Visão Geral
Este curso apresenta os conceitos, arquiteturas e práticas para criação de agentes inteligentes baseados em Large Language Models (LLMs). O foco está em transformar LLMs em agentes capazes de planejar, decidir, executar ações e interagir com ferramentas, sistemas e usuários de forma autônoma e controlada.
Conteúdo Programatico
Module 1 – Introduction to LLM-Based Agents
- What are LLM-based agents
- From chatbots to autonomous agents
- Capabilities and limitations of LLMs
- Agent use cases
Module 2 – Agent Architecture with LLMs
- Core components of an agent
- Memory, reasoning and action layers
- Stateless vs stateful agents
- Control flow patterns
Module 3 – Prompt Engineering for Agents
- System prompts and role definition
- Task decomposition
- Chain-of-thought and reasoning strategies
- Prompt robustness
Module 4 – Tool Use and Action Execution
- Tool calling concepts
- API integration
- Function calling patterns
- Error handling and recovery
Module 5 – Memory and Context Management
- Short-term vs long-term memory
- Vector databases fundamentals
- Context window management
- Knowledge retrieval strategies
Module 6 – Planning and Autonomous Decision Making
- Planning with LLMs
- ReAct and planning loops
- Task prioritization
- Autonomous execution cycles
Module 7 – Multi-Agent Systems with LLMs
- Single-agent vs multi-agent design
- Agent collaboration and delegation
- Communication protocols
- Coordination challenges
Module 8 – Safety, Ethics and Best Practices
- Hallucination and reliability
- Guardrails and constraints
- Security and data privacy
- Responsible deployment of LLM agents