Twelve curated topics on building production-grade LLM agents: LangChain primitives (chains, retrievers, memory), LangGraph state machines and human-in-the-loop, LlamaIndex RAG pipelines and query engines, AutoGen multi-agent conversations, CrewAI role-based crews, tool-calling orchestration, ReAct reasoning traces, planning vs reflection, agent observability, and evaluation of agent traces. Three worked examples per topic.