{"id":2833,"date":"2024-06-12T11:19:24","date_gmt":"2024-06-12T18:19:24","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/semantic-kernel\/?p=2833"},"modified":"2024-06-12T11:19:24","modified_gmt":"2024-06-12T18:19:24","slug":"azure-openai-on-your-data-with-semantic-kernel","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/agent-framework\/azure-openai-on-your-data-with-semantic-kernel\/","title":{"rendered":"Azure OpenAI On Your Data with Semantic Kernel"},"content":{"rendered":"

Hello, everyone!<\/p>\n

In the fast-paced world of technology, using artificial intelligence to get more out of your data is becoming essential. Azure OpenAI on Your Data<\/a> is a feature that helps businesses do just that. It combines the advanced AI models from OpenAI with the strong data management tools of Azure, allowing you to use AI with your own data to gain valuable insights and improve your operations.<\/p>\n

This article shows how to configure Azure OpenAI On Your Data and use it in the code with Semantic Kernel. What makes it powerful is the ability to combine your data, AI and other Semantic Kernel Plugins to achieve your business scenarios.<\/p>\n

Prerequisites<\/strong><\/h2>\n

To execute our code and get information about our data with AI, three services will be required:<\/p>\n

    \n
  1. Azure OpenAI service – to be able to talk to AI. Follow this link to create Azure OpenAI service: Quickstart: Get started using GPT-35-Turbo and GPT-4 with Azure OpenAI Service<\/a>.<\/li>\n
  2. Azure AI Search service – to index our data and make it searchable. Information on how to create Azure AI Search service can be found here: Create an Azure AI Search service in the portal<\/a>.<\/li>\n
  3. Azure Blob storage – to hold our data files. This service is optional for this functionality, but we will use it as an example how to upload, vectorize and index our files on Azure for testing purposes. This link provides information on how to create Azure Storage account and Azure Blob storage: Quickstart: Upload, download, and list blobs with the Azure portal<\/a>.<\/li>\n<\/ol>\n

    Our goal is to combine these services together using following algorithm:<\/p>\n

      \n
    1. Upload our data files to Azure Blob storage.<\/li>\n
    2. Vectorize and index data in Azure AI Search.<\/li>\n
    3. Connect Azure OpenAI service with Azure AI Search.<\/li>\n<\/ol>\n

      Each step can be achieved separately. For simplicity, we will use Add your data<\/em> view in Azure OpenAI Studio, which allows to perform these steps in one place. More information can be found here: Quickstart: Chat with Azure OpenAI models using your own data<\/a>.<\/p>\n

      If you already have your data in Azure AI Search, you can proceed to Accessing your data with Semantic Kernel<\/a> section of this article.<\/p>\n

      Upload data to Azure AI Search<\/strong><\/h2>\n

      Open Azure OpenAI Studio Chat Playground<\/a>, select Add your data<\/em> tab and press Add a data source<\/em> button:\n\"Chat<\/a><\/p>\n

      In the next window, select Subscription, Azure Blob storage, Azure AI Search service. Specify your index name and make sure that Add vector search to this search resource<\/em> is enabled. After enabling this option, new dropdown will appear to select embedding model. This is required to convert our test files to vectors and store them in Azure AI Search:\n\"Add<\/a><\/p>\n

      After clicking Next<\/em> button, it will redirect us to the next step to upload our files. For testing purposes, we will use PDF file from Azure AI Search Sample Data repository<\/a> with specific content, which AI shouldn’t know about by default:<\/p>\n

      Welcome to Contoso Electronics! We are excited to offer our employees two comprehensive health \r\ninsurance plans through Northwind Health. \r\n\r\nNorthwind Health Plus \r\nNorthwind Health Plus is a comprehensive plan that provides comprehensive coverage for medical, vision, and dental services. This plan also offers prescription drug coverage, mental health and substance abuse coverage, and coverage for preventive care services.\r\n\r\nNorthwind Standard \r\nNorthwind Standard is a basic plan that provides coverage for medical, vision, and dental services. This plan also offers coverage for preventive care services, as well as prescription drug coverage.<\/pre>\n
      Let’s upload this PDF file by clicking Upload files<\/em> button. It will be added to specified Azure Blob storage:\n\"Upload<\/a><\/p>\n
      Next window will ask us which Search type to enable. For this example, we are interested in Vector<\/em> search type (there is also Hybrid search type available, which combines vector and full text search):\n\"Data<\/a><\/p>\n
      On the next view we will be asked to provide authentication type. You can use System assigned managed identity<\/em> if your Azure AI Search has RBAC authentication enabled. For this example, we will use API key<\/em> authentication type:\n\"Data<\/a><\/p>\n
      The last step is to review our configuration and proceed with data upload, which will take a couple of minutes. In order to be sure that Azure AI Search index was created, we can go to Azure AI Search resource on Azure portal, open Search management > Indexes<\/em> page and verify if our index exists. In index, we can check if our data is searchable by using Search explorer<\/em>:\n\"Azure<\/a><\/p>\n
       <\/p>\n
      Accessing your data with Semantic Kernel<\/strong><\/h2>\n
      Now, we want to use AI and ask some questions about our data. Let’s create new console application and add Semantic Kernel:<\/p>\n
      dotnet new console -n azure-semantic-kernel-quickstart\r\ndotnet add package Microsoft.SemanticKernel<\/pre>\n
      We will also need following environment variables configured, so we can use it in our application. You can find Azure OpenAI keys and endpoint information on Resource Management > Keys and Endpoint<\/em> page on Azure portal. For Azure AI Search, use following routes on Azure portal:<\/p>\n