06-multi-agent-systems

06. Multi-Agent Systems

Agent orchestration and collaboration patterns.

Blog Post: https://arjunprabhulal.com/adk-multi-agent-systems/

Table of Contents

1. Overview
2. Prerequisites
3. Setup Steps
4. Delegation Patterns
5. Running the Agent
6. Next Steps

Overview

Multi-agent systems divide work among specialists, coordinate interactions, and manage shared context. Key benefits:

• Specialization - Each agent focuses on specific tasks
• Scalability - Add new specialists without changing the coordinator
• Maintainability - Easier to update individual components

Prerequisites

• Python 3.10+
• Gemini API key from AI Studio

Setup Steps

1. Navigate to this module:

cd 06-multi-agent-systems

2. Create and activate a virtual environment:

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies:

pip install -r ../requirements.txt

4. Set up environment variables in customer_support/.env:

GOOGLE_API_KEY=your-api-key-here

Delegation Patterns

Pattern	Description	Use Case
Hub-Spoke	Central coordinator delegates to specialists	Customer support routing
Pipeline	Chain of agents processing in sequence	Content creation workflow
Swarm	Agents collaborate dynamically	Complex research tasks

Running the Agent

Using ADK Web

adk web

Open http://127.0.0.1:8000 and select customer_support.

Test Queries:

• "Research AI trends and write a summary"
• "I need help with a technical issue"

Using ADK CLI

adk run customer_support

Next Steps

Continue to 07. Built-in Tools