What are multi-agent systems?

Building Agentic Workflows with LlamaIndex

Laurie Voss

VP Developer Relations, LlamaIndex

Beyond one agent

$$

  • Single agent: powerful, but limited

$$

  • Specialized agents can be more effective

$$

  • Can collaborate as a team of agents

team_of_agents.png

Building Agentic Workflows with LlamaIndex

Why use many agents?

Example:

  • Writing a research report
  • One agent juggling: searching, compiling, writing, checking, and returning the answer

agent_workflow.gif

Building Agentic Workflows with LlamaIndex

Meet the specialized agents

$$

  • Research Agent: gathers information

$$

  • Write Agent: drafts the report

$$

  • Review Agent: checks and improves content

workflow_agent_team.gif

Building Agentic Workflows with LlamaIndex

How they work together

$$

  • Research Agent collects notes

$$

  • Write Agent composes draft using notes

$$

  • Review Agent refines final output

workflow_agent_team.gif

Building Agentic Workflows with LlamaIndex

When to scale up

  • If tasks naturally split into parts

  • If one agent struggles with complex prompts or too many tools

complex_task.png

Building Agentic Workflows with LlamaIndex

Built-in AgentWorkflow

$$

  • LlamaIndex AgentWorkflow

$$

  • Supports handing tasks between agents automatically

$$

  • Brings a layer of complexity

agentworkflow.png

Building Agentic Workflows with LlamaIndex

Let's practice!

Building Agentic Workflows with LlamaIndex

Preparing Video For Download...