Skip to content

README.md

Latest commit

 

History

History
138 lines (105 loc) · 4.04 KB

README.md

File metadata and controls

138 lines (105 loc) · 4.04 KB

🤖 TellTimeAgent & Multi-Agent Demo – A2A with Google ADK

Welcome to TellTimeAgent and the Multi-Agent demo — a minimal Agent2Agent (A2A) implementation using Google's Agent Development Kit (ADK).

This example demonstrates how to build, serve, and interact with three A2A agents:

  1. TellTimeAgent – replies with the current time.
  2. GreetingAgent – fetches the time and generates a poetic greeting.
  3. OrchestratorAgent – routes requests to the appropriate child agent.

All of them work together seamlessly via A2A discovery and JSON-RPC.

📦 Project Structure

version_3_multi_agent/
├── .env                         # Your GOOGLE_API_KEY (not committed)
├── pyproject.toml              # Dependency config
├── README.md                   # You are reading it!
├── app/
│   └── cmd/
│       └── cmd.py              # CLI to interact with the OrchestratorAgent
├── agents/
│   ├── tell_time_agent/
│   │   ├── __main__.py         # Starts TellTimeAgent server
│   │   ├── agent.py            # Gemini-based time agent
│   │   └── task_manager.py     # In-memory task handler for TellTimeAgent
│   ├── greeting_agent/
│   │   ├── __main__.py         # Starts GreetingAgent server
│   │   ├── agent.py            # Orchestrator that calls TellTimeAgent + LLM greeting
│   │   └── task_manager.py     # Task handler for GreetingAgent
│   └── host_agent/
│       ├── entry.py            # CLI to start OrchestratorAgent server
│       ├── orchestrator.py     # LLM router + TaskManager for OrchestratorAgent
│       └── agent_connect.py    # Helper to call child A2A agents
├── server/
│   ├── server.py               # A2A JSON-RPC server implementation
│   └── task_manager.py         # Base in-memory task manager interface
└── utilities/
    ├── discovery.py            # Finds agents via `agent_registry.json`
    └── agent_registry.json     # List of child-agent URLs (one per line)

🛠️ Setup

  1. Clone & navigate

    git clone https://github.com/theailanguage/a2a_samples.git
cd a2a_samples/version_3_multi_agent

  2. Create & activate a venv

    python3 -m venv .venv
source .venv/bin/activate

  3. Install dependencies

    Using uv:

    uv pip install .

    Or with pip directly:

    pip install .

  4. Set your API key

    Create .env at the project root:

    echo "GOOGLE_API_KEY=your_api_key_here" > .env

🎬 Demo Walkthrough

Start the TellTimeAgent

python3 -m agents.tell_time_agent \
  --host localhost --port 10000

Start the GreetingAgent

python3 -m agents.greeting_agent \
  --host localhost --port 10001

Start the Orchestrator (Host) Agent

python3 -m agents.host_agent.entry \
  --host localhost --port 10002

Launch the CLI (cmd.py)

python3 -m app.cmd.cmd --agent http://localhost:10002

Try it out!

> What time is it?
Agent says: The current time is: 2025-05-05 14:23:10

> Greet me
Agent says: Good afternoon, friend! The golden sun dips low...

🔍 How It Works

  1. Discovery: OrchestratorAgent reads utilities/agent_registry.json, fetches each agent’s /​.well-known/agent.json.
  2. Routing: Based on intent, the Orchestrator’s LLM calls its tools:
    • list_agents() → lists child-agent names
    • delegate_task(agent_name, message) → forwards tasks
  3. Child Agents:
    • TellTimeAgent returns the current time.
    • GreetingAgent calls TellTimeAgent then crafts a poetic greeting.
  4. JSON-RPC: All communication uses A2A JSON-RPC 2.0 over HTTP via Starlette & Uvicorn.

📖 Learn More