Merge pull request #28 from Azure/websearch

Lesson 4 - Bing Web Search plugin
Azure · Jan 15, 2025 · bbe554e · bbe554e
2 parents 579ec7b + 2a81c1b
commit bbe554e
Show file tree

Hide file tree

Showing 23 changed files with 368 additions and 31 deletions.
diff --git a/docs/wksp/05-semantic-kernel-workshop/index.md b/docs/wksp/05-semantic-kernel-workshop/index.md
@@ -3,4 +3,4 @@
 This workshop is broken into 2 main sections:
 
 1. [Create simple AI Orchestration using Sematic Kernel](simple-ai-orchestration/index.md)
-1. [Create advanced AI Orchestration using Sematik Kernel with agents and planners](advanced-ai-orchestration/index.md)
+1. [Create advanced AI Orchestration using Semantic Kernel with agents and planners](advanced-ai-orchestration/index.md)
diff --git a/...ntic-kernel-workshop/simple-ai-orchestration/create-deployable-app/azd-infra.md b/...ntic-kernel-workshop/simple-ai-orchestration/create-deployable-app/azd-infra.md
@@ -54,6 +54,7 @@ use `bicep` templates which are organized as follows:
       * `web.bicep` - bicep templated for web application infrastructure
     * core - subdirectory with templates for core infrastructure components
       * **ai** - subdirectory for AI related components
+      * **bing** - subdirectory for Bing Search service
       * **host** - subdirectory for container app, environment and registry components
       * **monitor** - subdirectory for monitoring components (e.g. application insights)
       * **security** - subdirectory for security components (e.g. keyvault)
@@ -67,19 +68,20 @@ to generate an existing template is to find a template that uses similar compone
 
 You can build, provision all resources and deploy by following these steps:
 
-1. Switch to `workshop/donet` directory.
+1. Switch to `workshop/dotnet` directory.
 1. Ensure Docker desktop is running (if not using Github Codespace).
-1. Run `azd auth login` to login to your Azure account.
-1. Run `azd up` to provision Azure resources and deploy this sample to those resources.
-   You will be prompted for the following parameters:
-    * Environment name: sk-test
-    * Select an Azure subscription to use from list
-    * Select an Azure location to use: e.g. (US) East US 2 (eastus2)
-    * Enter a value for the infrastructure parameters:
-      * **openAIApiKey**
-      * **openAiChatGptDeployment**: e.g. gpt-4o
-      * **openAiEndpoint**
-      * **stockServiceApiKey**
+1. Open Git bash or WSL (Windows Subsystem for Linux) terminal and run the following commands
+    1. Run `azd auth login` to login to your Azure account.
+    1. Run `azd up` to provision Azure resources and deploy this sample to those resources.
+      You will be prompted for the following parameters:
+        * Environment name: sk-test
+        * Select an Azure subscription to use from list
+        * Select an Azure location to use: e.g. (US) East US 2 (eastus2)
+        * Enter a value for the infrastructure parameters:
+          * **openAIApiKey**
+          * **openAiChatGptDeployment**: e.g. gpt-4o
+          * **openAiEndpoint**
+          * **stockServiceApiKey**
 1. After the application has been successfully deployed you will see the API and Web Service URLs printed in the console.  
    Click the Web Service URL to interact with the application in your browser.
 

diff --git a/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/index.md b/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/index.md
@@ -4,9 +4,10 @@ This hands on workshop goes through the following lessons for creating a simple
 chatbot as a console application:
 
 1. [Prerequisites](pre-reqs.md)
-1. [Lesson 1: Create Simple Semantic Kernel chatbot](lesson1.md)
-1. [Lesson 2: Create Simple Semantic Kernel chatbot with history](lesson2.md)
-1. [Lesson 3: Create Simple Semantic Kernel chatbot with plugins](lesson3.md)
+1. [Lesson 1: Semantic Kernel chatbot](lesson1.md)
+1. [Lesson 2: Semantic Kernel chatbot with history](lesson2.md)
+1. [Lesson 3: Semantic Kernel chatbot with plugins](lesson3.md)
+1. [Lesson 4: Semantic Kernel chatbot with Web Search engine plugin](lesson4.md)
 
 Subsequently the following lessons guide you through converting that chatbot application
 into a deployable application with a backend API and Web UI frontend and deploying into Azure:

diff --git a/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson1.md b/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson1.md
@@ -1,4 +1,4 @@
-# Lesson 1: Simple Semantic Kernel chatbot
+# Lesson 1: Semantic Kernel chatbot
 
