Use Kernel Memory as Retrieval Process

[Skyquek]

I have a NL2SQL use case where I want to register a huge database schema into Kernel Memory. But, I don't want the bot to answer the question directly. I just want to retrieve relevant table and columns name and pass to other agent for SQL generation.

My question is:

1. How can I use Kernel Memory as just retrieval engine?
2. Is it possible for me make this retrieval process to be part of semantic kernel plugin?

Thanks in advance!

Tried the code below, but doesn't work. Its not showing which part of the document get retrieved.

// Check if document is ready
if (await memory.IsDocumentReadyAsync("schema001"))
{
    Console.WriteLine("Hi, The document is ready, you can start asking questions!");

    var readUserInput = Console.ReadLine();

    Func<string, Task> Chat = async (string input) =>
    {
        var searchResult = await memory.SearchAsync(input);
        var answer = searchResult;

        foreach (var result in answer.Results) 
        {
            Console.WriteLine($"Results: " + result.Index);
        }

        //Console.WriteLine($"Question: {input}\n\nAnswer: {searchResult}");

        //Console.WriteLine($"Sources:\n");

        //foreach (var x in answer.RelevantSources)
        //{
        //    Console.WriteLine($" - {x.SourceName} - {x.Link} [{x.Partitions.First().LastUpdate:D}]");
        //}
    };

    await Chat(readUserInput);
}

[vkilohani]

Try something like this inside the chat function.

        var searchResult = await memory.SearchAsync(input, limit: 2, minRelevance: 0.5);
       if (searchResult .Results.Count == 0)
            {
                  return "INFO NOT FOUND"; // or whatever you want
             }
       else
            {
                String answer = "";

                foreach (Citation resultByDoc in searchResult.Results)
                {
                    int i = 1;
                    answer += $"DocumentId: {resultByDoc.DocumentId}\n";
                    foreach (Citation.Partition result in resultByDoc.Partitions)
                    {
                        answer += $"\nChunk {i}: " + $"{result.Text}";
                        i++;
                    }
                }
                return answer;
           }