semantic caching working

README.md

@@ -342,7 +342,6 @@ To use Azure Cache for Redis with Vector Search, you need to enable the `USE_RED
 ```bash
 azd env set USE_REDIS true
 azd env set EMBEDDING_TYPE 4
-azd env set AZURE_USE_APPLICATION_INSIGHTS true
 azd up
 ```
 

app/Directory.Packages.props

@@ -45,6 +45,8 @@
     <PackageVersion Include="MudBlazor" Version="6.11.1" />
     <PackageVersion Include="PdfSharpCore" Version="1.3.62" />
     <PackageVersion Include="Pinecone.NET" Version="1.3.2" />
+    <PackageVersion Include="Redis.OM" Version="0.7.0" />
+    <PackageVersion Include="Redis.OM.Vectorizers" Version="0.7.0" />
     <PackageVersion Include="StackExchange.Redis" Version="2.7.27" />
     <PackageVersion Include="Swashbuckle.AspNetCore" Version="6.5.0" />
     <PackageVersion Include="System.CommandLine" Version="2.0.0-beta4.22272.1" />

app/backend/MinimalApi.csproj

@@ -26,6 +26,8 @@
     <PackageReference Include="Microsoft.VisualStudio.Azure.Containers.Tools.Targets" />
     <PackageReference Include="PdfSharpCore" />
     <PackageReference Include="Swashbuckle.AspNetCore" />
+    <PackageReference Include="Redis.OM" />
+    <PackageReference Include="Redis.OM.Vectorizers"/>
   </ItemGroup>
 
   <ItemGroup>

app/backend/Services/ReadRetrieveReadChatService.cs

@@ -4,6 +4,8 @@
 using Microsoft.SemanticKernel.ChatCompletion;
 using Microsoft.SemanticKernel.Connectors.OpenAI;
 using Microsoft.SemanticKernel.Embeddings;
+using Redis.OM;
+using Redis.OM.Vectorizers;
 
 namespace MinimalApi.Services;
 #pragma warning disable SKEXP0011 // Mark members as static
@@ -15,6 +17,7 @@ public class ReadRetrieveReadChatService
     private readonly IConfiguration _configuration;
     private readonly IComputerVisionService? _visionService;
     private readonly TokenCredential? _tokenCredential;
+    private readonly Redis.OM.Contracts.ISemanticCache? _semanticCache;
 
     public ReadRetrieveReadChatService(
         ISearchService searchClient,
@@ -54,6 +57,42 @@ public ReadRetrieveReadChatService(
         _configuration = configuration;
         _visionService = visionService;
         _tokenCredential = tokenCredential;
+        if (configuration["UseRedis"] == "true")
+        {
+            var SEredisConnectionString = configuration["AzureCacheServiceEndpoint"] ?? throw new ArgumentNullException("AzureCacheServiceEndpoint");
+            var redisConnectionString = ConvertStackExchangeRedisConnectionStringToRedisCloudConnectionURL(SEredisConnectionString);
+            var provider = new RedisConnectionProvider(redisConnectionString);
+
+            var deploymentId = configuration["AzureOpenAiEmbeddingDeployment"] ?? throw new ArgumentNullException("AzureOpenAiEmbeddingDeployment");
+            var url = configuration["AzureOpenAiServiceEndpoint"] ?? throw new ArgumentNullException("AzureOpenAiServiceEndpoint");
+            var apiKey = configuration["AzureOpenAiServiceKey"] ?? throw new ArgumentNullException("AzureOpenAiServiceKey");
+            int startIndex = url.IndexOf("https://") + "https://".Length;
+            int endIndex = url.IndexOf(".openai.azure.com/");
+
+            var resourceName = string.Empty;
+            if (startIndex >= 0 && endIndex >= 0)
+            {
+                // Extract the desired string
+                resourceName = url.Substring(startIndex, endIndex - startIndex);
+
+            }
+            else
+            {
+                throw new ArgumentNullException("AzureOpenAiServiceEndpoint is not in the correct format");
+            }
+
+            var dim = 1536;
+            _semanticCache = provider.AzureOpenAISemanticCache(apiKey, resourceName, deploymentId, dim);
+        }
+    }
+
+    private string ConvertStackExchangeRedisConnectionStringToRedisCloudConnectionURL(string stackExchangeRedisConnectionString)
+    {
+        var split = stackExchangeRedisConnectionString.Split(',');
+        var host = split[0].Split(':')[0];
+        var port = split[0].Split(':')[1];
+        var password = split[1].Split('=')[1];
+        return $"redis://:{password}=@{host}:{port}";
     }
 
     public async Task<ChatAppResponse> ReplyAsync(
@@ -110,7 +149,7 @@ standard plan AND dental AND employee benefit.
         }
         else
         {
-            documentContents = string.Join("\r", documentContentList.Select(x =>$"{x.Title}:{x.Content}"));
+            documentContents = string.Join("\r", documentContentList.Select(x => $"{x.Title}:{x.Content}"));
         }
 
