aaronpowell / GenerateAndChill

A demo app of how to use Azure OpenAI Service to generate verify prompts and generate images.

Image Generator Sample

This sample application uses Azure OpenAI Service to generate images from text.

How it works

There are two parts to the AI component of this application, initial text processing of the user prompt and image generation.

Text Processing

To aim for consistent generation of images the text prompt provided needs to be in English and needs to have a provided style for the image to be generated. To ensure this, an initial Chat Completion is performed using a System Prompt to guide the LLM into how it should generate the prompt.

The System Prompt can be found here.

Image Generation

Once the prompt has been returned from the Chat Completion it is provided to the image model to generate an image. The returned image is then stored in Azure Storage before being returned to the user.

Running the sample

Local development

This sample uses a devcontainer to prepare a development environment, meaning it can be opened in GitHub Codespaces or locally using the VSCode Remote Containers extension.

Manual setup

If you don't want to use the devcontainer you need to ensure you have the following:

• .NET 7 SDK
• Node.js 20.x
• Azurite
  • Alternatively, you can use a deployed Azure Storage account
• Azure OpenAI Service
  • You'll need to create a new resource and note the API key and endpoint
• Azure Developer CLI

Provisioning Azure resources

To provision the Azure resources you can use the Azure Developer CLI and run the following commands:

azd auth login
azd env new -n MyDemoEnv
azd provision

Once the deployment is complete, navgiate to the resource group in Azure and copy the OpenAI + Storage connection string into the appsettings.Development.json file like so:

{
  "Azure": {
    "OpenAIEndpoint": "<endpoint here>",
    "OpenAIKey": "<key here>",
    "BlobStorageConnectionString": "<storage connection string here>",
    "TableStorageConnectionString": "<storage connection string here>"
  }
}

Running the application

From VS Code launch the application with the Full Stack launch profile, which will start the .NET server, the React frontend and the Azure Static Web Apps CLI.

Navigate to http://localhost:4280 to see the application running.

Deploying to Azure

To deploy the application to Azure you can use the Azure Developer CLI and run the following commands:

azd up

Once the deployment is complete the URL for the frontend application will be displayed in the terminal.

License

MIT