 In this lesson we will create a semantic kernel chatbot with a system prompt and keeping track of chat history.
 

diff --git a/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson2.md b/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson2.md
@@ -1,4 +1,4 @@
-# Lesson 2: Simple Semantic Kernel chatbot with history
+# Lesson 2: Semantic Kernel chatbot with history
 
 In this lesson we will add chat history to our chat agent.
 

diff --git a/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson3.md b/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson3.md
@@ -1,4 +1,4 @@
-# Lesson 3: Simple Semantic Kernel chat agent with plugins
+# Lesson 3: Semantic Kernel chatbot with plugins
 
 In this lesson we will a semantic kernel plugins to be able to retrieve stock pricing.
 

diff --git a/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson4.md b/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson4.md
@@ -0,0 +1,84 @@
+# Lesson 4: Semantic Kernel chatbot with Web Search engine plugin
+
+In this lesson we will add a Web Search Engine plugin that uses Bing Search to our semantic kernel chatbot.
+
+1. Ensure all [pre-requisites](pre-reqs.md) are met and installed (including updating the `BingSearchService` `apiKey` value in the `appSettings.json` file using the key from **Bing Search Service v7** in Azure (https://portal.azure.com)).
+
+1. Switch to Lesson 4 directory:
+
+    ```bash
+    cd ../Lesson4
+    ```
+
+1. Start by copying `appsettings.json` from Lesson 1:
+
+    ```bash
+    cp ../Lesson1/appsettings.json .
+    ```
+
+1. Run program and ask what the sentiment on Microsoft stock is:
+
+    ```bash
+    dotnet run
+    ```
+
+   At the prompt enter
+
+   ```bash
+   What is the sentiment on Microsoft stock?
+    ```
+
+   Assistant will give a generic response:
+
+   ```txt
+   Assistant > The sentiment on Microsoft (ticker symbol: MSFT) largely hinges on factors like:
+
+    - Tech innovation (e.g., AI, Azure cloud service, and gaming)
+    - Quarterly earnings reports
+    - Overall market conditions
+    - How much caffeine traders have consumed
+   ```
+
+1. Notice it does not provide a specific answer. We can add the Web Search Engine plugin to be able to provide a better answer.
+
+1. Next locate **TODO: Step 1** in `Program.cs` and add the following import lines:
+
+    ```csharp
+    using Microsoft.SemanticKernel.Plugins.Web;
+    using Microsoft.SemanticKernel.Plugins.Web.Bing;
+    ```
+
+1. Next locate **TODO: Step 2** in `Program.cs` and provide the following lines to initialize and register the `WebSearchEnginePlugin`:
+
+    ```csharp
+    var bingApiKey = AISettingsProvider.GetSettings().BingSearchService.ApiKey;
+    if (!string.IsNullOrEmpty(bingApiKey))
+    {
+        var bingConnector = new BingConnector(bingApiKey);
+        var bing = new WebSearchEnginePlugin(bingConnector);
+        kernel.ImportPluginFromObject(bing, "bing");
+    }
+    ```
+
+1. Re-run the program and ask what the sentiment on Microsoft stock is:
+
+    ```bash
+    dotnet run
+    User > What is the sentiment on Microsoft stock?
+    ```
+
+    Assistant response:
+
+    ```txt
+    Assistant > Ah, Microsoft stock seems to be the belle of the ball with mixed but leaning-positive vibes. Let me serve up the sentiment soup:
+
+    - **Positives:** There's more excitement than usual, with higher-than-average media sentiment compared to other tech companies. Plus, analysts are dishing out 12-month price targets like cocktails, with an average around $489.55 and some saying it could go as high as $600—a nice little pie in the sky for investors! 🍰🚀
+
+    - **Concerns:** Clouds (pun intended!) aren't all silver-lining for Microsoft, as they've got capacity constraints in their cloud services. Choppy waters ahead, perhaps? 🌩️
+
+    - **Buzz:** Oh, it's trending alright! Tons of people are searching, sharing, and probably debating MSFT more than their weekend plans.
+
+    Feeling FOMO or ready to YOLO-invest? Don’t forget to watch that ticker—MSFT!    
+    ```
+
+Expect to see a more specific response. With the Web Search Engine plugin, you can now tap into any web search, so your agent will leverage that plugin to find information not available via other plugins or within the LLM being used.
diff --git a/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/pre-reqs.md b/docs/wksp/05-semantic-kernel-workshop/simple-ai-orchestration/pre-reqs.md
@@ -95,6 +95,28 @@ cd intelligent-app-workshop
         1. Update `appsettings.json` deploymentName field with your model deployment name.
         1. Use the **Deployment Name** value (e.g. gpt-4o) as the **deploymentName** value within the `OpenAI` element in the `appsettings.json` file.
 
