Bug: Plugin created by VectorStoreTextSearch.CreateWithGetTextSearchResults is not returning the results from the IVectorStore

[thomashauser]

Describe the bug
The method VectorizableTextSearchAsync is returning to document chunks, but the plugin said it is empty.

To Reproduce
ITextEmbeddingGenerationService textEmbeddingGenerationService = new OpenAITextEmbeddingGenerationService("text-embedding-small", openAiClient);
IVectorStore vectorStore = new SqlServerVectorStore(connectionString)
.UseTextEmbeddingGenerationService(textEmbeddingGenerationService);

var globalVectorStoreRecordCollection = vectorStore.GetCollection<int, DocumentRecord>("Global");
await globalVectorStoreRecordCollection.CreateCollectionIfNotExistsAsync();
var chatVectorStoreRecordCollection = vectorStore.GetCollection<int, DocumentRecord>($"Chat{chatId}");
await chatVectorStoreRecordCollection.CreateCollectionIfNotExistsAsync();

var globalSearch = new VectorStoreTextSearch(globalVectorStoreRecordCollection, textEmbeddingGenerationService);
var chatSearch = new VectorStoreTextSearch(chatVectorStoreRecordCollection, textEmbeddingGenerationService);

var builder = Kernel.CreateBuilder();
builder.AddAzureOpenAIChatCompletion(MODEL, azureOpenAIClient: openAiClient);
builder.Services.AddLogging(services => services.AddConsole().SetMinimumLevel(LogLevel.Trace));

var kernel = builder.Build();

// Add the plugin to the kernel
kernel.Plugins.Add(globalSearch.CreateWithGetTextSearchResults(nameof(globalSearch)));
kernel.Plugins.Add(chatSearch.CreateWithGetTextSearchResults(nameof(chatSearch)));

// Retrieve the chat completion service
var chatCompletionService = kernel.Services.GetRequiredService();

Expected behavior
The kernel should receive the two found document chunks.

Screenshots

Platform

• OS: Windows
• IDE: Visual Studio 2022 17.12.0
• Language: C#
• Azure OpenAI: gpt-4o (2024-08-06), text-embedding-3-small
• Source:
  Azure.AI.DocumentIntelligence: 1.0.0-beta.3
  Microsoft.Extensions.Configuration.Json: 9.0.0
  Microsoft.Extensions.Configuration.UserSecrets: 9.0.0
  Microsoft.Extensions.DependencyInjection: 9.0.0
  Microsoft.Extensions.Logging.Console: 9.0.0
  Microsoft.Extensions.VectorData.Abstractions: 9.0.0-preview.1.24523.1
  Microsoft.SemanticKernel: 1.29.0
  Microsoft.SemanticKernel.Connectors.SqlServer: 1.29.0-alpha
  Microsoft.SemanticKernel.Plugins.Core: 1.29.0-alpha
  Microsoft.SemanticKernel.Plugins.Document: 1.29.0-alpha
  Microsoft.SemanticKernel.Plugins.Memory: 1.29.0-alpha
  Microsoft.Data.SqlClient: 5.2.2
  Microsoft.SemanticKernel.Abstractions: 1.29.0
  Microsoft.SemanticKernel.Connectors.AzureOpenAI: 1.29.0
  Microsoft.SemanticKernel.Connectors.OpenAI: 1.29.0
  System.Linq.Async: 6.0.1

[markwallace-microsoft]

Hi @thomashauser
we have this sample VectorStore_TextSearch.cs which shows this working with our in-memory store. Can you reproduce the error by changing this sample?

Did you create your own SqlServerVectorStore implementation?

[inQuantum]

Objects from VectorSearch Results should be transformed/mapped to TextSearchResult. This should solve issue.

var result = new VectorStoreTextSearch<DataModel>(vectorStoreRecordCollection, textEmbeddingGeneration, stringMapper, resultMapper);


protected sealed class DataModelTextSearchStringMapper : ITextSearchStringMapper
{
    /// <inheritdoc />
    public string MapFromResultToString(object result)
    {
        if (result is DataModel dataModel)
        {
            return dataModel.Text;
        }
        throw new ArgumentException("Invalid result type.");
    }
}

/// <summary>
/// Result mapper which converts a DataModel to a TextSearchResult.
/// </summary>
protected sealed class DataModelTextSearchResultMapper : ITextSearchResultMapper
{
    /// <inheritdoc />
    public TextSearchResult MapFromResultToTextSearchResult(object result)
    {
        if (result is DataModel dataModel)
        {
            return new TextSearchResult(value: dataModel.Text) { Name = dataModel.Key.ToString(), Link = dataModel.Link };
        }
        throw new ArgumentException("Invalid result type.");
    }
}

[thomashauser]

Hi @thomashauser
we have this sample VectorStore_TextSearch.cs which shows this working with our in-memory store. Can you reproduce the error by changing this sample?

Did you create your own SqlServerVectorStore implementation?

Yes, it's my own implementation. There is currently no version that supports IVectoreStore and not the deprecated kernel memory, right?

[thomashauser]

Objects from VectorSearch Results should be transformed/mapped to TextSearchResult. This should solve issue.

var result = new VectorStoreTextSearch<DataModel>(vectorStoreRecordCollection, textEmbeddingGeneration, stringMapper, resultMapper);


protected sealed class DataModelTextSearchStringMapper : ITextSearchStringMapper
{
    /// <inheritdoc />
    public string MapFromResultToString(object result)
    {
        if (result is DataModel dataModel)
        {
            return dataModel.Text;
        }
        throw new ArgumentException("Invalid result type.");
    }
}

/// <summary>
/// Result mapper which converts a DataModel to a TextSearchResult.
/// </summary>
protected sealed class DataModelTextSearchResultMapper : ITextSearchResultMapper
{
    /// <inheritdoc />
    public TextSearchResult MapFromResultToTextSearchResult(object result)
    {
        if (result is DataModel dataModel)
        {
            return new TextSearchResult(value: dataModel.Text) { Name = dataModel.Key.ToString(), Link = dataModel.Link };
        }
        throw new ArgumentException("Invalid result type.");
    }
}

Awesome that worked. Thanks