From the course: Semantic Kernel in Action: Fundamentals
Join today to access over 24,200 courses taught by industry experts.
RAG types
From the course: Semantic Kernel in Action: Fundamentals
RAG types
- [Instructor] In the realm of AI, how we store and retrieve memories can make or break the intelligence of a system. Today, we are delving into templated and dynamic RAG, two ways of applying RAG, retrieval-augmented generation. Templated RAG operates on a principle similar to fill-in-the-blanks. Basically, if you know what information you need before asking the AI model, then what make the model look for this information? Isn't it better to retrieve this information beforehand and provide it in the prompt, in a templated prompt? This approach shines in the scenarios requiring high precision and structure, like generating records or answering fact-based queries. For example, what are the pending topics with my customer? When is the next call with him? The major benefit, performance, as well as position and consistency in the responses. First, you are delivering key data to the model in the prompt, so the model does not have to ask for this information and also wait for it and resume…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.