During this training day we will build a RAG solution using a bit more than the obvious components.

Instead of using only the LLM over the data, which provide unpredicted results, we will use a co-pilot with co-pilot studio and power automation, acquiring additional control over the result.

Instead of indexing relational data using cognitive services, we will build a query system where the LLM is capable to build a query for relational data.

Finally, to join the solution for structured and unstructured data in a single solution, we will using Azure Machine Learning Prompt Flow, illustrating a glimpse of the powerful capability to build a workflow over an LLM solution.

Let's break down what's fashion and what's reality and you will decide how to better use this architecture.