What does a data-processing pipeline mean?

A pipeline is a sequence of stages in which the output of one becomes the input of the next. In a simple example, I first fetch a page, then extract data, and finally classify and save it. Stages can be ordinary code, a model or a combination of both.

Splitting it up helps you see where an error happened and which stages need repeating. It's important to agree the data format between them. If the first stage loses important information, the next one won't magically recover it. A pipeline doesn't have to mean a queue or many separate servers — sometimes a few well-separated functions are enough.