+1. Create Azure Bing Search v7 Service in Azure and retrieve the API Key to update newly created `appsettings.json`. This value will be needed for [Lesson 4](lesson4.md).
+    1. Create Bing Search Service (from Azure Portal):
+        1. Go to the [Azure Portal](https://portal.azure.com).
+        1. Click on [Create A Resource](https://ms.portal.azure.com/#create/hub)
+        1. On the search bar type **Bing Search** and hit enter
+        1. Locate **Bing Search v7** and click **Create**
+        1. On the **Create a Bing Search Resource** page, provide the following information for the fields on the Basics tab:
+            * Subscription: The Azure subscription to used for your service.
+            * Resource group: The Azure resource group to contain your Bing service resource. You can create a new group or use a pre-existing group.
+            * Name: A descriptive and unique name for your Bing Search Service resource, such as `bing-search-myid`.
+            * Region: Global (default).
+            * Pricing Tier: The pricing tier for the resource. The `F1` tier is free and recommended for testing purposes.
+        1. Click **Next**.
+        1. On the **Tags** tab click **Next**
+        1. Click **Create**.
+    1. Get the Bing Search API Key and update `appsettings.json` file:
+        1. Go to the [Azure Portal](https://portal.azure.com).
+        1. On the search bar type **Bing Reources** and hit enter
+        1. Locate the **Bing Service** created above and click on it.
+        1. Expand **RESOURCE MANAGEMENT** and click on **Keys and Endpoint**
+        1. Copy the value from **Key 1** and paste it as the **ApiKey** value within the `BingSearchService` element in the `appsettings.json` file.
+
 1. Additionally, we need to obtain an API Key to be able to get stock prices from [polygon.io](https://polygon.io/dashboard/login). You can sign up for a free API Key by creating a login. This value will be needed for [Lesson 3](lesson3.md).
     1. Once logged in, from the [polygon.io Dashboard](https://polygon.io/dashboard) locate the **Keys** section. Copy the default key value and paste it as the **apiKey** value within the `StockService` element in the `appsettings.json` file.
 

diff --git a/mkdocs.yml b/mkdocs.yml
@@ -21,9 +21,10 @@ nav:
   - Module 5 - Hands-on AI Orchestration using Semantic Kernel Workshop: 
       - wksp/05-semantic-kernel-workshop/simple-ai-orchestration/index.md
       - Pre-requisites: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/pre-reqs.md
-      - Lesson 1 - Simple Semantic Kernel chatbot: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson1.md
-      - Lesson 2 - Simple Semantic Kernel chatbot with history: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson2.md
-      - Lesson 3 - Simple Semantic Kernel chatbot with plugins: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson3.md
+      - Lesson 1 - Semantic Kernel chatbot: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson1.md
+      - Lesson 2 - Semantic Kernel chatbot with history: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson2.md
+      - Lesson 3 - Semantic Kernel chatbot with plugins: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson3.md
+      - Lesson 4 - Semantic Kernel chatbot with Web Search engine plugin: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/lesson4.md
       - Build a deployable web application:
         - wksp/05-semantic-kernel-workshop/simple-ai-orchestration/create-deployable-app/index.md
         - Create Backend API: wksp/05-semantic-kernel-workshop/simple-ai-orchestration/create-deployable-app/backend-api.md

diff --git a/workshop/dotnet/App/backend/Extensions/ServiceExtensions.cs b/workshop/dotnet/App/backend/Extensions/ServiceExtensions.cs
@@ -5,6 +5,9 @@
 // Add import required for StockService
 using Core.Utilities.Services;
 using Microsoft.SemanticKernel;
+// Add imports for Bing Search plugin
+using Microsoft.SemanticKernel.Plugins.Web;
+using Microsoft.SemanticKernel.Plugins.Web.Bing;
 
 namespace Extensions;
 
