Solutions

The complete capability map, one runtime underneath.

Every entry below is a shipped feature with its own page: 77 capabilities across 10 areas, from document work down to the runtime and the bridges into your stack.

Documents & knowledge

Turn files and internal content into structured data, cited answers, and safe output.

Document Intelligence

Explore

Parse, understand, transform, and protect documents, entirely on your infrastructure.

Structured Data Extraction Typed fields with per-field confidence Document Chat & Q&A Ask PDFs, get cited answers Document RAG Retrieval over whole document sets Document Summarization PDFs and images, local summaries Document Classification Zero-shot categorization at scale Document Search Exact, regex, fuzzy, region OCR High-accuracy recognition, on-device Layout Understanding Deterministic document structure analysis Document to Markdown First-class PDF to Markdown Document Conversion PDF, DOCX, HTML, Markdown, EML Intelligent Document Splitting Boundary detection in mixed batches Email Processing EML, MBOX, ICS parsing Image Processing Deskew, binarize, OCR preparation PDF Toolkit Merge, split, render, searchable PDF PDF/A Conversion ISO 19005 archival output PDF Redaction Permanent removal of text and images Smart Redaction AI finds what must disappear

RAG & Knowledge

Explore

Retrieval with citations, from embedding to reranked answer.

RAG Chat Conversational retrieval, four query modes Built-in Vector Database Embedded storage at any scale Embeddings Dense vectors, on-device Query Expansion HyDE, multi-query, contextualization Reranker Second-pass relevance scoring Intelligent Data Extraction Schema-driven, from any source

Vision & Multimodal

Explore

Understand, classify, and search images with local vision-language models.

Image Understanding Visual question answering Chat with Image Multimodal multi-turn conversations Image Classification Zero-shot VLM categorization Image Labeling Multi-label tagging Image Embeddings Cross-modal vectors VLM-Driven OCR Vision-language text recognition Background Removal Subject isolation and preprocessing

Agents & language

Build assistants, agents, and language features on the same local runtime.

AI Agents

Explore

Plan, orchestrate, and govern agents that run against your documents, inside your perimeter.

Agent Templates 18 pre-built specialized agents Agent Reasoning Six planning strategies Multi-Agent Workflows Pipeline, parallel, router, supervisor Graph Orchestration Composable workflow nodes Agent Delegation Programmatic and model-driven hand-off Agent Memory Persistent, retrieval-grounded context Agent Skills Portable expertise via SKILL.md Chatbots Assistants that remember, reason, act Tools & Function Calling 70+ built-in tools, plus your own Single Function Calling Natural language to one C# method MCP Integration Full MCP client for local agents Permissions & Guardrails Centralized tool access control Filter Pipeline Middleware around every inference Real-Time Streaming Token, thinking, delegation streams Resilience Retry, circuit breaker, bulkhead Observability OpenTelemetry GenAI tracing

Text Analysis

Explore

Classify, extract, and score text and images with typed, confidence-scored results.

Custom Classification Text and image categorization Named Entity Recognition Custom entity types, multimodal PII Extraction Sensitive data detection, fully local Sentiment & Emotion Analysis Polarity and emotion scoring Keyword Extraction LLM-powered topic discovery Language Detection 100+ languages identified Multimodal Embeddings Text and image vectors, built-in store

Text Generation

Explore

Generate, rewrite, and constrain text, down to schema-valid output.

Conversation Primitives Single-turn, multi-turn, stateless Prompt Templates Composable prompt engine Structured Content Creation Grammar-constrained generation Content Summarization Adaptive length, overflow handling Grammar & Spell Correction Real-time text correction Text Enhancement Style and tone rewriting Text Translation 100+ languages, on-device

Speech & Audio

Explore

Transcribe, detect, and translate speech without a cloud call.

Real-Time Transcription Streaming speech-to-text Voice Activity Detection Speech isolation from audio Audio Language Detection Identify the spoken language Speech Translation Real-time, 34+ languages

Runtime & ecosystem

What every capability above runs on, and how it fits the stack you already have.

Local Inference

Explore

The shared runtime: models, hardware acceleration, and inference control.

Model Catalog Curated models, downloaded and cached Hardware Backends CPU, CUDA, Vulkan, Metal Multi-GPU & Tensor Overrides Large models on commodity GPUs Dynamic Sampling Adaptive inference quality Sampling & Generation Controls Logit bias, seeds, constraints Context Hibernation Pause and resume conversations Encrypted Models Stream-decrypted model loading

Model Optimization

Explore

Make models fit your hardware and your domain.

Model Quantization 30+ formats, FP32 to 1-bit LLM Fine-Tuning On-device LoRA training LoRA Integration Hot-swap adapters at runtime

Integrations

Explore

Bridges, compatibility endpoints, and the databases you already run.

Microsoft.Extensions.AI Drop-in IChatClient Semantic Kernel Local IChatCompletionService Amazon Bedrock Titan and Cohere embeddings

Prefer to start from the problem?

Eight scenario pages walk one problem end to end and name the capabilities each step uses.

AI assistants on documents Agentic document workflows Multi-agent document review Assistants with memory Compliance monitoring Invoice & document extraction Private company knowledge PII review & redaction

Frequently asked questions.

What can LM-Kit.NET do?

LM-Kit covers a broad surface of generative-AI capabilities, including AI agents, document intelligence, RAG and knowledge retrieval, vision, text analysis and classification, text generation, speech-to-text, local inference, model optimization and .NET ecosystem integrations. Every category is a feature already available in the SDK.

Is LM-Kit.NET one package or many?

LM-Kit is a single NuGet package that covers all of these capabilities. You pick the area you need and the rest stays out of your way.

Does LM-Kit.NET run locally on-device?

Yes. LM-Kit runs locally inside your .NET application, with capabilities such as embeddings and speech-to-text operating on-device.

What AI agent capabilities does LM-Kit.NET provide for .NET?

LM-Kit runs AI agents on local models, including planners, supervisors, parallel and pipeline orchestrators, persistent memory, MCP clients, custom tool calling, and agent skills.

What speech and text-analysis features does LM-Kit.NET offer?

For speech it provides a local, multilingual speech-to-text stack with streaming output, real-time translation, and 100+ languages. For text analysis it offers classification, NER, PII extraction, sentiment, keyword extraction, and language ID.

Get started

Add local AI to your .NET project today.

Download Read the docs