Agentic AI & Multi-Agent Workflows - Daftar Isi

Tujuan: Menguasai arsitektur, framework, dan production deployment AI agents — dari single-agent tool-use sampai multi-agent systems Level: Intermediate — Advanced Prasyarat: Familiar dengan LLMs, Python, dan API integration Estimasi Total: 10-14 jam baca + praktik

---

Struktur Materi

#	File	Topik	Estimasi
01	01-what-is-agentic-ai.md	What Is Agentic AI? Agents vs Models vs Workflows vs Automations	30 menit
02	02-agent-architecture.md	Agent Architecture: Planning, Memory, Tools, Action Loop	45 menit
03	03-langchain-and-langgraph.md	LangChain & LangGraph: Stateful Agents, Graph-Based Orchestration	55 menit
04	04-crewai-multi-agent.md	CrewAI: Role-Based Multi-Agent Collaboration, Task Delegation	50 menit
05	05-autogen-microsoft.md	AutoGen (Microsoft): Conversational Agents, Code Execution, Agent Networks	45 menit
06	06-openai-swarm-and-assistants.md	OpenAI Swarm & Assistants API: Lightweight Orchestration, Built-in Tools	35 menit
07	07-tool-use-and-function-calling.md	Tool Use & Function Calling: API Integration, Code Execution, Browser Automation	40 menit
08	08-agentic-rag.md	Agentic RAG: Multi-Hop Retrieval, Query Planning, Self-Correction	45 menit
09	09-human-in-the-loop.md	Human-in-the-Loop: Approval Workflows, Interrupts, Oversight Patterns	35 menit
10	10-memory-and-persistence.md	Memory & Persistence: Short-Term, Long-Term, Vector Memory, Checkpointing	35 menit
11	11-evaluation-and-testing.md	Evaluation & Testing: Agent Metrics, Trajectory Evaluation, Red-Teaming	35 menit
12	12-production-deployment.md	Production Deployment: Monitoring, Guardrails, Cost Control, Scaling	40 menit
13	13-real-world-agent-projects.md	Real-World Agent Projects: Research Assistant, Code Review Bot, Customer Support	50 menit

---

Framework Comparison

Framework	Best For	Key Feature	Learning Curve
LangGraph	Complex stateful workflows	Graph-based state management, persistence	Medium-High
CrewAI	Role-based collaboration	Intuitive role/task abstraction, quick to start	Low-Medium
AutoGen	Conversational multi-agent	Microsoft ecosystem, built-in code execution	Medium
OpenAI Swarm	Lightweight orchestration	Minimalist, experimental, easy to understand	Low
OpenAI Assistants	Quick prototyping	Built-in RAG, code interpreter, function calling	Low

The Agent Loop

THINK → PLAN → ACT → OBSERVE → REFLECT → (repeat)
  │       │      │       │          │
  │       │      │       │          └── Update memory, learn
  │       │      │       └───────────── Get tool output / environment feedback
  │       │      └────────────────────── Execute tool / API call
  │       └───────────────────────────── Break goal into steps
  └───────────────────────────────────── Understand goal + context

Prasyarat

• Python intermediate (functions, classes, async, API calls)
• Familiar dengan LLM APIs (OpenAI, Anthropic, atau open-source)
• Sudah membaca ai-fluency/15 (Advanced Prompting) dan ai-fluency/16 (AI Agents)
• Recommended: sudah membaca how-ai-works/06 (How LLMs Work)