AI-Powered Text Generation for .NET, Grammar-Constrained Outputs, LM-Kit

Elevate .NET apps with AI-powered text generation.

LM-Kit.NET brings cutting-edge AI-driven text generation directly to your .NET applications. Generate coherent, contextually relevant text across various domains, from structured content creation and summarization to grammar correction and text enhancement. By leveraging compact Large Language Models (LLMs) that run entirely on-device, LM-Kit ensures fast, secure, and private processing without the need for cloud services.

Powered by Dynamic Sampling technology, LM-Kit achieves up to 75% error reduction and 2x faster inference speeds. With support for grammar-constrained outputs, multimodal inputs, and reasoning capabilities, you can build sophisticated text generation pipelines that meet enterprise requirements.

New in 2026: Agent templates for text generation tasks including WriterAgentTemplate, EditorAgentTemplate, and SummarizerAgentTemplate. Build agentic text generation workflows with planning strategies and tool integration.

Sampling

Dynamic Sampling

Proprietary technology ensuring accurate, constrained outputs with optimal performance.

Multimodal

Multimodal support

Generate text from images, documents, and attachments with VLM integration.

Reasoning

Reasoning capabilities

Chain-of-thought prompting and intermediate reasoning steps with ReasoningLevel control.

Local

100% local

On-device inference with no cloud dependency. Your data never leaves your infrastructure.

Structured content creation

Grammar-constrained text generation.

Generate outputs that strictly conform to predefined formats using custom grammars, JSON schemas, and BNF definitions. Powered by Dynamic Sampling technology for up to 75% error reduction.

JSON

JSON schema enforcement

Create Grammar objects from JSON schemas to ensure generated content aligns with structure and data types. Build JSON outputs directly from .NET types including complex structures like dates and arrays.

BNF

BNF grammar definitions

Define custom grammars using Backus-Naur Form (BNF) for domain-specific output formats. Enforce exact syntax requirements for code generation, queries, and structured documents.

Schema

`TextExtractionElement`

Define extraction schemas with typed elements including String, Float, Date, Boolean, and Arrays. Supports formatting hints, case modes, and required field validation.

Built-in

Predefined grammar types

Use built-in grammars for common formats: Json, JsonArray, Arithmetic, List, and Boolean. Rapid prototyping with zero grammar definition required.

Constrain

Whitelisted values

Constrain outputs to specific allowed values using the WhitelistedValues property. Perfect for classification tasks and controlled vocabulary generation.

Discovery

Schema discovery

Automatically discover extraction schemas from unstructured content using SchemaDiscovery. Let the AI identify relevant fields and data types.

Grammar factory methods

Create a Grammar object from a JSON schema string to enforce structured JSON output generation.
Build grammar from TextExtractionElement collection for typed data extraction with validation.
Generate grammar from a list of allowed string values for constrained classification outputs.
Create JSON grammar from field definitions for rapid schema development without full JSON schema.

StructuredOutput.cs

using LMKit.Model;
using LMKit.TextGeneration;
using LMKit.TextGeneration.Sampling;
using LMKit.Extraction;

// Load the model
var model = LM.LoadFromModelID("qwen3.5:4b");

// Define extraction elements for structured output
var elements = new List<TextExtractionElement>
{
    new TextExtractionElement("ProductName", ElementType.String, "Name of the product"),
    new TextExtractionElement("Price", ElementType.Float, "Price in USD"),
    new TextExtractionElement("ReleaseDate", ElementType.Date, "Product release date"),
    new TextExtractionElement("Features", ElementType.StringArray, "List of key features")
};

// Create grammar from extraction elements
var grammar = Grammar.CreateJsonGrammarFromExtractionElements(elements);

// Create conversation with grammar constraint
var chat = new SingleTurnConversation(model) { Grammar = grammar };

// Generate structured output
var result = chat.Submit("Extract product info: The SuperWidget Pro costs $299.99 and launches March 15, 2026.");
Console.WriteLine(result.Completion);

Content summarization

Intelligent text condensation.

Condense lengthy texts into concise, informative summaries. Extract key insights from any content size with adaptive overflow strategies and multilingual support.

Adaptive summarization engine

The LM-Kit Summarizer is a sophisticated NLP engine that employs innovative techniques to condense content of any size or language into coherent summaries. Multiple overflow strategies ensure reliable handling of massive inputs, from short articles to lengthy legal documents.

Overflow resolution strategies

Truncate: Process the first portion that fits within context limits
RecursiveSummarize: Iteratively summarize chunks then combine
Reject: Return error for content exceeding limits
Automatic title generation with GenerateTitle property
Configurable summary length via MaxContentWords
Target language detection and generation
Multimodal support: summarize from images
Confidence score for quality assessment

Summarizer.cs

using LMKit.Model;
using LMKit.TextGeneration;

var model = LM.LoadFromModelID("mistral-small3.2");

var summarizer = new Summarizer(model)
{
    GenerateTitle = true,
    GenerateContent = true,
    MaxContentWords = 150,
    OverflowStrategy = OverflowResolutionStrategy.RecursiveSummarize,
    TargetLanguage = Language.English
};

var result = summarizer.Summarize(
    File.ReadAllText("annual-report.txt"));

Console.WriteLine($"Title: {result.Title}");
Console.WriteLine($"Summary: {result.Content}");

Correct text in multiple languages with language-specific rules and conventions.

Realtime

Real-time processing

Instant corrections as users type with async support for responsive applications.