         // step 2.5
@@ -140,7 +179,7 @@ standard plan AND dental AND employee benefit.
             }
         }
 
-        
+
         if (images != null)
         {
             var prompt = @$"## Source ##
@@ -185,12 +224,31 @@ You answer needs to be a json object with the following format.
             StopSequences = [],
         };
 
-        // get answer
-        var answer = await chat.GetChatMessageContentAsync(
-                       answerChat,
-                       promptExecutingSetting,
-                       cancellationToken: cancellationToken);
-        var answerJson = answer.Content ?? throw new InvalidOperationException("Failed to get search query");
+        string answerJson = string.Empty;
+        if (_semanticCache is not null)
+        {
+            var res = _semanticCache.GetSimilar(question);
+            if (res.Length > 0)
+            {
+                answerJson = res[0].Response;
+            }
+        }
+
+        if (string.IsNullOrEmpty(answerJson))
+        {
+            // get answer
+            var answer = await chat.GetChatMessageContentAsync(
+                           answerChat,
+                           promptExecutingSetting,
+                           cancellationToken: cancellationToken);
+            answerJson = answer.Content ?? throw new InvalidOperationException("Failed to get search query");
+
+            if (_semanticCache is not null)
+            {
+                _semanticCache.Store(question, answerJson);
+            }
+        }
+
         var answerObject = JsonSerializer.Deserialize<JsonElement>(answerJson);
         var ans = answerObject.GetProperty("answer").GetString() ?? throw new InvalidOperationException("Failed to get answer");
         var thoughts = answerObject.GetProperty("thoughts").GetString() ?? throw new InvalidOperationException("Failed to get thoughts");

app/prepdocs/PrepareDocs/AppOptions.cs

@@ -11,6 +11,7 @@ internal record class AppOptions(
     string? TenantId,
     string? SearchServiceEndpoint,
     string? AzureOpenAIServiceEndpoint,
+    string? AzureOpenAiServiceKey,
     string? SearchIndexName,
     string? EmbeddingModelName,
     bool Remove,

app/prepdocs/PrepareDocs/Program.Options.cs

@@ -29,6 +29,10 @@ internal static partial class Program
     private static readonly Option<string> s_azureOpenAIService =
         new(name: "--openaiendpoint", description: "Optional. The Azure OpenAI service endpoint which will be used to extract text, tables and layout from the documents (must exist already)");
 
+    private static readonly Option<string> s_azureOpenAIServiceKey =
+        new(name: "--openaiendpointkey", description: "Optional. The Azure OpenAI service endpoint key");
+
+
     private static readonly Option<string> s_embeddingModelName =
         new(name: "--embeddingmodel", description: "Optional. Name of the Azure Cognitive Search embedding model to use for embedding content in the search index (will be created if it doesn't exist)");
 
@@ -68,6 +72,7 @@ internal static partial class Program
             s_searchService,
             s_searchIndexName,
             s_azureOpenAIService,
+            s_azureOpenAIServiceKey,
             s_embeddingModelName,
             s_remove,
             s_removeAll,
@@ -88,6 +93,7 @@ internal static partial class Program
             SearchServiceEndpoint: context.ParseResult.GetValueForOption(s_searchService),
             SearchIndexName: context.ParseResult.GetValueForOption(s_searchIndexName),
             AzureOpenAIServiceEndpoint: context.ParseResult.GetValueForOption(s_azureOpenAIService),
+            AzureOpenAiServiceKey: context.ParseResult.GetValueForOption(s_azureOpenAIServiceKey),
             EmbeddingModelName: context.ParseResult.GetValueForOption(s_embeddingModelName),
             Remove: context.ParseResult.GetValueForOption(s_remove),
             RemoveAll: context.ParseResult.GetValueForOption(s_removeAll),

