LlamaIndex

What it is

A framework built for document-heavy retrieval.

LlamaIndex is purpose-built for RAG: 300+ data connectors pull from PDFs, Notion, Slack, databases and more; indexing and chunking are handled for you; and query engines, retrievers, and agentic retrieval sit on top so you can ask questions over your corpus without wiring every stage by hand. Where a general orchestration framework treats retrieval as one node among many, LlamaIndex treats the whole document-to-answer path as the main event.

The one job

Documents to a working retrieval pipeline, fast.

Point it at a pile of documents and get back something you can query. It collapses connect, chunk, index, and retrieve into a few lines so you spend your time on quality, not plumbing.

Reach for it when

• You have a document-heavy RAG problem, not a general agent problem.
• You need to pull from many data sources and want connectors off the shelf.
• You want query engines and retrievers out of the box rather than hand-built stages.
• You need agentic retrieval: an agent that decides what and how to fetch.

Skip it when

• You have only a few documents: roll your own with pgvector plus embeddings, which Lab 2 does to teach the mechanics.
• You want one orchestration framework and already run LangGraph across the app.
• Your corpus is tiny and static: an index and query engine are overhead you do not yet need.

A minimal look

Build an index, then query it.

import { VectorStoreIndex, Document } from "llamaindex";

// wrap your documents; chunking + embedding happen for you
const docs = files.map((f) => new Document({ text: f.text }));
const index = await VectorStoreIndex.fromDocuments(docs);

// a query engine retrieves the right chunks and answers
const engine = index.asQueryEngine();
const res = await engine.query({ query: "What is our refund window?" });

The principle it teaches

Where it fits your labs

This is a Week 2 tool. Lab 2 has you build the hybrid RAG pipeline by hand first; LlamaIndex is what you graduate to once you know exactly what each stage is doing and why.