GLOSSARY.
Semantic dictionary and quick references containing key concepts on artificial intelligence, RAG integration, MCP protocol, Next.js development, and Core Web Vitals.
RAG (Retrieval-Augmented Generation)
AI technique that optimizes Large Language Model (LLM) responses by retrieving information from a secure external database before generating the answer.
// Detalhe de Engenharia:RAG anchors the model's response in actual company facts (like manuals, ERP, and CRM), reducing the probability of hallucinations (factual errors) to practically zero.
MCP (Model Context Protocol)
Open protocol developed by Anthropic that establishes a universal and secure standard for connecting Artificial Intelligence models to data sources and software tools.
// Detalhe de Engenharia:Acts as a 'USB' standard for AIs, allowing a single MCP server to provide context and tools to different AI clients without the need for bespoke proprietary integrations for each LLM.
LLM (Large Language Model)
Deep learning Artificial Intelligence models trained on vast textual datasets to understand, generate, translate, and reason in natural language.
// Detalhe de Engenharia:Prominent examples include ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), and Llama (Meta), serving as the core cognitive processing tools in modern systems.
Vector Database (Banco de Dados Vetorial)
Database designed to store and search embeddings (mathematical vectors representing the semantic meaning of content).
// Detalhe de Engenharia:Tools like Pinecone, Qdrant, and pgvector allow semantic searches to be performed in milliseconds, finding document snippets conceptually similar to the user's question.
Embedding (Vetorização)
The process of converting text, images, or audio into sequences of numbers (vectors) that capture the conceptual meaning and context of the information.
// Detalhe de Engenharia:Thanks to embeddings, a search for 'development pricing' can return an article about 'software costs', as the algorithm understands the semantic equivalence.
Prompt Injection (Injeção de Prompt)
A security vulnerability where a malicious user manipulates AI inputs to make it ignore its original guidelines and expose secrets or execute forbidden actions.
// Detalhe de Engenharia:Shielding systems against Prompt Injection involves creating intermediary validation architectures and sanitizing all input strings before processing by the LLM.
DOM Bloating (Inchaço do DOM)
An excess of nested and unnecessary elements in the HTML tree of a web page, usually generated by visual builders or pre-made templates.
// Detalhe de Engenharia:Increases browser processing time and delays loading, harming Core Web Vitals scores (such as LCP) and Google Lighthouse ratings.
Headless Commerce (Comércio Desacoplado)
An e-commerce architecture where the front presentation layer (Frontend) is completely separated from the backend e-commerce logic (Backend).
// Detalhe de Engenharia:Communication occurs via fast APIs. Allows creating ultra-fast custom interfaces in Next.js while robust tools like VTEX or Shopify manage the shopping cart.
Core Web Vitals
A set of web performance metrics established by Google that assess the actual user experience in terms of loading speed, interactivity, and visual stability.
// Detalhe de Engenharia:Composed mainly of LCP (loading), INP (click interactivity), and CLS (visual stability). It is a direct SEO ranking factor.
LCP (Largest Contentful Paint)
Core Web Vitals metric that measures the time required to render the largest visual element (usually a hero image or title block) on the user's screen.
// Detalhe de Engenharia:To ensure a good experience and SEO score, LCP must happen in less than 2.5 seconds. At GHWD, we optimize code to achieve LCP under 1.0 second.
INP (Interaction to Next Paint)
Metric that evaluates the interactivity latency of a web page, measuring the time elapsed from the user's click to the subsequent visual update on the screen.
// Detalhe de Engenharia:Replaced the old FID in 2024. A good INP value is below 200 milliseconds. It is crucial for interactive systems like dashboards and e-commerces.
CLS (Cumulative Layout Shift)
Core Web Vitals metric that measures the visual instability of a page, observing if textual elements or buttons change places unexpectedly during loading.
// Detalhe de Engenharia:Unstable layouts cause accidental clicks and frustration. A good CLS index should be less than 0.1 and is avoided by reserving fixed sizes for media and fonts.
SSG (Static Site Generation)
Process of compiling web pages into pure static HTML files during build time, before the user accesses them.
// Detalhe de Engenharia:Eliminates the need for real-time database queries, allowing pages to be served via CDN Edge in milliseconds with unmatched performance.
SSR (Server-Side Rendering)
Rendering strategy where the Next.js server generates the dynamic HTML of a page on-demand for each user request.
// Detalhe de Engenharia:Useful for personalized pages, containing dynamic account or cart data, which still require SEO indexing on the first load.
API (Application Programming Interface)
Set of rules and specifications that allows different software applications to communicate and exchange data with each other.
// Detalhe de Engenharia:At GHWD, we create secure REST and GraphQL APIs to integrate legacy ERPs, commercial CRMs, and databases into Next.js interfaces and AI pipelines.
Webhook
An HTTP communication pattern that sends real-time data from one system to another immediately after a specific event occurs.
// Detalhe de Engenharia:Example: When an invoice is paid, the gateway sends a Webhook containing payment data to notify GHWD's automation to release the product automatically.
n8n
A powerful fair-code workflow automation tool focused on API integration and complex data flows.
// Detalhe de Engenharia:Can be hosted on the company's own infrastructure (on-premise), ensuring data security and eliminating variable costs based on volume of executions.
Schema.org
A collaborative, structured vocabulary of semantic tags and markups that developers add to HTML to help search engines understand the meaning of data.
// Detalhe de Engenharia:Essential for SEO and GEO, as they allow search engines like Google and ChatGPT to extract prices, dates, FAQs, and authors in a 100% structured format.
Token
The basic unit of text that an Artificial Intelligence model processes. It can be a word, part of a word, or special character.
// Detalhe de Engenharia:AI models charge per thousand tokens consumed. Optimizing prompts and RAG contexts aims to reduce token waste and lower the monthly bill.
Serverless (Sem Servidor)
Cloud execution model where the provider automatically manages the allocation of physical servers to run small code functions on-demand.
// Detalhe de Engenharia:Eliminates idle server costs and ensures unlimited scalability. If the site has zero visits, the cost is zero; if it has millions, it scales without crashing.