Skip to content

← All series & guides

Series

LlamaIndex from scratch

This path is for you if you want a library that turns files into something you can query. It is LlamaIndex: connectors, pieces of documents, an index, and a query step. No part is live on this page yet. Start with Practical AI for the plain retrieval lesson, then come back for the library.

  1. 1 What Is LlamaIndex? The Data Framework for Connecting LLMs to Your Files Use LlamaIndex when you need connectors, nodes, indexes, and cited answers over private files. Do not confuse the chat UI with the retrieval pipeline underneath. Scheduled · December 2, 2026
  2. 2 Installing LlamaIndex and Building Your First Ask My Folder Script Build an Ask My Folder script with llama-index-core, file readers, and persisted storage. Cache embeddings. Print source scores so weak matches do not look like proof. Scheduled · December 3, 2026
  3. 3 Documents, Nodes, Indexes, and Query Engines: The LlamaIndex Hierarchy Demystified Learn the Document, Node, Index, and Query Engine layers. Prefer sentence-aware nodes and the right index type over blind character slicing. Scheduled · December 4, 2026
  4. 4 Connecting Local Models (Ollama) vs Hosted APIs in LlamaIndex Point LlamaIndex at Ollama for air-gapped RAG, or at hosted APIs for frontier models. Match the provider to PHI rules, latency, and cost. Watch context size and embedding dimensions. Scheduled · December 5, 2026
  5. 5 Building a Production Pipeline for Complex PDFs, Tables, and Notes with LlamaParse Route hard PDFs through LlamaParse into Markdown, then chunk on headers. Keep local readers for plain notes. Intact tables beat OCR soup. Scheduled · December 6, 2026
  6. 6 LlamaIndex Agents and Query Tools Light: Going Beyond Basic RAG Use QueryEngineTool and FunctionTool with a ReAct agent for multi-hop work. Cap iterations, write clear tool descriptions, and watch token cost per loop. Scheduled · December 7, 2026
  7. 7 When Plain Python Scripts or Turnkey RAG UIs Are Better Than LlamaIndex Match tool weight to corpus size, churn, security, and query complexity. Skip framework overhead when a script or product UI already solves the job. Scheduled · December 8, 2026