For context stuffing LLMs with small token limits(AKA Retriever Augmentation), you will need to break up each article in few-sentence chunks. You can get to a large number of these chunks very rapidly.