Skip to content

This repo helps you to build a team of AI agents with Autogen

License

Notifications You must be signed in to change notification settings

Azure-Samples/dream-team

Repository files navigation

PRs Welcome LinkedIn

Build your dream team with Autogen

This repository leverages Microsoft Autogen 0.4, Azure OpenAI and integrates it with Streamlit, to build an end to end multi agents application, this repo makes it easy to build test and deploy an advanced multi agent framework, based on Magentic One

Architecture

🎉 December 3, 2024: The repo now support one click deployment with Azure Developer CLI, if you would like to run it with the full process localy you can check v0.21

🎉 November 18, 2024: we are porting this repo to Autogen 0.4, A new event driven, asynchronous architecture for AutoGen and Magentic One

Key Features

Dream Team offers the following key features:

  • Advanced multi agent framework: this solution is based on the popular framework Autogen(35K stars) and Magentic One
  • Friendly UI: easy way to build and share data apps powered by Streamlit
  • Single line deployment: developer-friendly deployment that accelerates your path from a local development environment to Azure with single line of code - azd up.
  • Secure code execution: Fast access to secure sandboxed with strong isolation environments that are ideal for running code or applications with Azure Container Apps dynamic sessions.
  • Managed Identities: Built in Azure Managed identities to eliminate the need for developers to manage these credentials
  • Observability & Debugging: Built-in features and tools for tracking, tracing, and debugging agent interactions and workflows, including PromptFlow tracing.

Prerequisites:

  1. Install Azure Developer CLI.
  2. Ensure you have access to an Azure subscription
  3. Docker - Follow the official Docker installation instructions
  4. Python version >= 3.8, < 3.13
  5. Install UV - optional for running locally

Step by step deployment

1. Clone the repository

git clone https://github.com/Azure-Samples/dream-team  

2. Login to your Azure account

azd auth login

You need to choose your preferred region (you can start with east us or sweden central or any other available region)

3. Deploy Azure Resources and the app

azd up

Working locally

cd src 

Set up a virtual environment (Preferred)

uv venv

Once you’ve created a virtual environment, you may activate it.

On Windows, run:

.venv\Scripts\activate

On Unix or MacOS, run:

source .venv/bin/activate

To deactivate :

deactivate

More information about virtual environments can be found here

Install dependencies

uv sync
git clone --depth 1 --branch v0.4.0.dev11 https://github.com/microsoft/autogen.git
cd autogen/python/packages/autogen-magentic-one
uv pip install -e .
  • Navigate to src folder
cd ../../../..
playwright install --with-deps chromium

Update configuration

  • If you used AZD to deploy the resources just run the code below
azd env get-values > .env
  • Alternatively, copy .env.sample (under src) into .env

Important: Magentic-One code uses code execution, you need to have Docker installed to run the examples if you use local execution

Run

streamlit run app.py

If your app is ready, you can browse to (typically) http://localhost:8501 to see the app in action. Screenshot

Learn

Check these resources:

  1. Blogpost - Build your dream team with Autogen
  2. Webinar - More agents is all you need