Tech Dictionary

A token is a chunk of text (a word, part of a word, or punctuation) that a language model reads and generates; models process text as sequences of tokens, not raw characters.

also context length

The context window is the maximum number of tokens a model can consider at once, including both your input prompt and its generated output.

Temperature is a setting (typically 0 to 1) that controls randomness in a model's output; lower values make responses more focused and deterministic, higher values make them more varied and creative.

also nucleus sampling

Top-p is a sampling method where the model only chooses from the smallest set of next tokens whose combined probability exceeds a threshold p, trimming off unlikely options.

Inference is the process of running a trained model to generate an output from an input, as opposed to training the model in the first place.

Fine-tuning is further training a pre-trained model on a specific dataset so it specializes in a particular task, tone, or domain.

A hallucination is when a model generates information that sounds plausible but is factually wrong or made up.

also vector embedding

An embedding is a list of numbers (a vector) that represents the meaning of a piece of text, so that similar meanings produce numerically similar vectors.

also vector store

A vector database stores embeddings and lets you quickly find the items whose vectors are most similar to a query vector.

Semantic search finds results based on meaning and intent rather than exact keyword matches, usually by comparing embeddings.

also base model

A foundation model is a large model trained on broad data that can be adapted to many tasks, such as Claude, GPT, or Gemini.

A multimodal model can understand or generate more than one type of input or output, such as text plus images, audio, or documents.

The knowledge cutoff is the date after which a model has no built-in knowledge of world events, because its training data stops there.

also system message

A system prompt is a top-level instruction that sets a model's role, rules, and behavior before the user's messages are processed.

Zero-shot prompting asks a model to perform a task using only instructions, without giving any examples of the desired output.

Few-shot prompting gives the model a handful of example input-output pairs in the prompt to show it exactly what you want.

also CoT

Chain-of-thought prompting asks the model to reason step by step before giving a final answer, which improves accuracy on complex tasks.

Prompt chaining breaks a task into multiple sequential prompts, where each step's output feeds the next, instead of one giant prompt.

Prompt injection is an attack where malicious text in the model's input tricks it into ignoring its instructions or performing unintended actions.

also JSON mode

Structured output is when a model returns data in a strict, machine-readable format like JSON that matches a defined schema.

Grounding is supplying a model with trusted source data in the prompt and instructing it to answer only from that data.

Context engineering is the practice of deciding what information, instructions, and examples to place in a model's context window, and how to structure them, for the best results.

An AI system that uses an LLM to decide and take actions in a loop, calling tools and reacting to results to accomplish a goal with minimal step-by-step human direction.

also Function Calling

A capability where the LLM outputs a structured request to invoke a defined function or tool, and your code runs it and returns the result for the model to use.

also Model Context Protocol

An open standard from Anthropic that defines how AI applications connect to external tools, data, and context through a consistent client-server protocol.

A program that exposes tools, resources, and prompts to AI applications over the Model Context Protocol so a client can discover and use them.

The component inside an AI application that connects to MCP servers, discovers their tools and resources, and relays calls between the model and the server.

The communication layer that carries MCP messages between client and server, typically local stdio or a network-based streamable HTTP transport.

One of the core building blocks an MCP server can expose: tools (actions the model can call), resources (readable data), and prompts (reusable templates).

An MCP feature that lets a server request a completion from the client's LLM, so the server can use model reasoning without holding its own API key.

also Retrieval-Augmented Generation

A technique that retrieves relevant documents from a knowledge source and inserts them into the prompt so the model answers from that grounded context.

A RAG approach where an agent actively decides when and how to retrieve, choosing sources, issuing multiple searches, and refining queries instead of doing a single fixed lookup.

also Agent2Agent

An open protocol for agent-to-agent communication that lets independent AI agents from different vendors discover each other and collaborate on tasks.

The coordination of multiple steps, tools, or agents into a controlled workflow that defines order, branching, and data flow toward a goal.

A system of multiple specialized agents that work together, often coordinated by an orchestrator, each handling a sub-part of a larger task.

