Distributed Multi-Agent eCommerce System (A2A)

This project implements a distributed agent architecture using the Agent2Agent (A2A) protocol. Each agent runs as a separate service in a Docker container, communicating via HTTP while sharing state through a central Redis database.

Architecture

graph TD
    User([User]) <--> Master[Master Router & UI]
    Master -- A2A --> Search[Search Agent Server]
    Master -- A2A --> Cart[Cart Agent Server]
    Master -- A2A --> Checkout[Checkout Agent Server]
    Master -- A2A --> Order[Order Status Agent Server]
    
    Search <--> Redis[(Redis State)]
    Cart <--> Redis
    Checkout <--> Redis
    Order <--> Redis

Loading

Services

Service	Port	Description
master-router	7860	Gradio UI & Request Coordinator (A2A Client)
search-agent	8000	Product Discovery (A2A Server)
cart-agent	8000	Cart Operations (A2A Server)
checkout-agent	8000	Order Processing (A2A Server)
order-agent	8000	Status Tracking (A2A Server)
redis	6379	Shared state for carts and orders

Setup & Running

1. Requirements:

   • Docker & Docker Compose
   • Python 3.10+ (for local development)

2. Start the system:

   docker-compose up --build

3. Access the UI: Open http://localhost:7860 in your browser.

Logic Implementation

The system uses the a2a-sdk to handle protocol-compliant message passing. Each remote agent serves an AgentCard and implements an AgentExecutor to process tasks. State is persisted in Redis, allowing agents to maintain context across the distributed environment.