Reinforcement learning - how is it done?

Question

Neural networks

Learning with a teacher is understandable. Using Keras, we create a neural network model, set two arrays with input data and correct answers. Then we just need to press the start button and wait for the result. Those. we set the parameters and the whole process goes automatically.
In my case, there is a stream of incoming data, the function calculates the correct answer depending on the incoming data and time. As I understand it, reinforcement training is just what is needed here. Those. we need a network training "module" that will compare the NN answer and the correct one (calculated by the function), and adjust the weights. And continue the cycle of learning further. Those. you need to somehow write with pens the part responsible for the learning process itself, which is performed "automatically" in training with a teacher.
Are there any standard solutions for this in Keras? Or how else can it be solved?
If there are any links to a similar topic in Russian, I will be very glad to see them.