A secondary agent invoked by a primary agent to handle a focused subtask, often with its own tools and isolated context, returning results to the caller.

Rules, checks, and constraints placed around an AI system to keep its inputs and outputs safe, on-policy, and within allowed actions.

also Evaluations

Structured tests that measure an AI system's quality on defined tasks using example inputs and scoring criteria, used to compare prompts, models, or versions.

A Claude capability where the model controls a computer by viewing screenshots and issuing mouse and keyboard actions to operate software through its graphical interface.

The distinction between memory that lasts only within one conversation or run (session) and memory stored across sessions (persistent) so the agent recalls past interactions later.

also HITL

A design pattern where a person reviews, approves, or corrects an AI system's output or actions at key points before they take effect.

also Opus · Sonnet · Haiku

Anthropic's three Claude model families balancing capability, speed, and cost: Opus is the most capable, Sonnet is balanced, and Haiku is the fastest and cheapest.

A Claude.ai feature that groups chats with shared context, custom instructions, and uploaded knowledge so conversations within the project draw on the same background.

A Claude.ai feature that renders generated content like code, documents, or web apps in a dedicated side panel you can view, edit, and iterate on live.

Anthropic's agentic coding tool that runs in the terminal and can read, write, and edit files, run commands, and use tools to complete software tasks in your codebase.

also Agent Skills

Reusable packaged instructions and resources (Agent Skills) that Claude loads on demand to perform specialized tasks, bundling guidance and optional scripts in a folder.

Prebuilt integrations that connect Claude to external apps and services through MCP, letting it access tools and data from those systems within Claude products.

Anthropic's training method that aligns models using a written set of principles (a constitution), having the model critique and revise its own outputs against those rules rather than relying only on human labels.

A Claude API feature that stores a reusable prefix of a prompt so repeated requests skip reprocessing it, cutting latency and cost for that cached portion.

A Claude API mode for submitting many requests together to be processed asynchronously at lower cost, with results returned within a target window rather than instantly.

also HTTP callback

A user-defined HTTP request that one app automatically sends to another's URL when an event happens, pushing data in real time instead of being polled.

also REST

An interface style where you access and modify resources over HTTP using methods like GET, POST, PUT, and DELETE, typically exchanging JSON.

Two ways to get new data: polling repeatedly checks a source on a schedule, while push (webhooks) has the source send data the instant an event occurs.

also request body

The actual data carried inside an HTTP request or webhook, usually a JSON body containing the event's details.

also JavaScript Object Notation

A lightweight, text-based data format using key-value pairs and arrays that is the standard way APIs and webhooks exchange structured data.

also event

The starting event that kicks off an automation, such as a new form submission, an incoming webhook, or a scheduled time.

also step

A step an automation performs after the trigger, such as creating a record, sending a message, or calling an API.

also condition

A step that allows an automation to continue only when specified conditions are met, otherwise stopping that run.

also loop

A module that splits an array into separate items so each one is processed individually by the following steps.

also bundle merger

A module that combines multiple separate items back into a single structure, such as one array, text string, or numeric total.

also branch

A step that splits an automation into multiple paths, each with its own conditions and actions, run in parallel or by priority.

also throttling

A cap an API sets on how many requests you can make in a given time window, beyond which requests are rejected or delayed.

also OAuth 2.0

An authorization standard that lets an app access your account on another service via tokens, without sharing your password.

also secret key

A secret string that identifies and authenticates your application to an API, usually sent in a header with each request.

also idempotent request

A property where making the same request multiple times produces the same result as making it once, preventing duplicate side effects.

also error handling

Logic that detects failed steps and decides whether to retry, route to an error path, or stop, so automations fail gracefully.

A no-code automation platform that connects thousands of apps through trigger-and-action workflows called Zaps.

A single automated workflow in Zapier, made up of one trigger and one or more action steps.

In Zapier, a task is one action a Zap successfully completes, and it is the unit Zapier bills against your plan.

