LLMKit 2.1.0

LLMKit

LLMKit is a thread-safe .NET library that provides a unified interface for interacting with various Large Language Models (LLMs) including OpenAI, Gemini, and DeepSeek.

Features

• Unified interface for multiple LLM providers
• Thread-safe implementation
• Conversation management
• Fluent API for message building
• Configurable parameters (tokens, temperature, etc.)
• Comprehensive error handling
• Dependency injection support
• Cancellation token support
• Custom endpoint support for all providers

Installation

Install-Package LLMKit

Requirements

• .NET 8.0 or later
• Valid API keys for the LLM providers

Quick Start

using LLMKit;
using LLMKit.Providers;

// Using statement ensures proper disposal
using var client = new LLMClient(
    new OpenAIProvider(apiKey: "your-api-key", model: "gpt-3.5-turbo")
);

string response = await client.GenerateTextAsync(
    "You are a helpful assistant.",
    "What is the capital of France?"
);

Usage Examples

Basic Conversation with Proper Disposal

using var client = new LLMClient(new OpenAIProvider("your-api-key", "gpt-3.5-turbo"));
var conversation = client.StartConversation();

await client.SendMessageAsync("Hello, how are you?");
await client.SendMessageAsync("What's the weather like?");
await client.SendMessageAsync("Tell me a joke");

string history = client.GetFormattedConversation();
// Client is automatically disposed here

Custom Parameters

var client = new LLMClient(
    new OpenAIProvider("your-api-key", "gpt-3.5-turbo"),
    defaultMaxTokens: 1000,
    defaultTemperature: 0.7
);

Dependency Injection

services.AddSingleton<ILLMProvider>(sp => 
    new OpenAIProvider(
        apiKey: Configuration["OpenAI:ApiKey"],
        model: Configuration["OpenAI:Model"]
    )
);
services.AddSingleton<LLMClient>();

Error Handling

try
{
    var response = await client.GenerateTextAsync(
        "You are a helpful assistant.",
        "What is the capital of France?"
    );
}
catch (LLMException ex)
{
    Console.WriteLine($"Error: {ex.Message}");
}

Cancellation

using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));
var response = await client.GenerateTextAsync(
    "You are a helpful assistant.",
    "What is the capital of France?",
    cts.Token
);

Supported Providers

OpenAI

var provider = new OpenAIProvider(
    apiKey: "your-api-key",
    model: "gpt-3.5-turbo"
);

Gemini

var provider = new GeminiProvider(
    apiKey: "your-api-key",
    model: "gemini-2.0-flash"
);

DeepSeek

var provider = new DeepSeekProvider(
    apiKey: "your-api-key",
    model: "deepseek-chat"
);

Custom Endpoints

Each provider supports custom endpoints. If not provided, the library will use the default endpoint for that provider.

// OpenAI with custom endpoint
var client = new LLMClient(new OpenAIProvider(
    apiKey: "your-api-key",
    model: "gpt-3.5-turbo",
    endpoint: new Uri("https://api.openai.com/v1/chat/completions")
));

// Gemini with custom endpoint
var client = new LLMClient(new GeminiProvider(
    apiKey: "your-api-key",
    model: "gemini-pro",
    endpoint: new Uri("https://generativelanguage.googleapis.com/v1beta/models")
));

// DeepSeek with custom endpoint
var client = new LLMClient(new DeepSeekProvider(
    apiKey: "your-api-key",
    model: "deepseek-chat",
    endpoint: new Uri("https://api.deepseek.com/v1/chat/completions")
));

License

MIT License. See LICENSE for details.

Support

For issues or questions, please open an issue in the GitHub repository.