infra/core/ai/cognitiveservices.bicep

@@ -53,3 +53,4 @@ resource deployment 'Microsoft.CognitiveServices/accounts/deployments@2023-05-01
 output endpoint string = account.properties.endpoint
 output id string = account.id
 output name string = account.name
+output apiKey string = account.listKeys().key1

infra/main.bicep

@@ -288,6 +288,10 @@ module keyVaultSecrets 'core/security/keyvault-secrets.bicep' = {
         name: 'AzureOpenAiEmbeddingDeployment'
         value: azureEmbeddingDeploymentName
       }
+      {
+        name: 'AzureOpenAiServiceKey'
+        value: azureOpenAi.outputs.apiKey
+      }
     ] : [
       {
         name: 'OpenAIAPIKey'
@@ -797,6 +801,7 @@ output AZURE_LOCATION string = location
 output AZURE_OPENAI_RESOURCE_LOCATION string = openAiResourceGroupLocation
 output AZURE_OPENAI_CHATGPT_DEPLOYMENT string = azureChatGptDeploymentName
 output AZURE_OPENAI_EMBEDDING_DEPLOYMENT string = azureEmbeddingDeploymentName
+output AZURE_OPENAI_EMBEDDING_API_KEY string = useAOAI? azureOpenAi.outputs.apiKey : ''
 output AZURE_OPENAI_ENDPOINT string = useAOAI? azureOpenAi.outputs.endpoint : ''
 output AZURE_OPENAI_RESOURCE_GROUP string = useAOAI ? azureOpenAiResourceGroup.name : ''
 output AZURE_OPENAI_SERVICE string = useAOAI ? azureOpenAi.outputs.name : ''
@@ -805,10 +810,10 @@ output AZURE_SEARCH_INDEX string = searchIndexName
 output AZURE_SEARCH_SERVICE string = searchService.outputs.name
 output AZURE_SEARCH_SERVICE_ENDPOINT string = searchService.outputs.endpoint
 output AZURE_SEARCH_SERVICE_RESOURCE_GROUP string = searchServiceResourceGroup.name
-output AZURE_CACHE_INDEX string = azureCacheIndexName
-output AZURE_CACHE_SERVICE string = azureCache.outputs.name
-output AZURE_CACHE_SERVICE_ENDPOINT string = azureCache.outputs.endpoint
-output AZURE_CACHE_SERVICE_RESOURCE_GROUP string = azureCacheResourceGroup.name
+output AZURE_CACHE_INDEX string = useRedis ? azureCacheIndexName : ''
+output AZURE_CACHE_SERVICE string = useRedis ? azureCache.outputs.name : ''
+output AZURE_CACHE_SERVICE_ENDPOINT string = useRedis ? azureCache.outputs.endpoint : ''
+output AZURE_CACHE_SERVICE_RESOURCE_GROUP string = useRedis ? azureCacheResourceGroup.name : ''
 output USE_REDIS bool = useRedis
 output AZURE_STORAGE_ACCOUNT string = storage.outputs.name
 output AZURE_STORAGE_BLOB_ENDPOINT string = storage.outputs.primaryEndpoints.blob

scripts/prepdocs.ps1

@@ -69,6 +69,7 @@ if ([string]::IsNullOrEmpty($env:AZD_PREPDOCS_RAN) -or $env:AZD_PREPDOCS_RAN -eq
     if ($env:USE_AOAI -eq "true") {
         Write-Host "Using Azure OpenAI"
         $dotnetArguments += " --openaiendpoint $($env:AZURE_OPENAI_ENDPOINT) "
+        $dotnetArguments += " --openaiendpointkey $($env:AZURE_OPENAI_EMBEDDING_API_KEY) "
         $dotnetArguments += " --embeddingmodel $($env:AZURE_OPENAI_EMBEDDING_DEPLOYMENT) "
     }
     else{

scripts/prepdocs.sh

@@ -37,6 +37,7 @@ if [ -z "$AZD_PREPDOCS_RAN" ] || [ "$AZD_PREPDOCS_RAN" = "false" ]; then
         echo "use azure openai"
         args="$args --openaiendpoint $AZURE_OPENAI_ENDPOINT"
         args="$args --embeddingmodel $AZURE_OPENAI_EMBEDDING_DEPLOYMENT"
+        args="$args --openaiendpointkey $AZURE_OPENAI_EMBEDDING_API_KEY"
     else
         echo "use openai"
         args="$args --embeddingmodel $OPENAI_EMBEDDING_DEPLOYMENT"