What is LM-Kit.NET?

LM-Kit.NET is a comprehensive local AI runtime for .NET developers. It provides a single SDK that covers AI agents, document intelligence, vision, RAG, text analysis, speech, and text generation, all running on-device without cloud calls. It supports CPU, GPU, and various model formats, and integrates with popular .NET AI libraries.

How to use LM-Kit.NET?

Install the LM-Kit.NET NuGet package, load a model using LM.LoadFromModelID, then use the provided APIs to build chatbots, agents, RAG pipelines, or other AI features. The quickstart guide shows how to set up chat, agent, and RAG examples in about five minutes.

Core features of LM-Kit.NET

Seven capability pillars: AI Agents, Document Intelligence, Vision & Multimodal, RAG & Knowledge, Text Analysis, Speech & Audio, Text Generation
On-device inference with CPU, GPU, and multi-GPU support
Built-in vector database and RAG engine with page-level citations
Dynamic Sampling for structured output and reduced hallucinations
Integrations with Microsoft.Extensions.AI, Semantic Kernel, Amazon Bedrock, and vector databases
Support for encrypted models, LoRA, and fine-tuning
Cross-platform: Windows, Linux, macOS