DeepSeek AI Integration with .NET MAUI Smart Components

19 Dec 20256 minutes to read

The Syncfusion .NET MAUI AI-powered components can enhance applications with intelligent capabilities. You can integrate DeepSeek using the IChatInferenceService interface, which standardizes communication between the editor and your custom AI service.

Setting Up DeepSeek

1. Obtain a DeepSeek API Key
   Create an account at DeepSeek Platform, sign in, and navigate to API Keys to generate an API key.
2. Review Model Specifications
   Refer to DeepSeek Models Documentation for details on available models (e.g., deepseek-chat).

Define Request and Response Models

Create a file named DeepSeekModels.cs in the Services folder and add:

DeepSeekModels.cs

public class DeepSeekMessage
{
    public string? Role { get; set; }
    public string? Content { get; set; }
}

public class DeepSeekChatRequest
{
    public string? Model { get; set; }
    public float Temperature { get; set; }
    public List<DeepSeekMessage>? Messages { get; set; }
}

public class DeepSeekChatResponse
{
    public List<DeepSeekChoice>? Choices { get; set; }
}

public class DeepSeekChoice
{
    public DeepSeekMessage? Message { get; set; }
}

Create a DeepSeek AI Service

This service manages requests to the DeepSeek Chat Completions endpoint and returns the generated text.

1. Create a Services folder in your project.
2. Add a new file named DeepSeekAIService.cs in the Services folder.
3. Implement the service as shown below:

DeepSeekAIService.cs

using Microsoft.Extensions.AI;
using Microsoft.Extensions.Configuration;
using System.Net;
using System.Text;
using System.Text.Json;

public class DeepSeekAIService
{
    private readonly string _apiKey = ""; // API key
    private readonly string _modelName = "deepseek-chat"; // Example model
    private readonly string _endpoint = "https://api.deepseek.com/v1/chat/completions";
    private static readonly HttpClient HttpClient = new(new SocketsHttpHandler
    {
        PooledConnectionLifetime = TimeSpan.FromMinutes(30),
        EnableMultipleHttp2Connections = true
    })
    {
        DefaultRequestVersion = HttpVersion.Version20 // Fallback to HTTP/2 for compatibility
    };
    private static readonly JsonSerializerOptions JsonOptions = new()
    {
        PropertyNamingPolicy = JsonNamingPolicy.CamelCase
    };

    public DeepSeekAIService()
    {
        if (!HttpClient.DefaultRequestHeaders.Contains("Authorization"))
        {
            HttpClient.DefaultRequestHeaders.Clear();
            HttpClient.DefaultRequestHeaders.Add("Authorization", $"Bearer {_apiKey}");
        }
    }

    public async Task<string> CompleteAsync(List<ChatMessage> chatMessages)
    {
        var requestBody = new DeepSeekChatRequest
        {
            Model = _modelName,
            Temperature = 0.7f, // Controls response randomness (0.0 to 1.0)
            Messages = chatMessages.Select(m => new DeepSeekMessage
            {
                Role = m.Role == ChatRole.User ? "user" : "system", // Align with DeepSeek API roles
                Content = m.Text
            }).ToList()
        };

        var content = new StringContent(JsonSerializer.Serialize(requestBody, JsonOptions), Encoding.UTF8, "application/json");

        try
        {
            var response = await HttpClient.PostAsync(_endpoint, content);
            response.EnsureSuccessStatusCode();
            var responseString = await response.Content.ReadAsStringAsync();
            var responseObject = JsonSerializer.Deserialize<DeepSeekChatResponse>(responseString, JsonOptions);
            return responseObject?.Choices?.FirstOrDefault()?.Message?.Content ?? "No response from DeepSeek.";
        }
        catch (Exception ex) when (ex is HttpRequestException || ex is JsonException)
        {
            throw new InvalidOperationException("Failed to communicate with DeepSeek API.", ex);
        }
    }
}

Implement IChatInferenceService

Create DeepSeekInferenceService.cs:

DeepSeekInferenceService.cs

using Syncfusion.Maui.SmartComponents;

public class DeepSeekInferenceService : IChatInferenceService
{
    private readonly DeepSeekAIService _deepSeekService;

    public DeepSeekInferenceService(DeepSeekAIService deepSeekService)
    {
        _deepSeekService = deepSeekService;
    }

    public async Task<string> GenerateResponseAsync(List<ChatMessage> chatMessages)
    {
        return await _deepSeekService.CompleteAsync(chatMessages);
    }
}

Register Services in MAUI

Update MauiProgram.cs:

MauiProgram.cs

using Syncfusion.Maui.Core.Hosting;
using Syncfusion.Maui.SmartComponents;

var builder = MauiApp.CreateBuilder();
builder
    .UseMauiApp<App>()
    .ConfigureSyncfusionCore();

builder.Services.AddSingleton<DeepSeekAIService>();
builder.Services.AddSingleton<IChatInferenceService, DeepSeekInferenceService>();