A Zapier feature that adds conditional branches to a Zap, running different action sequences based on if/then rules.

also Code by Zapier

A Zapier action that runs custom JavaScript or Python to transform data or perform logic that built-in steps can't.

also Integromat

A visual no-code automation platform (formerly Integromat) where you build workflows called scenarios from connected modules on a canvas.

In Make, a scenario is a complete automation workflow made of connected modules, with its own schedule and run history.

In Make, a module is a single step in a scenario that performs one app action, such as a trigger, search, or create.

In Make, an operation is a single execution of a module, and it is the unit Make bills against your plan.

In Make, a bundle is a single packet of data passed between modules, roughly equivalent to one record or item.

In Make, a data store is a built-in lightweight database for saving and retrieving data across scenario runs.

A source-available, node-based workflow automation tool that can be self-hosted and supports custom code and AI agent workflows.

In n8n, a node is a single step in a workflow that performs an action, trigger, or logic operation on the data passing through.

In n8n, an expression is a snippet of JavaScript wrapped in {{ }} that dynamically computes a field value from incoming data.

also sub-workflow

In n8n, a sub-workflow is a separate workflow called by another via the Execute Workflow node, enabling reusable, modular logic.

In n8n, credentials are securely stored authentication details (API keys, OAuth tokens) that nodes reuse to connect to services.

In n8n, items are the individual data objects flowing between nodes; most nodes run once per item, processing them as a list.

also AI Agent

An n8n node that runs an LLM-powered agent which can reason and call connected tools, memory, and models to complete tasks.

also LC · LeadConnector

GoHighLevel's white-label communication infrastructure that powers SMS, calls, and email inside the platform, billed through GHL rather than a separate Twilio/Mailgun account.

also LeadConnector Phone

GoHighLevel's built-in phone system (a managed Twilio integration) used to buy numbers, send/receive SMS, and make calls directly within a sub-account.

also 10DLC · Application-to-Person 10DLC

A US carrier registration standard for sending application-to-person SMS over standard 10-digit long codes, requiring brand and campaign registration to avoid filtering.

also email authentication · SPF · DKIM · DMARC

Three DNS-based email authentication standards that verify a sender is authorized for a domain, sign messages to prove they weren't altered, and tell receivers how to handle failures.

also SaaS

A GoHighLevel agency feature that lets you resell the platform as your own branded software, automatically provisioning sub-accounts and billing clients via Stripe with usage-based rebilling.

also usage rebilling · reselling

Charging sub-accounts a marked-up rate for usage-based services like SMS, calls, email, and AI by reselling the underlying GHL/Twilio/Mailgun consumption at your own price.

also agency account · location · sub-account

GoHighLevel's two-tier structure where the agency (top-level) account manages settings, snapshots, and billing across many client sub-accounts (locations), each an isolated CRM workspace.

also Conversation AI Bot

GoHighLevel's AI chatbot that auto-replies to leads across SMS, web chat, and other channels, with modes ranging from suggestive to fully autonomous booking based on a trained knowledge base.

also AI Voice Agent

GoHighLevel's AI agent that answers and handles inbound phone calls in a natural voice, able to respond to questions, qualify callers, and book appointments using a configured knowledge base and actions.

also AI content writer

GoHighLevel's generative text and image tool that drafts marketing content such as emails, social posts, and funnel copy directly inside the platform.

also AI workflow actions

AI-powered actions available inside GoHighLevel workflows that let an automation call a model to generate text, make decisions, or process data mid-flow as a step.

also GHL Agent Studio

GoHighLevel's framework for building, configuring, and managing custom AI agents with defined goals, knowledge, and actions that operate across channels like chat and voice.

also If/Else · branch

A GoHighLevel workflow action that splits the automation path into different branches based on whether contact data or conditions match defined criteria.

also delay · wait

A GoHighLevel workflow action that pauses a contact's progress for a set time, until a specific date/time, or until a condition or event is met before continuing.

