Magentic-One, Sample Implementation using Streamlit

This is a repo to demostrate an implementation of Magentic-One, using Streamlit and Microsoft Autogen 0.4. The deployment is done via GitOps (Flux and Kustomize) with Kubernetes, and the python apps can run locally as well without deploying to Kubernetes.

At the same time, this repo is also a demo of how to deploy apps to Azure Arc enabled Kubernetes clusters, which can be on-prem or in other clouds.

Folder Structure

• magentic-one-streamlit/: Contains the Streamlit app
• base: Contains the yaml to deploy the app to Kubernetes
• overlays: Contains the configuration for different environments (cloud, on-prem, etc)

How to run locally

Refers to the README in magentic-one-streamlit/ for instructions on how to run the app locally.

How to deploy to Kubernetes

This implementation is done in Azure Kubernetes Services, as well as Azure Arc enabled Kubernetes clusters. The deployment here assumes you have a Kubernetes cluster ready. The deployment leverages Azure Key Vault as secret provider for the Kubernetes cluster.

Step 0: Prerequisites

Have a Kubernetes cluster ready, either Azure Kubernetes or Azure Arc enabled Kubernetes. Refers to the following for setting up the cluster:

• Azure Kubernetes Services
• Azure Arc enabled Kubernetes

Next, set up Azure Key Vault and store the secrets required for the deployment. Refers to the following for setting up Azure Key Vault:

• Azure Key Vault

Lastly, setup a container registry to store the container images. Refers to the following for setting up Azure Container Registry:

• Azure Container Registry

Follows the instructions in the README in magentic-one-streamlit/ to build the container image.

Step 1: Create service principal for Kubernetes

Create a service principal for the Kubernetes cluster. Refers to the following for creating a service principal:

• Create service principal

This service principal will be used to perform the following actions:

• Pull the container image from the container registry
• Access secret from Azure Key Vault

Step 2: Enable Azure Key Vault Provider for Kubernetes

Azure Key Vault Provider for Kubernetes allows you to use Azure Key Vault as a secret provider for Kubernetes. Refers to the following for enabling Azure Key Vault Provider for Kubernetes:

• Azure Key Vault Provider for Azure Kubernetes Service
• Azure Key Vault Provider for Azure Arc enabled Kubernetes

Once the Azure Key Vault Provider is enabled, don't forget to grant the service principal access to the Key Vault. Then, create necessary secrets in Azure Key Vault as per the README in magentic-one-streamlit/.

For demo purposes, the apps will deploy in a new namespace within the Kubernetes cluster. Run the following command to create the namespace:

kubectl create namespace arc-app-demo

To enable the authentication via service principal, you need to create a secret in the Kubernetes cluster. Run the following command to create the secret:

kubectl create secret generic secrets-store-creds --from-literal clientid="xxxxxxxx-xxx-xxxx-xxxx-xxxxxxxxxxxx" --from-literal clientsecret="xxxxxxxxxxxxxx" -n arc-app-demo

kubectl label secret secrets-store-creds secrets-store.csi.k8s.io/used=true -n arc-app-demo

Lastly, deploy the Kubernetes sercet provider class. Refers to csi-secret/akv-sample.yaml for the sample configuration.

kubectl deploy -f csi-secret/akv-sample.yaml

Step 3: Deploy the app to Kubernetes

There are two ways to deploy the app to Kubernetes:

1. Using kubectl
2. Using Flux and Kustomize - GitOps

The deployment consist of a single replica of the app, and a service - Load Balancer. The deployment is done in the base folder, and the configuration for different environments is done in the overlays folder.

Using kubectl

Run the following command to deploy the app to Kubernetes:

kubectl apply -k overlays/cloud # For cloud environment
kubectl apply -k overlays/arc # For azure arc enabled kubernetes

Using GitOps (Flux and Kustomize) in Kubernetes

Refer to the following guide to setup GitOps in Azure Kubernetes Services or Azure Arc enabled Kubernetes:

• GitOps in Azure Kubernetes Services

Step 4: Access the app

Once the deployment is successful, you can access the app via the Load Balancer IP. You will have a public IP for Azure Kubernetes Services, and a private IP (local IP) for Azure Arc enabled Kubernetes, hence access the app accordingly.

Demo screenshot

Download the full video demo here: Magentic-One Demo Video