Skip to content

A2A

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History
 
 

A2A

README.md

A2A (Agent-to-Agent) — Learning Example

A minimal example of Google ADK's A2A protocol: one agent delegates to another over HTTP. Uses Azure OpenAI or Gemini as the LLM.

Teaching Materials

Structured course content on the A2A protocol is in the docs/ folder:

# Topic
1 The Problem A2A Solves
2 A2A Core Concepts
3 A2A vs MCP vs APIs
4 A2A Architecture
5 Environment Setup
6 Build First Agent
7 Expose Agent with A2A
8 A2A Client
9 Agent Orchestration
10 This Repository's Example

→ Full index

Get the Repository

git clone git@github.com:MSKazemi/ai-agent-systems-course.git
cd ai-agent-systems-course/A2A

Or with HTTPS:

git clone https://github.com/MSKazemi/ai-agent-systems-course.git
cd ai-agent-systems-course/A2A

Quick Start

1. Install

Create and activate a Python virtual environment:

python3 -m venv .venv
source .venv/bin/activate   # Linux/macOS
# .venv\Scripts\activate    # Windows

Install dependencies:

pip install -r requirements.txt

2. Configure

cp .env.example .env

Edit .env. For Azure OpenAI:

LLM_PROVIDER=azure
AZURE_OPENAI_API_KEY=<from Azure Portal → Keys and Endpoint>
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_API_VERSION=2024-08-01-preview
AZURE_PRIMARY_LLM_DEPLOYMENT_NAME=gpt-4o

For Gemini:

LLM_PROVIDER=gemini
GOOGLE_API_KEY=<from aistudio.google.com/apikey>

3. Run

Terminal 1 — Math agent (port 8001):

source .venv/bin/activate   # Linux/macOS
uvicorn remote_math_agent:a2a_app --host 127.0.0.1 --port 8001

Terminal 2 — Prime agent (port 8003):

source .venv/bin/activate   # Linux/macOS
uvicorn remote_prime_agent:a2a_app --host 127.0.0.1 --port 8003

Terminal 3 — Coordinator (chat UI):

source .venv/bin/activate   # Linux/macOS
adk web . --port 8002

Open http://127.0.0.1:8002. Select coordinator or dynamic_coordinator, and try:

  • "What is 17 + 23?" → delegates to math agent
  • "Is 17 a prime number?" → delegates to prime agent

Verify Azure

If you get 401 errors, test your credentials:

python verify_azure.py

Troubleshooting

Problem Solution
No agents found Run adk web from project root
Port in use Use adk web . --port 8003
Azure 401 Regenerate key in Azure Portal, update .env
Remote unreachable Start both uvicorn servers (8001, 8003) before adk web
pip: cannot execute: required file not found Recreate venv (see below)

Recreate broken venv:

deactivate
rm -rf .venv
python3 -m venv .venv
source .venv/bin/activate   # Linux/macOS
pip install -r requirements.txt

Links