also goal · goal trigger

A GoHighLevel workflow element that lets a contact jump ahead to a designated point when a desired action occurs, skipping remaining steps once the goal is achieved.

also GHL webhook action · inbound webhook

GoHighLevel's workflow capability to send outbound HTTP requests to external URLs or receive inbound webhook triggers, passing JSON data between GHL and other systems.

also client-side vs server-side

Frontend is the part of an app users see and interact with in the browser, while backend is the server-side logic, data, and APIs that power it behind the scenes.

also endpoint

A specific URL on a server that accepts requests and returns data or performs an action, representing one accessible operation of an API.

also verbs · HTTP verbs

Standard request types that tell a server what action to perform, most commonly GET (read), POST (create), PUT/PATCH (update), and DELETE (remove).

also response codes

Three-digit numbers a server returns to indicate a request's outcome, grouped as 2xx success, 3xx redirect, 4xx client error, and 5xx server error.

also env vars · .env

Named values stored outside your code and read at runtime, used to hold configuration and secrets like API keys without hardcoding them.

also version control

A distributed version-control system that tracks changes to files over time, letting you commit snapshots, branch, and collaborate without losing history.

also deploy · release

The process of taking code from development and making it live and accessible to users on a server or hosting platform.

also continuous integration · continuous delivery

Continuous Integration and Continuous Delivery/Deployment: automated pipelines that build, test, and release code whenever changes are pushed.

also Next

A React framework for building full-stack web apps with server-side rendering, file-based routing, and built-in API routes.

A cloud hosting platform optimized for frontend frameworks like Next.js, providing automatic Git-based deployments, global CDN, and serverless functions.

also Domain Name System

The system that translates human-readable domain names into IP addresses and stores records (A, CNAME, MX, TXT) that route traffic and verify domains.

also Content Delivery Network

A geographically distributed network of servers that caches and serves content from a location near each user to speed up load times.

also Cross-Origin Resource Sharing

A browser security mechanism that controls whether a web page from one origin is allowed to make requests to a server on a different origin, governed by server response headers.

also HyperText Markup Language

The markup language that defines the structure and content of a web page — headings, paragraphs, buttons, images, and links.

also Cascading Style Sheets

The language that styles a web page — colors, fonts, spacing, layout, and animations.

also JS

The programming language of the browser that makes pages interactive — clicks, forms, popups, and live updates.

also Tailwind

A utility-first CSS framework where you style elements by composing small classes directly in your markup instead of writing separate CSS files.

A JavaScript library by Meta for building user interfaces out of reusable components that update automatically when data changes.

also Node

A runtime that lets JavaScript run on a server instead of only in the browser, powering the backend of most JS apps.

also Application Programming Interface

A defined way for two pieces of software to talk to each other and exchange data or trigger actions.

A query language for APIs that lets the client ask for exactly the data it needs in a single request, instead of fixed endpoints returning fixed shapes.

An open-source backend platform giving you a Postgres database, auth, file storage, and auto-generated APIs in one dashboard.

A TypeScript ORM that lets your app read and write a database with type-safe code instead of raw SQL.

also Object Relational Mapper

A tool that maps database tables to objects in your code so you can query data in your programming language instead of writing SQL.

A tool that packages an app with all its dependencies into a container that runs identically on any machine.

A cloud platform for hosting Git repositories, collaborating on code, and triggering deployments.

also cloud computing · AWS · GCP · Azure

The big cloud platforms (Amazon, Google, Microsoft) that rent powerful servers and services to run large applications.

also Command Line Interface · terminal

A text-based way to control your computer or a tool by typing commands instead of clicking a graphical interface.

also Ext · plugin

An add-on that installs new features into a code editor like VS Code.

also Large Language Model

A large AI model trained on massive text to understand and generate human-like language.

also open weights

An AI model whose weights are publicly released so anyone can download, run, and self-host it.

also proprietary model

An AI model you access through an API or app but cannot download or see inside.

A benchmark that tests how well AI models solve real software bugs from open-source projects, widely used to rank coding ability.