Learning LangChain

Chapter 1. LLM Fundamentals with LangChain

The Preface gave you a taste of the power of LLM prompting, where we saw firsthand the impact that different prompting techniques can have on what you get out of LLMs, especially when judiciously combined. The challenge in building good LLM applications is, in fact, in how to effectively construct the prompt sent to the model and process the model’s prediction to return an accurate output (see Figure 1-1).

Figure 1-1. The challenge in making LLMs a useful part of your application

If you can solve this problem, you are well on your way to building LLM applications, simple and complex alike. In this chapter, you’ll learn more about how LangChain’s building blocks map to LLM concepts and how, when combined effectively, they enable you to build LLM applications. But first, the sidebar “Why LangChain?” is a brief primer on why we think it useful to use LangChain to build LLM applications.

Why LangChain?

You can of course build LLM applications without LangChain. The most obvious alternative is to use the software development kit (SDK)—the package exposing the methods of their HTTP API as functions in the programming language of your choice—of the LLM provider you tried first (for example, OpenAI). We think learning LangChain will pay off in the short term and over the long run because of the following factors:

Prebuilt common patterns