Workshop: building agents and multi-agent systems with AgentWorkflow in LlamaIndex
In this comprehensive, step-by-step workshop originally delivered at ODSC's AI Builders Summit, Laurie Voss walks you through LlamaIndex's features for building agents and multi-agent systems. First we walk you through using AgentWorkflow, a high-level construct for building agents and multi-agent systems. Then we drop down one level to Workflows, the building blocks of multi-agent systems in LlamaIndex, and build a multi-agent system from scratch.
The full list of topics covered includes:
Introduction:
What is LlamaIndex?
Why should you use it?
AgentWorkflow
Connecting to an LLM
Creating an agent tool
Creating a single agent using AgentWorkflow
Adding state management
Streaming events and output
Getting human-in-the-loop input
Building a multi-agent system using AgentWorkflow
Workflows
Creating a basic Workflow
Visualizing Workflows
Multi-step workflows
Looping
Branching
Concurrent events
Streaming Workflow events
Building a multi-agent system from Workflows from scratch
Adding self-reflection
Full notebook here: https://colab.research.google.com/dri...