First, I'm quite new to Keras/Tenserflow.

But I want to build a Bidirectional(LSTM) network.

Now I understand that I learn/train the network with:

model.fit(X_train, Y_train,........ with my dataset

after I can test the model with model.predict(X_test)and get my prediction.

As I have a live feed of data I would like to constantly update also the RNN that it learns also new behaviour. Is this possible? Can a Keras LSTM RNN learn (fit) and predict continuously with new data? Or do I need to run every time when I get a new row of data the model.fit and build a new net with the additional dataset?