autogen/docs/github-flow-getting-started.md

Prerequisites

• Access to gpt3.5-turbo or preferably gpt4 - Get access here
• Setup a Github app
• Install the Github app
• Provision the azure resources
• Create labels for the dev team skills

How do I setup the Github app?

• Register a Github app, with the options listed below:

  • Give your App a name and add a description
  • Homepage URL: Can be anything (Example: repository URL)
  • Add a dummy value for the webhook url, we'll come back to this setting
  • Enter a webhook secret, which you'll need later on when filling in the WebhookSecret property in the appsettings.json file
  • Setup the following permissions
    • Repository
      • Contents - read and write
      • Issues - read and write
      • Metadata - read only
      • Pull requests - read and write
  • Subscribe to the following events:
    • Issues
    • Issue comment
  • Allow this app to be installed by any user or organization

• After the app is created, generate a private key, we'll use it later for authentication to Github from the app

Which labels should I create?

In order for us to know which skill and persona we need to talk with, we are using Labels in Github Issues.

The default bunch of skills and personnas are as follow:

• PM.Readme
• Do.It
• DevLead.Plan
• Developer.Implement

Add them to your repository (They are not there by default).

Once you start adding your own skills, just remember to add the corresponding label to your repository.

How do I run this locally?

Codespaces are preset for this repo. For codespaces there is a 'free' tier for individual accounts. See: https://github.com/pricing Start by creating a codespace: https://docs.github.com/en/codespaces/developing-in-a-codespace/creating-a-codespace-for-a-repository

Within this repo there is a folder "scr\sample-apps\gh-flow" Within this folder there are two files called appsettings.azure.template.json and appsettings.local.template.json. If you run this demo locally, use the local template and if you want to run it within Azure use the Azure template. Rename the selected file to appsettings.json and fill out the config values within the file.

In the Explorer tab in VS Code, find the Solution explorer, right click on the gh-flow project and click Debug -> Start new instance

We'll need to expose the running application to the GH App webhooks, for example using DevTunnels, but any tool like ngrok can also work. The following commands will create a persistent tunnel, so we need to only do this once:

TUNNEL_NAME=_name_yout_tunnel_here_
devtunnel user login
devtunnel create -a $TUNNEL_NAME
devtunnel port create -p 5244 $TUNNEL_NAME

and once we have the tunnel created we can just start forwarding with the following command:

devtunnel host $TUNNEL_NAME

Copy the local address (it will look something like https://yout_tunnel_name.euw.devtunnels.ms) and append /api/github/webhooks at the end. Using this value, update the Github App's webhook URL and you are ready to go!

Before you go and have the best of times, there is one last thing left to do load the WAF into the vector DB

Also, since this project is relying on Orleans for the Agents implementation, there is a dashboard available at https://yout_tunnel_name.euw.devtunnels.ms/dashboard, with useful metrics and stats related to the running Agents.

How do I deploy the azure bits?

This repo is setup to use azd to work with the Azure bits. azd is installed in the codespace.

Let's start by logging in to Azure using

azd auth login

After we've logged in, we need to create a new environment provision the azure bits.

ENVIRONMENT=_name_of_your_env
azd env new $ENVIRONMENT
azd provision -e $ENVIRONMENT

After the provisioning is done, you can inspect the outputs with the following command

azd env get-values -e dev

As the last step, we also need to load the WAF into the vector DB

Load the WAF into Qdrant.

If you are running the app locally, we have Qdrant setup in the Codespace and if you are running in Azure, Qdrant is deployed to ACA. The loader is a project in the src\apps folder, called seed-memory. We need to fill in the appsettings.json file in the config folder with the OpenAI details and the Qdrant endpoint, then just run the loader with dotnet run and you are ready to go.