Skip to content

Vivid Labs AI

Now Watching

RAG Crash Course

RAG Crash Course /What Is RAG?

What Is RAG?

Understand how retrieval-augmented generation grounds an LLM with relevant external context before it produces an answer.

Queue

12 videos in this course

RAG Crash Course

Jump across the full chapter sequence without leaving the player.

01

What Is RAG?

Understand how retrieval-augmented generation grounds an LLM with relevant external context before it produces an answer.

02

House Rules

Set the boundaries of a retrieval system: what belongs in the source collection, what the model may infer, and where grounding must win.

03

Search by Meaning

Move beyond exact keyword matching and retrieve passages whose meaning aligns with the user's question.

04

Rebuild the Diary

Trace how a retrieval pipeline reconstructs the right context from a larger source collection instead of sending everything to the model.

05

Meaning as Numbers

Learn how embeddings encode semantic meaning as vectors that a retrieval system can store, compare, and search.

06

The Ruler

Compare vectors with similarity and distance measures, and see how the chosen metric changes what retrieval considers relevant.

07

Semantic Search

Combine embeddings, a vector index, and a query to retrieve the passages that best match intent rather than surface wording.

08

Write It Down

Index source text with its embeddings and metadata so the system can retrieve both meaning and the evidence needed to use it.

09

Cut to Size

Choose chunk boundaries and sizes that preserve meaning while keeping retrieval precise and model context manageable.

10

Practice Cuts

Compare practical chunking strategies and learn how to test whether each split preserves the context a real query will need.

11

The Whole Table

Assemble retrieved chunks into coherent context, preserve their source relationships, and hand the model evidence it can use together.

12

The Dinner Rush

Bring the pipeline together and design retrieval that remains responsive and reliable when query volume and context pressure increase.