Agent Framework Playground

This repository is a lightweight playground for experimenting with the agent-framework-core package and Azure OpenAI. It contains a simple Python chat client plus a set of companion notebooks that demonstrate how to build and orchestrate agents with the magnetic workflow building blocks.

Repository layout

Path	Description
chat_client.py	Minimal agent chat loop that connects to Azure OpenAI via AzureOpenAIChatClient.
config.py	Loads configuration from .env and exposes Azure-specific constants.
requirements.txt	Runtime dependencies for the Python scripts and notebooks.
simple_agent.ipynb / supervisor.ipynb / tool*.ipynb	Notebook experiments that accompany the agent framework.
.env.sample	Template for the environment variables required by config.py.

Prerequisites

• Python 3.10 or newer
• Azure OpenAI resource with a GPT‑4o (or compatible) deployment
• (Optional) Jupyter Notebook or VS Code for exploring the notebooks

Setup

1. Create and activate a virtual environment:

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

2. Install dependencies:

   pip install -r requirements.txt

3. Configure secrets by copying the sample and filling in your values:

   cp .env.sample .env

   Required variables:

   • AZURE_OPENAI_ENDPOINT
   • AZURE_OPENAI_API_KEY
   • AZURE_OPENAI_DEPLOYMENT_GPT_4O_ID
   • AZURE_OPENAI_DEPLOYMENT_ID (used by config.py, add this entry to .env as needed)

Running the chat client

After setup, execute:

python chat_client.py

The script instantiates AzureOpenAIChatClient, wires it into a MagenticBuilder, and streams agent responses to the console. Adjust logging levels inside chat_client.py if you need more verbose tracing.

Working with the notebooks

1. Launch Jupyter (or VS Code’s notebook interface).
2. Open any of the *.ipynb files:
   • 00.install.ipynb walks through environment bootstrap steps.
   • magentic_one.ipynb and simple_agent.ipynb illustrate basic agent patterns.
   • supervisor.ipynb, tool.ipynb, and tools.ipynb demonstrate orchestration examples.
3. Ensure the notebook kernel uses the same virtual environment that has the dependencies installed and the .env file available so that config.py can resolve credentials.

Troubleshooting

• Auth failures: Double-check that the .env file contains the exact deployment IDs and endpoint/key for the Azure resource region you created.
• Package import errors: Re-run pip install -r requirements.txt inside the active environment to ensure agent-framework-core and friends are available.
• Notebook cannot find variables: Restart the kernel after editing .env, or call load_dotenv() manually at the top of a notebook cell.

Contributing

This is still a personal sandbox, so feel free to fork or adapt to your own Azure OpenAI setup. If you expand the framework usage (e.g., add tools or workflows), document them here to keep the README in sync.