Agentic AI Patterns: ReAct, Plan-and-Execute, and Multi-Agent
«Agentic AI Patterns: ReAct, Plan-and-Execute, and Multi-Agent» è un corso online di RAG e AI engineering di livello avanzato, tenuto da InsightLearn Curated AI su InsightLearn. Il percorso comprende 9 lezioni organizzate in 3 sezioni, per una durata complessiva di 13h 0m. Build production-grade AI agents with ReAct, Plan-and-Execute, and multi-agent patterns with reliability guarantees.
- Docente: InsightLearn Curated AI
- Categoria: RAG e AI engineering
- Livello: avanzato
- Durata: 13h 0m
- Accesso: gratuito in questa fase di validazione, in cambio di un feedback al termine
- Certificato: certificato di completamento verificabile online
Descrizione del corso
InsightLearn è in fase di sviluppo (work in progress): i contenuti di questa pagina sono pubblicati esclusivamente a fini di verifica e test.
Move beyond chatbots to building reliable autonomous agents. Covers the ReAct (Reason + Act) pattern, Plan-and-Execute agents, Reflexion and self-critique loops, agent memory architectures (working, episodic, semantic), tool selection strategies, multi-agent communication patterns (hierarchical, collaborative, adversarial), agent state machines with LangGraph, human-in-the-loop interruption points, agent debugging and tracing, and production reliability patterns for long-running agents.
Cosa imparerai
- Curated from public resources. Build production-grade AI agents with ReAct, Plan-and-Execute, and multi-agent patterns with reliability guarantees.
Programma del corso
- ReAct: combining reasoning and acting in LLMs (32 min)
- Plan-and-Execute: separating planning from execution (28 min)
- Reflexion: self-critique and iterative improvement (30 min)
- Agent memory architectures: working, episodic, semantic (35 min)
- Agent state machines with LangGraph (40 min)
- Human-in-the-loop: interruption and approval patterns (30 min)
- Multi-agent communication: hierarchical and collaborative (38 min)
- Agent reliability: retries, timeouts, and circuit breakers (32 min)
- Debugging agents with LangSmith traces (28 min)
Dalle lezioni
Estratto della trascrizione di «ReAct: combining reasoning and acting in LLMs»
Hey everyone, welcome to this video and today we are going to learn what are React agents. We'll start by understanding why do we need agents, what exactly is a React agent and then we are going to use an open source framework called dynamic to build a React agent. Dynamic comes with Apache 2 license and you can also use it to build drag applications or for doing multi-agent orchestration as well as managing any complex LLM workflow. So with that, let's come to the very first question, why do we need agents in the first place? So LLMs in their original state have ability to reason or think, but they cannot take actions or they cannot interact with…
Estratto della trascrizione di «Plan-and-Execute: separating planning from execution»
Hi all, this is Will from LangChain, today I'm going to walk through how to create plan and execute style agents in LangGraph. For those of you not familiar with LangGraph, we've got a great set of videos on YouTube that you can go through to get a little more of an introduction, but a short summary of it would be that it's a framework built on top of LangChain Core that provides a graph based syntax that gives you a great balance between expressivity and control when building things like agents and state machines that require loop type workflows. You can go on, you can see. it here, check out the repo in the Linechain AI organization in GitHub.…
Estratto della trascrizione di «Reflexion: self-critique and iterative improvement»
Welcome, architects and developers. Today, we are diving into one of the most powerful paradigms in agentic AI, self-reflection. Specifically, we will master the iterative refinement of outputs to elevate our agents from simple answering machines to methodical reasoning engines. Let's get started. Let's define our core concept. Iterative refinement is a self-reflection pattern wherein agent evaluates and improves its own output over multiple cycles. Why do we need this? By default, large language models operate using what psychologists call system one thinking fast, instinctive, and reactive. They generate the most statistically likely next tokens without pausing to evaluate the overall strategy. However, complex tasks require system two, thinking methodical, slow, and reflective reasoning. By forcing the LLM to draft…
Altri corsi di RAG e AI engineering
- AI Engineering: Building Production LLM Applications
- AI Safety: Alignment, Red-Teaming, Responsible Deployment
- Fine-Tuning LLMs: LoRA, QLoRA, and Full Fine-Tuning
- Function Calling & Tool Use: Multi-Model Production Patterns
- LangSmith: Tracing, Debugging, and Monitoring LLM Apps
- LLM Evals: RAGAS, DeepEval, and Automated Testing
Domande frequenti
Quanto dura il corso Agentic AI Patterns: ReAct, Plan-and-Execute, and Multi-Agent?
Il corso dura 13h 0m in totale, con 9 lezioni suddivise in 3 sezioni, da seguire online al proprio ritmo.
Serve esperienza pregressa?
Il corso è di livello avanzato e non indica requisiti specifici di ingresso.
Il corso rilascia un certificato?
Sì: al completamento del corso ricevi un certificato di completamento verificabile online.