course-creation-agent (Distributed)

A multi-agent system built with Google's Agent Development Kit (ADK) and Agent-to-Agent (A2A) protocol. It features a team of microservice agents that research, judge, and build content, orchestrated to deliver high-quality results.

Architecture

This project uses a distributed microservices architecture where each agent runs in its own container and communicates via A2A:

• Orchestrator Service (orchestrator): The main entry point. It manages the workflow using LoopAgent and SequentialAgent, and connects to other agents using RemoteA2aAgent.
• Researcher Service (researcher): A standalone agent that gathers information using Google Search.
• Judge Service (judge): A standalone agent that evaluates research quality.
• Content Builder Service (content_builder): A standalone agent that compiles the final course.
• Agent App (app): A web application that queries the Orchestrator agent, displays progress and results.

Project Structure

course-creation-agent/
├── agents/
    ├── orchestrator/        # Main Orchestrator agent, ADK API Service
    ├── researcher/          # Researcher agent, A2A microservice
    ├── judge/               # Judge agent, A2A microservice
    └── content_builder/     # Content Builder agent, A2A microservice
├── app/                     # Web App service application
    └── frontend/            # Frontend application
├── shared/                  # Files used by all agents
└── ...

Shared files

There are some files in shared directory that are shared across all agents and the web app. To avoid duplication, these files are linked into respective subdirectories as symlinks.

• a2a_utils.py - contains code for rewriting agent URLs in A2A AgentCard when deployed in Cloud Run.
• adk_app.py - ADK API Service implementation with additional A2A functionality.
• authenticated_httpx.py - httpx client extension for service-to-service requests.

Requirements

• uv: Python package manager (required for local development).
• Google Cloud SDK: For GCP services and authentication.

Quick Start

1. Install Dependencies:

   uv sync

2. Set up credentials: Ensure you have Google Cloud credentials available. You might need to run:

   gcloud auth application-default login

   And ensure your GOOGLE_CLOUD_PROJECT environment variable is set.

3. Run Locally:

   ./run_local.sh

   This will start all 4 agents and the web app in background processes

4. Access the App: Open http://localhost:8000 in your browser.

Deployment

To deploy to Google Cloud Run, you need to deploy each service individually and then configure the Orchestrator with the URLs of the other services.

1. Deploy Researcher, Judge, Content Builder, and Orchestrator: Deploy each of these folders as a separate Cloud Run service. Note down their URLs (e.g., https://researcher-xyz.a.run.app).

2. Deploy Agent App: Deploy the app/ folder to Cloud Run. Set the following environment variables on the Agent App service:

   • RESEARCHER_AGENT_CARD_URL: https://<researcher-url>/a2a/agent/.well-known/agent.json
   • JUDGE_AGENT_CARD_URL: https://<judge-url>/a2a/agent/.well-known/agent.json
   • CONTENT_BUILDER_AGENT_CARD_URL: https://<content-builder-url>/a2a/agent/.well-known/agent.json
   • AGENT_URL: https://<orchestrator-url>

3. Access: Open the App's URL in your browser.