@@ -27,6 +30,15 @@ public static void AddSkServices(this IServiceCollection services)
             StockDataPlugin stockDataPlugin = new(new StocksService(httpClient));
             kernel.Plugins.AddFromObject(stockDataPlugin);
 
+            // Initialize Bing Search plugin
+            var bingApiKey = AISettingsProvider.GetSettings().BingSearchService.ApiKey;
+            if (!string.IsNullOrEmpty(bingApiKey))
+            {
+                var bingConnector = new BingConnector(bingApiKey);
+                var bing = new WebSearchEnginePlugin(bingConnector);
+                kernel.ImportPluginFromObject(bing, "bing");
+            }
+
             return kernel;
         });
     }

diff --git a/workshop/dotnet/Core.Utilities/Config/AISettingsProvider.cs b/workshop/dotnet/Core.Utilities/Config/AISettingsProvider.cs
@@ -5,9 +5,9 @@
 
 namespace Core.Utilities.Config;
 
-internal static class AISettingsProvider
+public static class AISettingsProvider
 {
-    internal static AppSettings GetSettings()
+    public static AppSettings GetSettings()
     {
         IConfigurationRoot config = new ConfigurationBuilder()
             .AddJsonFile("appsettings.json", optional: true)

diff --git a/workshop/dotnet/Core.Utilities/Core.Utilities.csproj b/workshop/dotnet/Core.Utilities/Core.Utilities.csproj
@@ -1,6 +1,6 @@
 <Project Sdk="Microsoft.NET.Sdk">
   <PropertyGroup>
-		<NoWarn>SKEXP0110;SKEXP0001;SKEXP0101</NoWarn>
+    <NoWarn>SKEXP0110;SKEXP0001;SKEXP0101</NoWarn>
 	</PropertyGroup>
 
   <ItemGroup>
@@ -14,5 +14,6 @@
     <PackageReference Include="Microsoft.Extensions.Logging.Console" />
     <PackageReference Include="Microsoft.SemanticKernel" />
     <PackageReference Include="Microsoft.SemanticKernel.Agents.Core" />
+    <PackageReference Include="Microsoft.SemanticKernel.Plugins.Web" />
   </ItemGroup>
 </Project>
diff --git a/workshop/dotnet/Core.Utilities/Models/AppSettings.cs b/workshop/dotnet/Core.Utilities/Models/AppSettings.cs
@@ -4,7 +4,8 @@ namespace Core.Utilities.Models;
 
 public record AppSettings (
     OpenAI OpenAI,
-    StockService StockService
+    StockService StockService,
+    BingSearchService BingSearchService
 );
 
 public record OpenAI (
@@ -15,4 +16,8 @@ string ApiKey
 
 public record StockService (
     string ApiKey
+);
+
+public record BingSearchService (
+    string ApiKey
 );
diff --git a/workshop/dotnet/Core.Utilities/Plugins/StockDataPlugin.cs b/workshop/dotnet/Core.Utilities/Plugins/StockDataPlugin.cs
@@ -1,5 +1,4 @@
 using Core.Utilities.Services;
-using Core.Utilities.Models;
 using Core.Utilities.Extensions;
 
 using Microsoft.SemanticKernel;

diff --git a/workshop/dotnet/Directory.Build.props b/workshop/dotnet/Directory.Build.props
@@ -4,7 +4,7 @@
         <TargetFramework>net8.0</TargetFramework>
         <ImplicitUsings>enable</ImplicitUsings>
         <Nullable>enable</Nullable>
-		<NoWarn>SKEXP0110;SKEXP0001;SKEXP0101</NoWarn>
+		<NoWarn>SKEXP0110;SKEXP0001;SKEXP0101;SKEXP0050</NoWarn>
         <ManagePackageVersionsCentrally>true</ManagePackageVersionsCentrally>
 	</PropertyGroup>
 

diff --git a/workshop/dotnet/Directory.Packages.props b/workshop/dotnet/Directory.Packages.props
@@ -12,6 +12,7 @@
     <PackageVersion Include="Microsoft.Extensions.DependencyInjection" Version="8.0.1" />
     <PackageVersion Include="Microsoft.SemanticKernel" Version="1.32.0" />
     <PackageVersion Include="Microsoft.SemanticKernel.Agents.Core" Version="1.19.0-alpha" />
+    <PackageVersion Include="Microsoft.SemanticKernel.Plugins.Web" Version="1.32.0-alpha" />
     <PackageVersion Include="Swashbuckle.AspNetCore" Version="6.8.1" />
     <PackageVersion Include="xunit" Version="2.4.1" />
     <PackageVersion Include="xunit.runner.visualstudio" Version="2.4.3" />

