VeloraAI 1.0.2

🚀 VeloraAI is a modern and flexible C# library designed to simplify local LLM integration. It allows developers to interact with quantized AI models directly from .NET 8.0 applications — with a single line of code. Whether you're building chatbots, creative tools, or AI companions, VeloraAI is optimized for speed, reliability, and customization.

✨ Features

• ⚡ Quick-Start Model Loading — Choose from pre-integrated models or load your own via TestingMode.
• 🧠 Support for Multiple Models — CrystalThink, Qwen, Mistral, DeepSeek, Llama and more.
• 🔁 Event-driven Response System — React to TextGenerated, ResponseStarted, and ResponseEnded in real-time.
• 🔐 Customizable System Prompts — Use friendly or aggressive instruction styles (e.g., NoBSMode).
• 📦 Model Downloader — Automatically fetches models from Hugging Face if not already available.
• 📷 Experimental Vision Mode — Send image + prompt for visual reasoning (WIP).

🧱 Built With

• LLamaSharp — Backbone inference engine.
• .NET 8.0 — Modern C# support.
• WinForms & Console — Sample UI and CLI clients included.

📂 Models Available

🔧 Usage

1. Authenticate and Start Chatting

var result = await VeloraAI.AuthenticateAsync(VeloraAI.Models.Crystal_Think_V2_Q4);
if (result == VeloraAI.AuthState.Authenticated)
{
    await VeloraAI.AskAsync("What is the capital of France?");
}

2. Hook Into Events

VeloraAI.TextGenerated += (_, text) => Console.Write(text);
VeloraAI.ResponseStarted += (_, __) => Console.WriteLine("\n[VELORA is typing...]");
VeloraAI.ResponseEnded += (_, __) => Console.WriteLine("\n\n[Done]");

3. Use Custom Models

VeloraAI.TestingMode = true;
VeloraAI.TestingModelPath = @"C:\path\to\your_model.gguf";
await VeloraAI.AuthenticateAsync(VeloraAI.Models.TestingModel);

⚙️ Advanced Prompt Modes

Friendly Assistant (Default)

Follows a natural conversational tone with emojis and personality.

NoBS Mode

Blunt, hyper-logical response style with no emotional overhead or filler.

await VeloraAI.AuthenticateAsync(VeloraAI.Models.Crystal_Think_V2_Q4, NoBSMode: true);

📥 Model Auto-Download

Models are downloaded on first use to:

%APPDATA%/VeloraAI

Progress can be tracked using:

VeloraAI.CurrentDownloadProgress;

🔄 Reset History

VeloraAI.ResetHistory(); // or use custom system prompt

🎯 Custom Inference Parameters

You can fine-tune Velora's behavior using the following optional parameters in AskAsync:

await VeloraAI.AskAsync(
    prompt: "Summarize this paragraph.",
    temperature: 0.25f,
    TopP: 0.2f,
    TopK: 0,
    RepeatPenalty: 1.1f,
    maxTokens: 80
);

🛠️ Contributing

Pull requests are welcome! Please submit improvements, optimizations, or new model integrations.

📄 License

MIT

💬 Example Console Output

Authenticating model...
Authentication result: Authenticated

> What is 21 * 2?
[VELORA is typing...]

42

[Done]

🧪 Credits

• Developed by voidZiAD
• Powered by LLamaSharp, GGUF models, and the C#/.NET 8.0 ecosystem

🌐 Links

• LLamaSharp GitHub
• HuggingFace Models (Hosted by Velora)
• VeloraAI Console/WinForms Example

🧠 "VELORA" Personality

"I'm VELORA — not just another chatbot. I'm here to help you code, reason, and think clearer. No nonsense, just clarity."