Intro to Agentic Systems in AI

Literature

Zhihua Duan, Jialin Wang (2024): Exploration of LLM Multi-Agent Application Implementation Based on LangGraph+CrewAI

Resources

• CrewAI documentation
  • Focus on intro, agents, tasks, crews, flows, and tools
  • Also, check out CrewAIs framework examples. Specifically, the workshop will depart from the Book writing crew example.
• Pydantic
• Python packaging convention
• Youtube video with examples and explanation of flows and states in agentic systems.

Building Multi-Agent Systems with CrewAI

This workshop explores the development of agentic systems - building AI applications where multiple specialized agents collaborate to accomplish complex tasks. Using CrewAI as our framework, we’ll examine both theoretical concepts and practical implementation.

Key Topics

1. Agentic Systems Fundamentals

• Role specialization and agent collaboration
• Task delegation and coordination
• State management in multi-agent systems

2. CrewAI Framework

• Agents: Specialized roles with defined goals
• Crews: Orchestrating agent collaboration
• Tasks: Structured work units
• Flow: Managing state and process control

3. Project Structure and Best Practices

• Organizing code with clear folder hierarchies
• Configuration management using YAML
• Data validation with Pydantic models
• Testing and debugging agent interactions

Real-world application

We’ll examine a tender document generation system that demonstrates:

• Analysis and writing agent collaboration
• Structured data handling
• State management across multiple tasks
• Output validation and formatting