Chapter 2: LlamaIndex Framework

LlamaIndex Overview

LlamaIndex (formerly GPT Index) focuses on data indexing and querying:

• Multiple index types
• Query engines
• Response synthesis
• Evaluation tools

Basic RAG with LlamaIndex

from llama_index import VectorStoreIndex, SimpleDirectoryReader
from llama_index.llms import OpenAI

# Load documents
documents = SimpleDirectoryReader("./data").load_data()

# Create index
index = VectorStoreIndex.from_documents(documents)

# Query
query_engine = index.as_query_engine()
response = query_engine.query("What is RAG?")
print(response)

Index Types

from llama_index import (
    VectorStoreIndex,
    SummaryIndex,
    TreeIndex,
    KeywordTableIndex
)

# Vector Index - similarity search
vector_index = VectorStoreIndex.from_documents(docs)

# Summary Index - summarization
summary_index = SummaryIndex.from_documents(docs)

# Tree Index - hierarchical
tree_index = TreeIndex.from_documents(docs)

# Keyword Index - keyword extraction
keyword_index = KeywordTableIndex.from_documents(docs)

Query Engines

# Basic query engine
query_engine = index.as_query_engine(
    similarity_top_k=5,
    response_mode="compact"  # or "tree_summarize", "refine"
)

# Chat engine with memory
chat_engine = index.as_chat_engine(
    chat_mode="condense_question",
    verbose=True
)

# Use chat engine
response = chat_engine.chat("What is RAG?")
response = chat_engine.chat("Tell me more")  # Remembers context

Response Modes

Mode	Description	Best For
compact	Fit into single prompt	Default
refine	Iterative refinement	Long contexts
tree_summarize	Hierarchical summary	Many docs
simple_summarize	Basic summary	Quick overview

LangChain vs LlamaIndex

Feature	LangChain	LlamaIndex
Focus	General chains	Data indexing
Flexibility	Higher	Medium
Learning Curve	Steeper	Gentler
Best For	Complex pipelines	Document Q&A

---

Next Steps

Continue to Chapter 3: Custom RAG.