{"id":52761,"date":"2024-07-22T10:05:00","date_gmt":"2024-07-22T17:05:00","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/dotnet\/?p=52761"},"modified":"2024-07-22T10:05:00","modified_gmt":"2024-07-22T17:05:00","slug":"add-ai-to-your-dotnet-apps-easily-with-prompty","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/dotnet\/add-ai-to-your-dotnet-apps-easily-with-prompty\/","title":{"rendered":"Add AI to Your .NET Apps Easily with Prompty"},"content":{"rendered":"

Adding AI features to .NET development is a new and exciting experience. In this blog post, we will explore Prompty and how to use it to integrate a Large Language Model, like GPT-4o, into your development flow and .NET applications.<\/p>\n

Introduction to Prompty<\/h2>\n

As AI enthusiasts and .NET developers, we constantly seek tools that simplify our workflows and enhance our productivity. One such powerful tool is Prompty, a Visual Studio Code extension designed to facilitate the integration of Large Language Models (LLMs) like GPT-4o into your applications. Prompty provides an intuitive interface to interact with LLMs directly from your development environment, making it easier than ever to add AI features to your projects.<\/p>\n

Prompty is developed by Microsoft and is available for free on the Visual Studio Code marketplace. Whether you are building chatbots, creating content generators, or experimenting with other AI-driven applications, Prompty can significantly streamline your development process.<\/p>\n

Description of a Typical Flow of a Developer Using Prompty in Visual Studio Code<\/h2>\n

\"What<\/p>\n

Let’s take a look at a sample flow on how to use Prompty. This process usually involves a few key steps:<\/p>\n

    \n
  1. \n

    Installation<\/strong>: Begin by installing the Prompty extension from the Visual Studio Code Marketplace<\/a>.<\/p>\n<\/li>\n

  2. \n

    Setup<\/strong>: After installation, configure the extension by providing your API keys and setting up the necessary parameters to connect to the LLM of your choice, such as GPT-4o.<\/p>\n<\/li>\n

  3. \n

    Integration<\/strong>: Prompty integrates seamlessly with your development workflow. Start by creating a new file or opening an existing one where you want to use the LLM. Prompty provides commands and snippets to easily insert prompts and handle responses.<\/p>\n<\/li>\n

  4. \n

    Development<\/strong>: Write prompts directly in your codebase to interact with the LLM. Prompty supports various prompt formats and provides syntax highlighting to make your prompts readable and maintainable.<\/p>\n

    You can use the extension to generate code snippets, create documentation, or even debug your applications by asking the LLM specific questions. Once you have your prompt ready, you can use the extension to generate code snippets, create documentation, or even debug your applications by asking the LLM specific questions.<\/p>\n<\/li>\n

  5. \n

    Testing<\/strong>: Test your prompts and adjust them as needed to get the desired responses from the LLM. Prompty allows you to iterate quickly, refining your prompts to improve the accuracy and relevance of the AI’s responses.<\/p>\n<\/li>\n<\/ol>\n

    Real Sample Using a WebAPI Application<\/h2>\n

    Let’s walk through a practical example of using Prompty in a .NET WebAPI application.<\/p>\n

    Step 1: Set up the WebAPI Project<\/h3>\n

    First, create a new WebAPI project, named PromptyWebAPI<\/code>, using the .NET CLI:<\/p>\n

    dotnet new webapi -n PromptyWebAPI\ncd PromptyWebAPI<\/code><\/pre>\n
    Add the following dependencies:<\/p>\n
    dotnet add package Microsoft.SemanticKernel --version 1.15.1\ndotnet add package Microsoft.SemanticKernel.Prompty --version 1.15.1-alpha\ndotnet add package Microsoft.Extensions.Configuration.UserSecrets --version 8.0.0<\/code><\/pre>\n
    Run the project directly from Visual Studio Code or with the command:<\/p>\n
    dotnet run<\/code><\/pre>\n
    We will see the standard Weather Forecast API Endpoint.<\/p>\n
    \"WebAPI<\/p>\n
    Note:<\/strong> The WebAPI project uses User Secrets to access the GPT-4o Azure OpenAI Model. Set up the user secrets with the following commands:<\/p>\n
    dotnet user-secrets init\ndotnet user-secrets set \"AZURE_OPENAI_MODEL\" \"< model >\"\ndotnet user-secrets set \"AZURE_OPENAI_ENDPOINT\" \"< endpoint >\"\ndotnet user-secrets set \"AZURE_OPENAI_APIKEY\" \"< api key >\"<\/code><\/pre>\n
    Step 2: Create a prompty for a more descriptive summary for the Forecast<\/h3>\n
    The weather forecast returns a fictitious forecast including dates, temperatures (C and F) and a summary. Our goal is to create a prompt that can generate a more detailed Summary.<\/p>\n
    Let’s add a new prompty file to the root of the PromptyWebAPI<\/code> folder. This is as easy as right-click<\/code> and New Prompty<\/code>. Rename the created file to weatherforecastdesc.prompty<\/code>.<\/p>\n
    Our solution should look like this:<\/p>\n
    \"Solution<\/p>\n
    Now it’s time to complete the sections of our prompty file. Each section has specific information related to the use of the LLM. In example, the model section<\/strong> will define the model to be used, the sample section<\/strong> will provide a sample for the expected output and finally we have the prompt to be used. <\/p>\n
    In the following sample, the prompt defines a System message<\/strong>, and in the Context<\/strong>, we provide the parameters for the weather.<\/p>\n
    Replace the content of your prompty file with the following content.  <\/p>\n
    ---\nname: generate_weather_detailed_description\ndescription: A prompt that generated a detaled description for a weather forecast\nauthors:\n  - Bruno Capuano\nmodel:\n  api: chat\n  configuration:\n    type: azure_openai\n    azure_endpoint: ${env:AZURE_OPENAI_ENDPOINT}\n    azure_deployment: ${env:AZURE_OPENAI_MODEL}\n  parameters:\n    max_tokens: 3000\nsample:\n  today: > \n    2024-07-16\n\n  date: > \n    2024-07-17\n\n  forecastTemperatureC: >\n    25\u00b0C\n---\n\n# System:\nYou are an AI assistant who generated detailed weather forecast descriptions. The detailed description is a paragraph long.\nYou use the full description of the date, including the weekday.\nYou also give a reference to the forecast compared to the current date today.\nAs the assistant, you generate descriptions using a funny style and even add some personal flair with appropriate emojis.\n\n# Context\nUse the following context to generated a detailed weather forecast descriptions \n- Today: {{today}}\n- Date: {{date}}\n- TemperatureC: {{forecastTemperatureC}}<\/code><\/pre>\n
    Once our prompty is ready, we can test the prompt pressing F5<\/strong>. Once we run our prompty, we should see the results in the output window:<\/p>\n
    \"Testing<\/p>\n
    This is the moment to start to refine our prompt!<\/p>\n
    Bonus:<\/strong> Right click on the prompty file, also allow the generation of C# code using Semantic Kernel to use the current file.<\/em><\/p>\n
    \"Generate<\/p>\n
    Note:<\/strong> Previous to this, we need to provide the necessary information for the LLM to be used by prompty. Follow the extension configuration to do this. The easiest way is to create a .env<\/code> file with the LLM information.<\/em><\/p>\n
    Step 3: Update the endpoint using prompty for the forecast summary<\/h3>\n
    Now we can use this prompty, using Semantic Kernel directly in our project. Let’s edit the program.cs<\/code> file and apply the following changes:<\/p>\n