Enroll in "Building AI Applications with Haystack," our new short course!

I'm delighted to introduce Building AI Applications with Haystack. Built in partnership with deepset, the creator of Haystack and taught by Tuana Jellick, Haystack is a framework that simplifies the process of creating LLM based applications. One of the examples you see in this course is creating a news summarization app. Building this from scratch without using a framework is possible, but own technologies and libraries are evolving quickly and user needs can also change. Having a higher level orchestration framework can help you make sure that your application is flexible, extendable and maintainable. Even as the tech stack changes, new user needs arise and you add new features. Frameworks provide abstractions that make your app development process more modular and extensible. You see all of this by creating several apps in this course, including including a chat agent with function calling and a self reflecting agent with loops. I'm delighted to introduce the instructor for this course, Taner Schellig, who is Developer Relations Lead for Haystack at Deepset. Tanner has been helping many developers create AI applications using Haystack.

Thanks Andrew. Haystack offers a robust and flexible architecture for building complex AI applications and in this course you'll learn key concepts that are needed for building these apps. Components are the building blocks of Haystack. Ready made components include retrievers, generators, rankers, various pre processors and more. Pipelines. Combine these components to implement one of many AI applications like rag, chat, document search and so on. Document stores are where you store your data and retrieve them from as your app grows. There will be cases where the default components in Haystack will not provide all of the features you need. Haystack provides a common interface with which new components can be created. In this course, through building a new Summarizer app, you will learn how to create your own custom components and add them to your pipeline. You will create a custom component that fetches the top news from Hacker News which will be passed to the LLM in your pipeline to summarize.

So in building LLM apps there are many use cases where you want to create a conditional branch or an if statement in your pipeline. Say you want to build a rag app. Sometimes retrieved context may not contain all the information needed to answer the user question. In this case, maybe you want your app to do a web search and use those results as a context for the LLM. You're going to learn to do this by implementing a fallback to web search mechanism in a rag app. Or say you're extracting named entities from a document in a specific format of person, location, date. But at times maybe the LLM output is in an incorrect format. You learn to use a self reflecting pipeline with a looping mechanism where the LLM is fed back to the LLM for it to check and reflect on how to improve it. You can then iteratively prompt the LLM to improve its response. Finally, you also learn to use haystack with tools or function calling. One cool thing you learn is that a haystack pipeline can itself become a tool for a higher level agent. So with these abstractions you can build some very sophisticated workflows. I hope you sign up for and enjoy the course.