RAG

Building Retrieval-Augmented Generation (RAG) systems sounds straightforward in theory. But after working on several real-world RAG use cases—FAQ chatbot builder, image-to-image search, product recommendation engines and voice agents—I learned that the reality is far messier. Rather than covering everything at once, I’m breaking this down into a short series of posts—each one focusing on a specific issue I encountered and how I approached fixing it. In this first part, we’ll look at one of the most frequent problems: follow-up queries that fail retrieval—and how query expansion helped solve it. ...

How Maximal Marginal Relevance (MMR) Enhances Diversity and User Experience in Product Suggestions.

LLM is powerful, but it cannot answer questions it doesn’t know before. Thanks to RAG we can inject some knowledge to extend LLM capability. Now let’s build our first RAG!

We moved from in-house training model to hosted models and ready-to-use APIs. With the existence of free LLM APIs, let’s explore how to create our own free chatbot!