How-To Guides
Use this section when you want to build something specific with LM-Kit.NET.
Unlike API reference pages, these guides are practical implementation paths. Each guide focuses on a real outcome, includes runnable code, and helps you move from idea to working feature quickly.
Tip
New to LM-Kit.NET? Start with Your First AI Agent, then come back here to build your next capability.
Quick Start
- Pick your immediate goal from What are you trying to build?
- Open the recommended starting guide.
- Run the sample code as-is first.
- Customize one thing at a time, then retest.
This approach gives you fast wins and lowers integration risk.
What are you trying to build?
| Your goal | Start here |
|---|---|
| Load a model and generate responses | Load a Model and Generate Your First Response |
| Build an agent that can call tools | Create an AI Agent with Tools |
| Build a robust production agent | Build a Resilient Production Agent |
| Work with PDFs and document Q&A | Build a Private Document Q&A System |
| Build an end-to-end RAG pipeline | Build a RAG Pipeline for Retrieval-Augmented Generation |
| Extract structured data from documents | Extract Structured Data from Documents |
| Add memory and session continuity to agents | Build a Conversational Assistant with Memory |
| Connect agents to external MCP servers | Connect to MCP Servers from Your Application |
| Add speech transcription and post-processing | Transcribe Audio with Local Speech-to-Text |
| Improve performance and GPU utilization | Configure GPU Backends and Optimize Performance |
Learning Paths
If you want a guided progression, follow one of these paths.
1) Agent Builder Path
- Load a Model and Generate Your First Response
- Create an AI Agent with Tools
- Build a Function-Calling Agent
- Build a Multi-Agent Workflow
- Build a Resilient Production Agent
2) Document and RAG Path
- Build a Private Document Q&A System
- Build a Multi-Format Document Ingestion Pipeline
- Build a RAG Pipeline for Retrieval-Augmented Generation
- Improve RAG Results with Reranking
- Build a Persistent Document Knowledge Base
3) Speech and Audio Intelligence Path
- Transcribe Audio with Local Speech-to-Text
- Transcribe and Reformat Audio with LLM Post-Processing
- Extract Action Items and Tasks from Meeting Recordings
- Generate Structured Meeting Notes from Audio Recordings
How to get the most value from each guide
- Start with the exact dependencies and model recommendations in the guide.
- Keep prompts, schemas, and tool signatures under source control.
- Add telemetry early so you can track latency, token usage, and failures.
- Validate with small, known test inputs before large production data.
- For agent workflows, test fallback behavior and timeout handling.
- For document pipelines, test noisy scans and mixed file formats.
Browse by Theme
Use the left sidebar to explore guides by capability area:
- Foundations: model loading, quantization, GPU optimization, sampling
- Agents and orchestration: tools, planning, multi-agent workflows, memory
- Document intelligence: OCR, extraction, layout analysis, search
- RAG and retrieval: chunking, embeddings, reranking, persistent knowledge
- Text intelligence: classification, sentiment, emotions, rewriting, translation
- Speech and audio: transcription, chaptering, action item extraction, translation
- Training and optimization: LoRA dataset prep and adapter workflows
Related Resources
- Getting Started: environment setup and first steps.
- Choosing the Right Model: model selection by task and hardware.
- Samples Overview: runnable projects you can open immediately.
- Glossary: key AI and LM-Kit.NET concepts.