Topolearn

Learn LlamaIndex

Data framework for LLM apps: RAG over your documents, agents and workflows

26 lessons · 5 hours · 4 levels · covers LlamaIndex 0.14 (llama-index-core 0.14.x, Workflows 2.x, Oct 2026)

LlamaIndex is a Python framework for connecting LLMs to your own data. It loads documents, splits and embeds them into indexes, and answers questions through query engines, chat engines and retrievers; on top of that it offers tool-using agents and event-driven Workflows. It integrates with hundreds of model providers and vector databases, and with LlamaParse for document parsing.

Sign in free to track progressOfficial docs ↗

Before you start

LlamaIndex builds on these; the tree links to the exact lessons you need.

PythonPractical levelAI & LLM fundamentalsPractical level

How deep do you want to go?

1. EssentialsRead and write simple code, do the everyday basics9 lessons · 2 hours in total
2. PracticalUse it productively at work, with the usual idioms18 lessons · 3 hours in total
3. AdvancedThe less common features, performance and trade-offs23 lessons · 4 hours in total
4. ExpertInternals and the edge cases experts know26 lessons · 5 hours in total

The LlamaIndex tree

LlamaIndexWhat LlamaIndex isWhat LlamaIndex isYour first RAG app in five linesYour first RAG app infive linesSettings: LLMs and embedding modelsSettings: LLMs andembedding modelsLoading data with readersLoading data withreadersDocuments, nodes and metadataDocuments, nodesand metadataChunking with node parsersChunking with nodeparsersVectorStoreIndexVectorStoreIndexQuery engines and responsesQuery engines andresponsesSaving and loading indexesSaving and loadingindexesChat enginesChat enginesRetrievers and custom query enginesRetrievers andcustom query engi…Metadata filtersMetadata filtersVector store integrationsVector storeintegrationsResponse synthesis and promptsResponse synthesisand promptsIngestion pipelinesIngestion pipelinesStructured outputsStructured outputsAgents with FunctionAgentAgents withFunctionAgentParsing complex documents with LlamaParseParsing complexdocuments with Lla…Retrieval-augmented generation (RAG)Retrieval-augmentedgeneration (RAG)Virtual environments and pipVirtual environmentsand pipEmbeddings and semantic similarityEmbeddings andsemantic similarityReading and writing filesReading and writingfilesDictionariesDictionariesPrompts, messages and rolesPrompts, messagesand rolesWriting clear promptsWriting clear promptsStructured output (JSON)Structured output(JSON)Type hints basicsType hints basicsTool calling (function calling)Tool calling (functioncalling)What an AI agent isWhat an AI agent is
Essentials and Practical shown (18 of 26 lessons); the full list is below. Click a lesson to read it.

All 26 lessons, in order

Getting started

  1. What LlamaIndex is · Essentials · 8 min

    Understand what LlamaIndex does, its RAG pipeline, and how the core and integration packages fit together.

  2. Your first RAG app in five lines · Essentials · 10 min

    Load a folder, build a VectorStoreIndex, and ask questions with a query engine.

  3. Settings: LLMs and embedding models · Essentials · 10 min

    Choose the LLM and embedding model globally with Settings or per component, including local models via Ollama.

Loading and indexing data

  1. Loading data with readers · Essentials · 10 min

    Load files with SimpleDirectoryReader options and other sources with LlamaHub readers.

  2. Documents, nodes and metadata · Essentials · 10 min

    Work with Document and TextNode objects, attach metadata, and control what metadata the LLM and embeddings see.

  3. Chunking with node parsers · Essentials · 12 min

    Split documents into nodes with SentenceSplitter and structure-aware parsers, and choose chunk size and overlap.

  4. VectorStoreIndex · Essentials · 10 min

    Build a VectorStoreIndex from documents or nodes, add and delete content, and understand what it stores.

  5. Saving and loading indexes · Essentials · 8 min

    Persist an in-memory index to disk with storage_context.persist and load it back with load_index_from_storage.

  6. Vector store integrations · Practical · 12 min

    Store embeddings in a real vector database such as Chroma via StorageContext, and reconnect with from_vector_store.

  7. Ingestion pipelines · Practical · 12 min

    Run documents through an IngestionPipeline of transformations with caching and de-duplication into a vector store.

  8. Parsing complex documents with LlamaParse · Practical · 10 min

    Know when to use LlamaIndex's hosted LlamaParse platform for PDFs with tables and layouts, and how to call it.

Querying and chatting

  1. Query engines and responses · Essentials · 12 min

    Configure as_query_engine (top-k, response mode, streaming) and read answers, sources and metadata from the Response.

  2. Chat engines · Practical · 10 min

    Hold multi-turn conversations over your data with as_chat_engine and the condense_plus_context mode.

  3. Retrievers and custom query engines · Practical · 12 min

    Use retrievers directly, inspect NodeWithScore results, and assemble a RetrieverQueryEngine with postprocessors.

  4. Metadata filters · Practical · 10 min

    Restrict retrieval to nodes whose metadata matches conditions, using MetadataFilters and FilterOperator.

  5. Response synthesis and prompts · Practical · 12 min

    Choose response modes such as compact, refine and tree_summarize, and customise the QA prompt.

  6. Structured outputs · Practical · 10 min

    Extract typed Pydantic objects with structured_predict, as_structured_llm and query engines with output_cls.

Agents and workflows

  1. Agents with FunctionAgent · Practical · 14 min

    Build a tool-using FunctionAgent from Python functions and query engine tools, and run it with async code.

  2. Agent memory, state and streaming · Advanced · 14 min

    Keep conversations with Context, store state, serialise it between requests, and stream agent events.

  3. Workflows: steps and events · Advanced · 15 min

    Write event-driven Workflows with @step methods, custom Events, StartEvent/StopEvent and shared Context state.

  4. Multi-agent systems with AgentWorkflow · Advanced · 14 min

    Coordinate several specialised agents with AgentWorkflow handoffs, shared state, or an orchestrator agent.

Better retrieval

  1. Reranking and hybrid search · Advanced · 14 min

    Improve retrieval precision with rerankers and recall with hybrid BM25 + vector search fused by QueryFusionRetriever.

  2. Advanced retrieval patterns · Expert · 15 min

    Apply sentence-window and auto-merging retrieval, query routing and sub-question decomposition to hard RAG problems.

Evaluation and production

  1. Evaluating RAG · Advanced · 14 min

    Measure retrieval with hit rate and MRR, and answers with faithfulness and relevancy evaluators, over a test set.

  2. Observability and instrumentation · Expert · 12 min

    Trace every retrieval, LLM call and tool call with the instrumentation module, global handlers and tracing integrations.

  3. Running LlamaIndex in production · Expert · 14 min

    Design for scale, freshness, cost, latency and security: async ingestion, caching, permissions, and serving workflows.