Core Capabilities

LM-Kit.NET delivers a complete AI stack: the fastest .NET inference engine, domain-tuned models that solve real-world problems out of the box, and a comprehensive orchestration layer for building agents and RAG applications.

🤖 AI Agents and Orchestration

Build autonomous AI agents that reason, plan, and execute complex workflows within your applications.

• Agent Framework - Complete agent infrastructure with Agent, AgentBuilder, AgentExecutor, and AgentRegistry for building production-ready AI agents
• Multi-Agent Orchestration - Coordinate multiple agents with PipelineOrchestrator, ParallelOrchestrator, RouterOrchestrator, and SupervisorOrchestrator
• Planning Strategies - Multiple reasoning approaches including ReAct, Chain-of-Thought, Tree-of-Thought, Plan-and-Execute, and Reflection handlers
• 56 Built-in Tools - Ready-to-use tools across categories: Data (JSON, XML, CSV, YAML), Text (Diff, Regex, Template), Numeric (Calculator, Stats), Security (Hash, Crypto, JWT), IO (FileSystem, HTTP, Network), and more
• Agent-to-Agent Delegation - Enable agents to delegate tasks to specialized sub-agents with DelegationManager and DelegateTool
• Agent Templates - 18 pre-built templates including Chat, Assistant, Code, Research, Analyst, Planner, and more for rapid development
• Resilience Policies - Production-grade reliability with Retry, Circuit Breaker, Timeout, Rate Limit, Bulkhead, and Fallback policies
• Streaming Support - Real-time response streaming with buffered, multicast, and delegate handlers
• Agent Observability - Full tracing and metrics with AgentTracer, AgentMetrics, and JSON export capabilities
• MCP Client Support - Connect to Model Context Protocol servers for extended capabilities including resources, prompts, and tool discovery
• Agent Memory - Persistent memory that survives across conversation sessions with RAG-based recall
• Reasoning Control - Adjust reasoning depth for models that support extended thinking
• Function Calling - Let models dynamically invoke your application's methods with structured parameters

🔍 Multimodal Intelligence

Process and understand content across text, images, documents, and audio.

• Vision Language Models (VLM) - Analyze images, extract information, answer questions about visual content
• VLM-Based OCR - High-accuracy text extraction from images and scanned content
• Speech-to-Text - Transcribe audio with voice activity detection and multi-language support
• Document Processing - Native support for PDF, DOCX, XLSX, PPTX, HTML, and image formats
• Image Embeddings - Generate semantic representations of images for similarity search

📚 Retrieval-Augmented Generation (RAG)

Ground AI responses in your organization's knowledge with a flexible, extensible RAG framework.

• Modular RAG Architecture - Use built-in pipelines or implement custom retrieval strategies
• Built-in Vector Database - Store and search embeddings without external dependencies
• PDF Chat and Document RAG - Chat and retrieve over documents with dedicated workflows
• Multimodal RAG - Retrieve relevant content from both text and images
• Advanced Chunking - Markdown-aware, semantic, and layout-based chunking strategies
• Reranking - Improve retrieval precision with semantic reranking
• External Vector Store Integration - Connect to Qdrant and other vector databases

📊 Structured Data Extraction

Transform unstructured content into structured, actionable data.

• Schema-Based Extraction - Define extraction targets using JSON schemas or custom elements
• Named Entity Recognition (NER) - Extract people, organizations, locations, and custom entity types
• PII Detection - Identify and classify personal identifiers for privacy compliance
• Multimodal Extraction - Extract structured data from images and documents
• Layout-Aware Processing - Detect paragraphs and lines, support region-based workflows

💡 Content Intelligence

Analyze and understand text and visual content.

• Sentiment and Emotion Analysis - Detect emotional tone from text and images
• Custom Classification - Categorize text and images into your defined classes
• Keyword Extraction - Identify key terms and phrases
• Language Detection - Identify languages from text, images, or audio
• Summarization - Condense long content with configurable strategies

✍️ Text Generation and Transformation

Generate and refine content with precise control.

• Conversational AI - Build context-aware chatbots with multi-turn memory
• Constrained Generation - Guide model outputs using JSON schemas, templates, or custom grammar rules
• Translation - Convert text between languages with confidence scoring
• Text Enhancement - Improve clarity, fix grammar, adapt tone

🛠️ Model Customization

Tailor models to your specific domain.

• Fine-Tuning - Train models on your data with LoRA support
• Dynamic LoRA Loading - Switch adapters at runtime without reloading base models
• Quantization - Optimize models for your deployment constraints
• Training Dataset Tools - Prepare and export datasets in standard formats

Supported Models

LM-Kit.NET includes domain-tuned models optimized for real-world tasks, plus broad compatibility with models from leading providers:

Text Models: LLaMA, Mistral, Mixtral, Qwen, Phi, Gemma, Granite, DeepSeek, Falcon, and more

Vision Models: Qwen-VL, MiniCPM-V, Pixtral, Gemma Vision, LightOnOCR

Embedding Models: BGE, Nomic, Qwen Embedding, Gemma Embedding

Speech Models: Whisper (all sizes), with voice activity detection

Browse production-ready models in the Model Catalog, or load models directly from any Hugging Face repository.

Performance and Hardware

The Fastest .NET Inference Engine

LM-Kit.NET automatically leverages the best available acceleration on any hardware:

• NVIDIA GPUs - CUDA backends with optimized kernels
• AMD/Intel GPUs - Vulkan backend for cross-vendor GPU support
• Apple Silicon - Metal acceleration for M-series chips
• Multi-GPU - Distribute models across multiple GPUs
• CPU Fallback - Optimized CPU inference when GPU unavailable

Dual Backend Architecture

Choose the optimal inference engine for your use case:

• llama.cpp Backend - Broad model compatibility, memory efficiency
• ONNX Runtime - Optimized inference for supported model formats

Observability

Gain full visibility into AI operations with comprehensive instrumentation:

• OpenTelemetry Integration - GenAI semantic conventions for distributed tracing and metrics
• Inference Metrics - Token counts, processing rates, generation speeds, context utilization, perplexity scores, and sampling statistics
• Event Callbacks - Fine-grained hooks for token sampling, tool invocations, and generation lifecycle

Platform Support

Operating Systems

• Windows - Windows 7 through Windows 11
• macOS - macOS 11+ (Intel and Apple Silicon)
• Linux - glibc 2.27+ (x64 and ARM64)

.NET Frameworks

Compatible from .NET Framework 4.6.2 through the latest .NET releases, with optimized binaries for each version.

Integration

Zero Dependencies

LM-Kit.NET ships as a single NuGet package with absolutely no external dependencies:

No Python runtime. No containers. No external services. No native libraries to manage separately. The entire AI stack runs in-process within your .NET application, making deployment as simple as any other NuGet package.

Ecosystem Connections

• Semantic Kernel - Use LM-Kit.NET as a backend for Microsoft Semantic Kernel
• Vector Databases - Integrate with Qdrant via open-source connectors
• MCP Servers - Connect to Model Context Protocol servers for extended tool access

Data Privacy and Security

Running inference locally provides inherent security advantages:

• No data transmission - Content never leaves your network
• No third-party access - No external services process your data
• Audit-friendly - Complete visibility into AI operations
• Air-gapped deployment - Works in disconnected environments

This architecture simplifies compliance with GDPR, HIPAA, SOC 2, and other regulatory frameworks.

Getting Started

Basic Chat

AI Agent with Tools

Explore more:

• Documentation
• GitHub Demo Repository
• Model Catalog