In this chapter, learn how to choose between RAG, hybrid search, and text-to-SQL for your data and sketch the architecture of a conversational app on Snowflake.

Getting answers from data

Getting answers from data: Using RAG

Getting answers from data: Using text-to-SQL

What we will build: Success in this course

Reference documentation on COMPLETE

Querying unstructured data: RAG

Querying Unstructured Data: Hybrid Search

Querying structured data: Text-to-SQL

Setting up your environment

Let's dive in

Chapter 1 review - Question 1

Chapter 1 review - Question 2

Chapter 1 review - Question 3

Chapter 1 review - Question 4

Chapter 1 review - Question 5

Introduction to building conversational apps with data

In this chapter, learn how to build a RAG app with Cortex Search by loading, parsing, and chunking documents, measuring answer quality, and shipping it with Streamlit.

Opening up unstructured data with RAG

FOMC meeting minutes

What is Cortex Search

Reference documentation on Arctic Embed 2.0

Permissions for AI Observability

Loading unstructured data to a Snowflake stage

Parsing and Chunking Text

Documentation on Parsing and Splitting

Creating the Cortex Search service

Building a RAG with Cortex Search

How to measure success

Why should you trust LLM-as-a-judge

Automatic Processing of New Documents

Tasks and Streams

Building out the front-end

Well done!

Chapter 2 review - Question 1

Chapter 2 review - Question 2

Chapter 2 review - Question 3

Chapter 2 review - Question 4

Chapter 2 review - Question 5

Finding answers from unstructured data with Cortex Search

In this final chapter, learn how to turn structured tables into a natural-language interface with Cortex Analyst, using a semantic model to generate accurate SQL and serve answers through a Streamlit app.

Avoiding death by dashboard

The code you need for this module

Setting up the App

Understanding the Semantic Model

Working with the Semantic Model

Using Cortex Analyst

Under the hood of Cortex Analyst

From text-to-SQL to TAG - Creating Table Assisted Generation

LLM context window token limits

Expanding the Scope of the Semantic Model

Using Cortex Search to enhance Cortex Analyst

Building the Streamlit app

Congratulations!

Chapter 3 review - Question 1

Chapter 3 review - Question 2

Chapter 3 review - Question 3

Chapter 3 review - Question 4

Chapter 3 review - Question 5

Answer questions from structured data with Cortex Analyst

Most of your organization's knowledge is locked inside tables, PDFs, and documents that only technical users can query. What if anyone could just ask? In this course, you'll build generative AI apps on Snowflake that let people talk to their data in plain language. You'll learn when to reach for RAG, hybrid search, or text-to-SQL, then build working chatbots with Cortex Search for documents and Cortex Analyst for tables, each with a Streamlit front end. By the end, you'll turn raw data into answers the whole business can use.

Generative AI Essentials with Snowflake

Build conversational AI apps that answer questions from your data with Cortex Search and Cortex Analyst on Snowflake.


Building Generative AI Apps to Talk to Your Data

Why should you trust LLM-as-a-judge

Building Generative AI Apps to Talk to Your Data

Interaktywne ćwiczenie praktyczne