TextCorrection.cs

using LMKit.Model;
using LMKit.TextEnhancement;

var model = LM.LoadFromModelID("qwen3.5:4b");
var corrector = new TextCorrection(model);

string input = "My cat don't never flyed to the moon " +
    "because he too much tired from baking pizza.";

string corrected = corrector.Correct(input);

Console.WriteLine("Original: " + input);
Console.WriteLine("Corrected: " + corrected);
// Output: "My cat has never flown to the moon
// because he was too tired from baking pizza."

// Async support for real-time applications
var asyncResult = await corrector.CorrectAsync(input);

TextRewriter.cs

using LMKit.Model;
using LMKit.TextEnhancement;

var model = LM.LoadFromModelID("qwen3.5:4b");
var rewriter = new TextRewriter(model);

string casual = "Hey! So basically the project's done and it works great!";

// Transform to professional style
string professional = rewriter.Rewrite(casual, WritingStyle.Professional);
Console.WriteLine("Professional: " + professional);
// "The project has been completed successfully and is functioning as expected."

// Transform to concise style
string concise = rewriter.Rewrite(casual, WritingStyle.Concise);
Console.WriteLine("Concise: " + concise);
// "Project complete. Fully operational."

Demos

Get started with examples.

Explore working code samples demonstrating text generation capabilities.

Web content extractor

Compose prompts like programs.

Every prompt LM-Kit produces, including the system prompts driving conversations, agents, and skills, can flow through a full templating engine. Three syntaxes, filters, conditionals, loops, custom helpers. Compiled once, reused everywhere.

Prompts

Prompt templates

Mustache, dollar, or percent syntax. Variable substitution, filter library, conditionals (#if, #unless), loops (#each), scoping (#with), custom helpers.

Explore prompt templates

Demos & docs

Build it. Read it. Try it.

Working console demos on GitHub, step-by-step how-to guides on the docs site, and the API reference for the classes used on this page.

Demo

Single-turn chat

Console demo: one-shot prompt with a local LLM.

Open on GitHub → Demo

Multi-turn chat

Console demo: history-aware chat with streaming.

Open on GitHub → Demo

Text summarizer

Console demo: long-form summaries with overflow policies.

Open on GitHub → How-to guide

Load a model and generate

Foundational how-to: from NuGet to first inference.

Read the guide →

LM-Kit.NET pillars

Seven pillars, one foundation.

The seven pillars of LM-Kit.NET, plus the local runtime they share. Highlighted card is where you are now.

01 · AI Agents

Orchestration patterns

ReAct planning, supervisors, parallel and pipeline orchestrators, persistent memory, MCP clients, custom tools.

AI Agents

02 · Document Intelligence

Parse PDFs, images, EML

PDF text and table extraction, on-device OCR reaching SOTA benchmark scores, structured field extraction with grammar-constrained generation.

Document Intelligence

03 · Vision & Multimodal

VLMs, image classification, chat with image

Image understanding, classification, labeling, multimodal chat, image embeddings, VLM-OCR, background removal. Same conversation surface as LLMs.

Vision & Multimodal

04 · RAG & Knowledge

Vector search and retrieval

Built-in vector store, Qdrant connector, embeddings, hybrid retrieval, document chunking, source citations.

RAG & Knowledge

05 · Text Analysis

Classification, NER, PII, sentiment

Built-in classifiers and an extractor that emits typed C# objects via grammar-constrained sampling. Sentiment, keywords, language detection.

Text Analysis

06 · Speech & Audio

Audio transcription, STT

A growing local speech-to-text stack: hallucination suppression, Voice Activity Detection, real-time translation, streaming output, 100+ languages.

Speech & Audio

07 · Text Generation

Conversations, rewriting, summaries

Single-turn, multi-turn, and stateless conversation primitives. Translate, correct, rewrite, summarise. Prompt templates, streaming, grammar-constrained outputs.

You are here

The foundation

Every capability above runs on this runtime.

Foundation

Local Inference

The runtime all seven pillars sit on. The LM-Kit.NET NuGet ships the complete inference system: open-weight LLMs, vision-language models, embeddings, on-device speech-to-text, OCR and classifiers, accelerated on CPU, AVX2, CUDA 12/13, Vulkan or Metal. One package, zero cloud calls, predictable latency, full data and technology sovereignty.

Explore the foundation

Ready to build AI-powered text generation?

Full control over content creation, summarization, and enhancement. 100% local, 100% your infrastructure.

Download free API documentation

AI-powered text generation for .NET.

Structured content creation

Content summarization

Grammar & spell correction

Text enhancement

Dynamic Sampling

Multimodal support

Reasoning capabilities

100% local

JSON schema enforcement

BNF grammar definitions

TextExtractionElement

Predefined grammar types

Whitelisted values

Schema discovery

Overflow resolution strategies

Briefing

Key points

Abstract

Custom

Grammar correction

Spell checking

Multilingual support

Real-time processing

Concise

Professional

Friendly

Custom

Brand consistency

Language adaptation

Audience targeting

Async processing

Web content extractor

Text summarizer

Text corrector

Text rewriter

Keyword extraction

Multi-turn chat

SingleTurnConversation

MultiTurnConversation

Summarizer

Grammar

TextCorrection

TextRewriter

KeywordExtraction

TextExtraction

Prompt templates

Single-turn chat

Multi-turn chat

Text summarizer

Load a model and generate

Orchestration patterns

Parse PDFs, images, EML

VLMs, image classification, chat with image

Vector search and retrieval

Classification, NER, PII, sentiment

Audio transcription, STT

Conversations, rewriting, summaries

Local Inference

`TextExtractionElement`

`SingleTurnConversation`

`MultiTurnConversation`

`Summarizer`

`Grammar`

`TextCorrection`

`TextRewriter`

`KeywordExtraction`

`TextExtraction`