Prompt Engineering for LLMs

Chapter 6. Assembling the Prompt

In the previous chapters, you gathered a wealth of content that will serve as the building blocks for your prompt. Now, it’s time to put these pieces together and craft a prompt that effectively communicates your needs. This chapter will guide you through the process of shaping your prompt by first exploring the different structures and options available to you. How you choose to organize these individual snippets will play a crucial role in the effectiveness of your final prompt.

The next step involves triaging your content—deciding what to keep and what to discard so that it will fit within any size constraints you might have. This process is key to refining your prompt and ensuring it remains focused and relevant. With your content finalized, you’ll then move on to assembling your prompt, which will be your tool for eliciting relevant, coherent, and contextually accurate responses from the model. Let’s dive in.

Anatomy of the Ideal Prompt

Before we go into the details of how to get there, let’s visualize where we want to go. Take a look at Figure 6-1, which gives a bird’s eye view of how your prompt should look. We’ll go through its elements one at a time. Concise and crisp prompts are generally more effective—plus, they use less computational power and are processed more quickly. Additionally, you have a hard cut-off with the context window size.

As discussed in Chapter 5, a prompt consists of elements drawn from dynamic context and static ...