DeepSeek.NET

A modern .NET (C#) wrapper for the DeepSeek AI API – integrate conversational AI into your applications with ease.

• Model Discovery: Retrieve available AI models effortlessly.
• Smart Conversations: Execute both single-turn and streaming multi-turn dialogues.
• Streaming Support: Real-time response handling for dynamic interactions.
• Modern .NET (C#) Integration: Built with .NET 8 and HttpClient best practices.

📋 Prerequisites

• DeepSeek API Key: Obtain from DeepSeek Platform
• .NET 8+: Target framework requirement

📦 Installation

dotnet add package DeepSeek.NET

🛠️ Quick Start

Initialize Client

// Simple initialization
var client = new DeepSeekClient("your_api_key_here");

// Or with custom HttpClient
var httpClient = new HttpClient();
var factoryClient = new DeepSeekClient(httpClient, "your_api_key_here");

(You can change the HttpClient Timeout with client.setTimeout(int seconds))

Discover Available Models

DeepSeek offers two AI models for different use cases:

• Models.ModelChat: Standard conversational AI model, ideal for chat interactions.
• Models.ModelReasoner: Includes an additional reasoning phase, useful for deeper analytical responses.

var models = await client.ListModelsAsync();
if (models?.Data is null)
{
    Console.WriteLine($"Error: {client.ErrorMsg}");
    return;
}

foreach (var model in models.Data)
{
    Console.WriteLine($"- {model.Id}: {model.Capabilities}");
}

Basic Chat Interaction

When constructing a conversation, you can use different types of messages to define roles:

• Message.NewSystemMessage(content): Used for system instructions or context setting.
• Message.NewAssistantMessage(content): Represents a response from the AI assistant.
• Message.NewUserMessage(content): Represents a message from the user.

var chatRequest = new ChatRequest
{
    Messages = [
        Message.NewUserMessage("Explain quantum computing in 3 sentences"),
        Message.NewAssistantMessage("..."),
        Message.NewUserMessage("Now simplify it for a 5th grader")
    ],
    Model = Models.ModelChat
};

var response = await client.ChatAsync(chatRequest);
var responseContent = response?.Choices.First().Message.Content ?? "No response";

Console.WriteLine(responseContent);

Real-Time Streaming

var streamRequest = new ChatRequest
{
    Messages = [
        Message.NewUserMessage("Tell a 100-word sci-fi story") }
    ],
    Model = Models.ModelChat
};

var stream = await client.ChatStreamAsync(streamRequest);
if (stream is null) return;

await foreach (var chunk in stream)
{
    Console.Write(chunk.Delta?.Content);
}

Read Reasoning Content

Only in case of Models.Reasoner, you will have available also the Reasoning Content

var chatRequest = new ChatRequest
{
    Messages = [
        Message.NewUserMessage("Hello! Tell me an interesting fact about space.")
    ],
    Model = Models.ModelReasoner
};

var response = await client.ChatAsync(chatRequest);
var reasoningContent = response?.Choices.First().Message.ReasoningContent ?? String.Empty;

Console.WriteLine(reasoningContent);

🤝 Contributing

We welcome contributions! Please follow our contribution guidelines when:

• Reporting issues
• Suggesting enhancements
• Submitting pull requests

License

MIT License – See LICENSE for details.