Knowledge · AI glossary
AI terms, explained simply.
LLM, token, RAG, agent: a lot of terms circulate around AI. Here we explain the most important ones briefly and without jargon — with everyday examples and a note on what they mean for you.
01 · 9 terms
Fundamentals
The building blocks modern AI is built on.
Artificial Intelligence (AI)
Artificial intelligence (AI) refers to computer systems that take on tasks that normally require human thinking — for example understanding text, recognizing patterns or preparing decisions.
Context Window
The context window states how much text a model can take into account at once — instruction, documents, conversation history and answer combined. It is measured in tokens.Learn more →Matching tools: Claude, Google Gemini
Generative AI
Generative AI creates new content — text, images, audio or code — instead of merely sorting or classifying existing data.
Large Language Model (LLM)
A large language model (LLM) is a language model trained on enormous amounts of text. It generates text by predicting, piece by piece, the most likely next chunk of text.Learn more →Matching tools: ChatGPT, Claude, Google Gemini, Mistral Vibe
Machine Learning
Machine learning is a branch of AI: instead of following fixed rules, a system learns relationships from example data and applies them to new data.
Multimodal AI
Multimodal AI processes several types of input at once, such as text, images, audio and video, and can connect them with one another.Matching tools: Google Gemini
Neural Network
A neural network is a mathematical model made of many connected computing nodes, loosely modeled on the brain. It learns by adjusting its connections during training.
Token
A token is the smallest unit of text a language model works with — usually a word or part of a word. The length and cost of a request are measured in tokens.Learn more →Matching tools: OpenRouter
Transformer
The transformer is the model architecture most of today's language models are built on. Its attention mechanism recognizes which words in a text belong together.
02 · 11 terms
Working with language models
How you address language models, give them knowledge and adapt them.
Benchmark
A benchmark is a standardized test that makes models comparable. It shows tendencies but does not replace a test with your own tasks.Matching tools: Arena
Embedding
An embedding is the representation of a text (or image) as a series of numbers that captures its meaning. Similar content ends up close together in it.Matching tools: LlamaIndex
Fine-Tuning
In fine-tuning, an already trained model is trained further with your own examples so that it hits a style, format or task more reliably.Learn more →Matching tools: Hugging Face
Hallucination
Hallucination is when a language model makes up statements that sound convincing but are false or unsupported.Learn more →Matching tools: Perplexity, Gemini Notebook
Mixture of Experts (MoE)
In a mixture-of-experts model only part of the parameters, the so-called active parameters, compute each request. The model therefore answers faster and more cheaply, but still needs memory for all parameters.Learn more →
Prompt
A prompt is the input to an AI: a question, instruction or task, often with context, examples and wishes for the format of the answer.
Prompt Engineering
Prompt engineering means designing inputs to an AI so that the results are reliably usable: with a clear task, context, examples and the desired format.Learn more →Matching tools: Claude, ChatGPT
Reasoning Model
A reasoning model produces intermediate steps before the actual answer and "thinks" the task through. This helps with logic, math and code but costs more time and compute.Matching tools: ChatGPT, Claude, DeepSeek
Retrieval-Augmented Generation (RAG)
Retrieval-augmented generation (RAG) connects a language model with your own sources: for each question, matching passages of text are looked up and handed to the model as the basis for its answer.Learn more →Matching tools: LlamaIndex, AnythingLLM, Open WebUI
System Prompt
The system prompt is the basic instruction that tells a model who it is, how it behaves and what it must not do. Users usually do not see it.
Vector Database
A vector database stores embeddings and quickly finds the entries most similar in meaning to a query. It is the search foundation of many RAG systems.Matching tools: LlamaIndex
03 · 12 terms
Agents & automation
When AI does not just answer but gets tasks done.
Agent Framework
An agent framework is a software library with building blocks for AI agents — for example for tool connections, memory, planning and workflows.Matching tools: LangChain, LangGraph, LlamaIndex, CrewAI
AI Agent
An AI agent is a system that pursues goals with a language model: it plans steps, uses tools such as search, databases or email, and checks intermediate results.Learn more →Matching tools: n8n, LangGraph, CrewAI, Claude Code
Chatbot
A chatbot holds conversations in text form. Modern chatbots built on language models understand free wording and can access company knowledge.Matching tools: Open WebUI, ChatGPT
Coding Agent
A coding agent is an AI agent for software development: it reads code, proposes changes, runs tests and carries out tasks step by step — ideally with human review.Matching tools: Claude Code, Replit, Kiro, Devin
Function Calling
In function calling, a language model does not output running text but a structured call to a function, which your application then runs — for example "create appointment".Matching tools: LangChain
Human-in-the-Loop
Human-in-the-loop means that a person checks or approves important steps before an AI decision takes effect.Matching tools: LangGraph
Low-Code / No-Code
Low-code and no-code platforms build applications and processes with a visual editor instead of a lot of programming. For AI workflows they are a quick way in.Matching tools: n8n, Zapier, Make
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard through which AI applications access data and tools in a uniform way, such as files, databases or services.Learn more →Matching tools: Claude, Claude Code, n8n
Multi-Agent System
A multi-agent system lets several AI agents with different roles work together, such as planner, researcher and reviewer, passing intermediate results to one another.Matching tools: AutoGen, CrewAI, LangGraph
Orchestration
Orchestration controls in what order models, tools and data sources work together, when an agent continues and when a human approves.Matching tools: LangGraph, n8n
Sandbox
A sandbox is an isolated environment in which programs or AI-generated code run without being able to access your real systems and data.Matching tools: E2B
Workflow Automation
Workflow automation connects systems and work steps into processes that run automatically. With AI, unstructured inputs such as emails or documents can be processed as well.Matching tools: n8n, Zapier, Make, Activepieces
04 · 6 terms
Operation & law
Where AI runs, who owns the data and what the law requires.
EU AI Act
The EU AI Act is the European regulation on AI. It classifies applications by risk and attaches graded obligations to them, which apply step by step.
GDPR and AI
The GDPR also applies when you use AI, as soon as personal data is processed. Legal basis, data processing agreements, storage location and transparency matter.Learn more →Matching tools: Ollama, n8n, Open WebUI
LLM Gateway (Router)
An LLM gateway bundles access to many language models behind one interface and routes requests to suitable models depending on cost, speed or quality.Matching tools: OpenRouter, Eden AI
Local Model (On-Premise)
A local model runs on your own hardware or in your own data center instead of at a cloud provider. The data does not leave your company in the process.Matching tools: Ollama, LM Studio, Open WebUI
Open Weights
With open-weights models, the trained model weights are published, so you can run and adapt the model yourself. That is not automatically the same as open source.Learn more →Matching tools: Ollama, LM Studio, Hugging Face
Quantization
Quantization stores a language model's weights with fewer bits, for example 4 instead of 16. The model then needs only about a quarter of the memory and runs on much smaller hardware, usually with little loss in quality.Learn more →Matching tools: Ollama, LM Studio, Hugging Face
05 · 2 terms
Visibility in AI search
How content shows up in AI answers.
Generative Engine Optimization (GEO)
Generative engine optimization (GEO) means preparing content so that AI search and chat assistants understand, use and cite it.Learn more →Matching tools: Perplexity
llms.txt
llms.txt is a text file in a website's root directory that gives AI systems an overview of the most important content. It is a proposal, not an official standard.