diff --git a/workshop/dotnet/Lessons/Lesson4/Lesson4.csproj b/workshop/dotnet/Lessons/Lesson4/Lesson4.csproj
@@ -0,0 +1,8 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+    <ItemGroup>
+        <None Update="appsettings.json">
+            <CopyToOutputDirectory>Always</CopyToOutputDirectory>
+        </None>
+    </ItemGroup>
+</Project>
diff --git a/workshop/dotnet/Lessons/Lesson4/Program.cs b/workshop/dotnet/Lessons/Lesson4/Program.cs
@@ -0,0 +1,74 @@
+using Core.Utilities.Config;
+// Add import for Plugins
+using Core.Utilities.Plugins;
+// Add import required for StockService
+using Core.Utilities.Services;
+using Microsoft.SemanticKernel;
+using Microsoft.SemanticKernel.Connectors.OpenAI;
+// TODO: Step 1 - Add imports for Bing Search plugin
+
+// Add ChatCompletion import
+using Microsoft.SemanticKernel.ChatCompletion;
+// Temporarily added to enable Semantic Kernel tracing
+using Microsoft.Extensions.DependencyInjection;
+using Microsoft.Extensions.Logging;
+
+
+// Initialize the kernel with chat completion
+IKernelBuilder builder = KernelBuilderProvider.CreateKernelWithChatCompletion();
+// Enable tracing
+//builder.Services.AddLogging(services => services.AddConsole().SetMinimumLevel(LogLevel.Trace));
+Kernel kernel = builder.Build();
+
+// Initialize Time plugin and registration in the kernel
+kernel.Plugins.AddFromObject(new TimeInformationPlugin());
+
+// Initialize Stock Data Plugin and register it in the kernel
+HttpClient httpClient = new();
+StockDataPlugin stockDataPlugin = new(new StocksService(httpClient));
+kernel.Plugins.AddFromObject(stockDataPlugin);
+
+// TODO: Step 2 - Initialize Bing Search plugin
+
+
+// Get chatCompletionService and initialize chatHistory with system prompt
+var chatCompletionService = kernel.GetRequiredService<IChatCompletionService>();
+ChatHistory chatHistory = new("You are a friendly financial advisor that only emits financial advice in a creative and funny tone");
+// Remove the promptExecutionSettings and kernelArgs initialization code
+// Add system prompt
+OpenAIPromptExecutionSettings promptExecutionSettings = new()
+{
+    // Add Auto invoke kernel functions as the tool call behavior
+    ToolCallBehavior = ToolCallBehavior.AutoInvokeKernelFunctions
+};
+
+// Initialize kernel arguments
+KernelArguments kernelArgs = new(promptExecutionSettings);
+
+// Execute program.
+const string terminationPhrase = "quit";
+string? userInput;
+do
+{
+    Console.Write("User > ");
+    userInput = Console.ReadLine();
+
+    if (userInput != null && userInput != terminationPhrase)
+    {
+        Console.Write("Assistant > ");
+        // Initialize fullMessage variable and add user input to chat history
+        string fullMessage = "";
+        chatHistory.AddUserMessage(userInput);
+
+        // Step 4 - Provide promptExecutionSettings and kernel arguments
+        await foreach (var chatUpdate in chatCompletionService.GetStreamingChatMessageContentsAsync(chatHistory, promptExecutionSettings, kernel))
+        {
+            Console.Write(chatUpdate.Content);
+            fullMessage += chatUpdate.Content ?? "";
+        }
+        chatHistory.AddAssistantMessage(fullMessage);
+
+        Console.WriteLine();
+    }
+}
+while (userInput != terminationPhrase);
diff --git a/workshop/dotnet/Solutions/Lesson4/Lesson4.csproj b/workshop/dotnet/Solutions/Lesson4/Lesson4.csproj
@@ -0,0 +1,8 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+    <ItemGroup>
+        <None Update="appsettings.json">
+            <CopyToOutputDirectory>Always</CopyToOutputDirectory>
+        </None>
+    </ItemGroup>
+</Project>