LLamaSharp

Why we included this project

LLamaSharp gives C# developers a managed path to local LLM inference, wrapping llama.cpp so you can load a GGUF model and generate text from your own code instead of dealing with native C++ or a separate service. It runs both text-only LLaMA-family models and multimodal ones like LLaVA, with CPU and GPU backends (CUDA and Vulkan among them) distributed as NuGet packages, so most hardware setup is just an install. The higher-level APIs and built-in integrations for Semantic Kernel and kernel-memory make it straightforward to connect these models to existing .NET agents, chat features, or retrieval flows. If your team is comfortable staying in C# and wants on-premises or privacy-sensitive assistants, this is a practical foundation to build on.