By using a vector database, we can bypass the finite context limitation of current LLMs and give the model only the context relevant to the task in hand.