period:  
    # show error message  
    raise SystemExit('Error: num_prices < period')

# this could be named gains/losses to save time/memory in the future  
changes = prices[1:] - prices[:-1]  
#num_changes = len(changes)

rsi_range = num_prices - period

rsis = np.zeros(rsi_range)

gains = np.array(changes)  
# assign 0 to all negative values  
masked_gains = gains < 0  
gains[masked_gains] = 0

losses = np.array(changes)  
# assign 0 to all positive values  
masked_losses = losses > 0  
losses[masked_losses] = 0

for idx in range(0, rsi_range):  

    avg_gain = np.mean(gains[idx:(idx+period)])  
    avg_loss = np.mean(losses[idx:(idx+period)])

    if avg_loss == 0:  
        rsis[idx] = 100  
    else:  
        rs = avg_gain / (-avg_loss)  
        rsis[idx] = 100 - (100 / (1 + rs))

return rsis