Mythosia.AI 1.7.0

Mythosia.AI

Package Summary

The Mythosia.AI library provides an abstraction for various AI models, including support for OpenAI GPT-4, Anthropic Claude, DeepSeek, Sonar, Gemini, and more.
This library allows easy interaction with AI services, enabling both synchronous completion and streaming capabilities for GPT-3.5, GPT-4, Claude, DeepSeek, Sonar, Gemini, etc.

Supported Models

• OpenAI: GPT-4, GPT-3.5 (Turbo), GPT-4o variants
• Anthropic Claude: Claude 3.x (Claude3_5Sonnet, Claude3Opus, etc.)
• DeepSeek: various chat/reasoner models
• Sonar: sonar, sonar-pro, sonar-reasoning
• Gemini: gemini-1.5-flash (and possible future expansions)

Key Features

• Synchronous completion and streaming support for multiple AI providers
• Customizable model selection, temperature, and max tokens
• Single-prompt token counting and entire-conversation token counting built-in
• Unified abstraction over HTTP requests/responses for each service
• Extendable structure for easily adding new AI model integrations

How to Use

Allocation

using Mythosia.AI;
using System.Net.Http;

var httpClient = new HttpClient();

// Example usage for each supported service:
var aiService = new ChatGptService("your_openai_api_key", httpClient);
var aiService = new ClaudeService("your_anthropic_api_key", httpClient);
var aiService = new DeepSeekService("your_deepseek_api_key", httpClient);
var aiService = new SonarService("your_sonar_api_key", httpClient);
var aiService = new GeminiService("your_gemini_api_key", httpClient);

completion

string response = await aiService.GetCompletionAsync("What is AI?");

Token Counting

// Entire conversation token count
uint totalTokens = await aiService.GetInputTokenCountAsync();

// Single-prompt token count
uint promptTokens = await aiService.GetInputTokenCountAsync("One-off prompt to analyze");

Model Switching Example

// Add to existing ChatBlock usage
aiService.ActivateChat.ChangeModel(AIModel.Gpt4oLatest);
aiService.ActivateChat.ChangeModel("o1-2024-12-17"); // change with string

Streaming Responses

await aiService.StreamCompletionAsync("Explain quantum computing", content => {
Console.WriteLine(content);
});

ASP.NET Core Integration

Below is an example of how to register and consume these AI services in an ASP.NET Core application.

public void ConfigureServices(IServiceCollection services)
{
    services.AddHttpClient();
    
    // Register your AI services with the needed keys
    services.AddScoped<ChatGptService>(sp => 
        new ChatGptService("openai_key", sp.GetRequiredService<HttpClient>()));

    services.AddScoped<ClaudeService>(sp => 
        new ClaudeService("claude_key", sp.GetRequiredService<HttpClient>()));

    services.AddScoped<DeepSeekService>(sp => 
        new DeepSeekService("deepseek_key", sp.GetRequiredService<HttpClient>()));

    services.AddScoped<SonarService>(sp =>
        new SonarService("sonar_key", sp.GetRequiredService<HttpClient>()));

    services.AddScoped<GeminiService>(sp =>
        new GeminiService("gemini_key", sp.GetRequiredService<HttpClient>()));
}

Service Usage in Controllers

[ApiController]
[Route("api/[controller]")]
public class AIController : ControllerBase
{
    private readonly ChatGptService _chatGptService;
    private readonly ClaudeService _claudeService;
    private readonly DeepSeekService _deepseekService;
    private readonly SonarService _sonarService;
    private readonly GeminiService _geminiService;

    public AIController(
        ChatGptService chatGptService, 
        ClaudeService claudeService,
        DeepSeekService deepseekService,
        SonarService sonarService,
        GeminiService geminiService)
    {
        _chatGptService = chatGptService;
        _claudeService = claudeService;
        _deepseekService = deepseekService;
        _sonarService = sonarService;
        _geminiService = geminiService;
    }

    [HttpPost("deepseek-completion")]
    public async Task<IActionResult> GetDeepSeekCompletion([FromBody] string prompt)
    {
        var result = await _deepseekService.GetCompletionAsync(prompt);
        return Ok(result);
    }
    
    [HttpPost("sonar-completion")]
    public async Task<IActionResult> GetSonarCompletion([FromBody] string prompt)
    {
        var result = await _sonarService.GetCompletionAsync(prompt);
        return Ok(result);
    }
}

Conclusion

The Mythosia.AI library simplifies integration with multiple AI providers (OpenAI GPT, Anthropic Claude, DeepSeek, Sonar, Gemini) by offering:

• Synchronous and streaming completion methods
• Single-prompt and full-conversation token counting
• Unified HTTP request/response handling
• Extendable model and service structure

It provides a convenient way to switch models, manage prompts, and retrieve results without worrying about each AI provider’s low-level API differences.