I want to integrate AI Search into Copilot Studio Bot , but copilot studio expect AI Search to be vector. Still i'm using REST API's for Sharepoint online . Can you guide me how can i create vector search for all documents

Question

Hi Team,

I want to integrate AI Search into Copilot Studio Bot , but copilot studio expect AI Search to be vector. Still i'm using REST API's for Sharepoint online .

Can you guide me how can i create vector search for all documents ?

Please provide code snippet to work around it .

Also , guide me how to re sync the index when a new / updated document for this i need to prefer to Power Automate / Azure Functions on Sharepoint data source.

Please guide me.

Thanks,

Balaji

Answer 1

Hi @Balaji Mogadali since you're using SharePoint Online and REST APIs, you'll need to generate embeddings (vectors) for your SharePoint documents and then index them in Azure AI Search. Here's a loose guide to help you get started:

1. Generating Embeddings (Vectors):

You'll need an embedding model to convert your document content into vectors. Azure OpenAI's embeddings models (like text-embedding-ada-002) are a good choice. You can use the Azure OpenAI REST API or SDK to generate embeddings.

Here's a C# code snippet using the Azure.AI.OpenAI NuGet package:

using Azure;
using Azure.AI.OpenAI;
using System;
using System.Collections.Generic;
using System.Net.Http;
using System.Net.Http.Headers;
using System.Text;
using System.Text.Json;
using System.Threading.Tasks;

public class EmbeddingGenerator
{
    private readonly OpenAIClient _openAIClient;

    public EmbeddingGenerator(string endpoint, string key)
    {
        _openAIClient = new OpenAIClient(new Uri(endpoint), new AzureKeyCredential(key));
    }

    public async Task<List<float>> GetEmbeddingsAsync(string text)
    {
        try
        {
            EmbeddingsOptions embeddingsOptions = new EmbeddingsOptions(text);
            Response<Embeddings> embeddingsResponse = await _openAIClient.GetEmbeddingsAsync("text-embedding-ada-002", embeddingsOptions);
            return embeddingsResponse.Value.Data[0].Embedding;
        }
        catch (RequestFailedException ex)
        {
            Console.WriteLine($"Error generating embeddings: {ex.Message}");
            return null; // Or handle the error as needed
        }
    }


    public static async Task<string> GetSharePointFileContentAsync(string siteUrl, string relativeUrl, string accessToken)
    {
        using (HttpClient client = new HttpClient())
        {
            client.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", accessToken);

            string fileUrl = $"{siteUrl}/_api/web/GetFileByServerRelativeUrl('{relativeUrl}')/$value";

            try
            {
                HttpResponseMessage response = await client.GetAsync(fileUrl);
                response.EnsureSuccessStatusCode(); // Throw if not successful
                return await response.Content.ReadAsStringAsync();
            }
            catch (HttpRequestException ex)
            {
                Console.WriteLine($"Error getting SharePoint file: {ex.Message}");
                return null;
            }
        }
    }
}

// Example usage:
public static async Task Main(string[] args)
{

    string openAiEndpoint = Environment.GetEnvironmentVariable("OPENAI_ENDPOINT");
    string openAiKey = Environment.GetEnvironmentVariable("OPENAI_API_KEY");
    string sharepointSiteUrl = Environment.GetEnvironmentVariable("SHAREPOINT_SITE_URL");
    string sharepointRelativeUrl = "/sites/yoursite/Shared Documents/yourdocument.txt";
    string accessToken = Environment.GetEnvironmentVariable("SHAREPOINT_ACCESS_TOKEN");

    var generator = new EmbeddingGenerator(openAiEndpoint, openAiKey);

    string fileContent = await EmbeddingGenerator.GetSharePointFileContentAsync(sharepointSiteUrl, sharepointRelativeUrl, accessToken);


    if (fileContent != null)
    {
        List<float> embeddings = await generator.GetEmbeddingsAsync(fileContent);

        if (embeddings != null)
        {
            Console.WriteLine("Embeddings generated successfully:");
            Console.WriteLine(JsonSerializer.Serialize(embeddings));
        }
    }

    Console.ReadKey();
}

2. Indexing in Azure AI Search:

Your index schema needs a field to store the vectors. A Collection(Edm.Single) is suitable.

{
  "name": "contentVector",
  "type": "Collection(Edm.Single)",
  "searchable": true, // Important for vector search
  "filterable": false,
  "retrievable": true,
  "vectorSearchConfiguration": "my-vector-config" // Link to your vector config
}

You'll also need a vector search configuration:

{
  "name": "my-vector-config",
  "kind": "hnsw",
  "hnswParameters": {
    "m": 4,
    "efConstruction": 400,
    "efSearch": 500,
    "metric": "cosine"
  }
}

Use the Azure AI Search REST API or SDK to index your documents, including the generated vectors.

Steps to Re-sync the Index with New/Updated Documents

1. Using Power Automate:
   • Create a flow in Power Automate that triggers when a new document is created or an existing document is updated in SharePoint.
   • Use the HTTP action to call an Azure Function that generates embeddings and updates the search index.
2. Using Azure Functions:
   • Create an Azure Function that generates embeddings for new or updated documents and updates the search index.
   • Use the Azure Cognitive Search REST API to update the search index with the new embeddings.
3. Power Automate Flow Example:
   • Trigger: When a file is created or modified in SharePoint.
   • Action: HTTP request to Azure Function.

Hope that helps

-Grace