AI Glossary

Key terms and concepts for self-hosted AI agents and small business automation.

A B C D E F G H I L M O P R S T V W

A

AI Agent

An autonomous software system that can perceive its environment, make decisions, and take actions to achieve specific goals without constant human intervention.

API (Application Programming Interface)

A set of rules that allows different software applications to communicate with each other. AI agents use APIs to connect to tools like email, calendars, and databases.

Automation

Using AI agents to perform repetitive tasks without human intervention. Examples: email responses, appointment scheduling, data entry.

B

Backup

Copies of data and system configurations. Critical for self-hosted systems to prevent data loss from hardware failures.

C

Context Window

The maximum amount of text (measured in tokens) an AI model can process at once. Larger context windows allow the AI to understand longer documents or conversations.

D

Docker

A platform for containerizing applications. Provides isolation, reproducibility, and easy deployment for AI agents and their dependencies.

E

Embedding

Converting text into numerical vectors that capture semantic meaning. Used in RAG systems to find relevant information in databases.

F

Fine-tuning

Training an AI model on specific data to improve its performance on particular tasks. Makes general models specialized for your business needs.

G

GPU (Graphics Processing Unit)

Specialized hardware that accelerates AI computations. Self-hosted setups often use GPUs for faster inference.

H

Hallucination

When an AI generates confident but incorrect or fabricated information. Understanding this helps clients know why verification and human oversight remain important.

I

Inference

The process of an AI model generating a response or prediction. This is what happens when you send a prompt and get an answer back.

L

Latency

The time delay between sending a request and receiving a response. Lower latency means faster AI responses.

LLM (Large Language Model)

A neural network trained on vast text data to understand and generate human-like language. Examples: GPT-4, Claude, Llama, Mistral.

M

MCP (Model Context Protocol)

A standard for connecting AI models to external tools and data sources. Enables agents to interact with your existing business systems.

O

Ollama

A tool for running large language models locally on your own hardware. Makes it easy to deploy and manage AI models without cloud dependencies.

P

Prompt Engineering

The practice of crafting effective instructions to AI models to elicit desired responses. Critical for getting reliable results from agents.

R

RAG (Retrieval-Augmented Generation)

A technique that combines LLM generation with external knowledge retrieval to provide more accurate, up-to-date, and contextually relevant responses.

S

Sandboxing

Isolating an application in a restricted environment to prevent it from accessing sensitive system resources or causing unintended side effects.

Self-Hosted

Running software on your own infrastructure (VPS, dedicated server, or local hardware) rather than relying on third-party cloud services.

SSH (Secure Shell)

A protocol for securely accessing remote servers. Used to manage and configure self-hosted AI systems without physical access to the hardware.

T

Tailscale

A secure networking tool that creates encrypted mesh VPNs between devices. Used for safe remote access to self-hosted services without exposing ports.

Token

The basic unit of text processing in LLMs. Roughly 4 characters or 0.75 words in English. Pricing is often per 1K or 1M tokens.

V

Vector Database

A specialized database that stores and searches data as mathematical vectors. Essential for RAG systems to quickly find relevant context.

VPS (Virtual Private Server)

A virtualized server that provides dedicated resources within a shared physical machine. Common providers: Hetzner, Contabo, DigitalOcean.

W

Workflow

A sequence of automated steps that process information or complete tasks. AI agents can orchestrate complex workflows across multiple tools.