SkillByAIOpen interactive version →

LangChain / LlamaIndex

Build LLM applications with LangChain and LlamaIndex: prompts, runnables, parsers, tools, retrieval, indexes, query engines, agents and testing, with examples run offline using fake models.

Start course →

Syllabus

The Framework Landscape

  1. Why Use an LLM Framework
  2. LangChain, LlamaIndex and Their Packages
  3. LangChain vs LlamaIndex: Which to Use When
  4. A Mental Model: Everything Is a Pipeline of Typed Steps

LangChain Core: Prompts, Models and Chains

  1. Prompt Templates and Messages
  2. Chat Models and the Pipe (LCEL) Chain
  3. Runnables: Lambda, Parallel, Passthrough and Batch
  4. Output Parsers and Structured Output
  5. Streaming and Async

LangChain: Tools, Routing and Reliability

  1. Defining Tools
  2. Routing and Branching
  3. Retries, Fallbacks and Timeouts
  4. Callbacks, Tracing and Debugging
  5. Chat History and Memory

Retrieval with LangChain

  1. Documents, Loaders and Text Splitters
  2. Embeddings, Vector Stores and Retrievers
  3. Building a RAG Chain
  4. Making Retrieval Better: Filters, MMR, Reranking

LlamaIndex: Data, Indexes and Query Engines

  1. LlamaIndex Data Model: Documents and Nodes
  2. Building a Vector Index and Retriever
  3. Query Engines and Response Synthesis
  4. Persisting and Reloading an Index

Agents and Workflows

  1. Agents: The Tool-Calling Loop
  2. LangGraph: Stateful Workflows
  3. Guardrails for Agentic Apps

Testing, Evaluating and Shipping

  1. Unit Testing with Fake Models
  2. Evaluating Quality: Datasets, Metrics, Tracing
  3. Versions, Cost, Latency and Security
  4. When to Drop the Framework

Putting It Together

  1. Case Study: A Documentation Assistant
  2. Revision: Cheat Sheet and Self-Check