Skip to content
Nappai Documentation

General AI and Automation Concepts

This page serves as a reference glossary for the most common concepts in the world of artificial intelligence and automation that you will apply within NappAI.

Vector Store

A Vector Store (or vector database) is a type of database designed to store “embeddings” (numerical representations of information). It is used in conjunction with retrieval components to create a knowledge base that your AI Agent can access to answer specific questions.

API

An API (Application Programming Interface) offers programmatic access to a service’s data and functionalities. APIs allow NappAI to interact with external systems in an automated way.

AI Chain

A “Chain” refers to the connection of multiple components in a logical sequence within a Flow. It enables multi-step operations where the output of one component becomes the input of the next. In NappAI, a Flow is essentially a chain (or a more complex graph) of components.

Grounding

In the context of RAG (Retrieval-Augmented Generation), “grounding” measures how accurately a model’s response is based on the information from the provided source documents. A well-”grounded” response is faithful to the source, while an ungrounded response may be a “hallucination.”

Completion

A “Completion” is simply the response generated by an AI model (like Gemini or GPT) after receiving an input (a “prompt”).

Retrieval-Augmented Generation (RAG)

RAG is a technique that gives LLMs access to external and up-to-date information. A RAG system first “retrieves” relevant documents from a knowledge base (like a Vector Store) and then uses that information to “generate” an accurate and contextualized response. NappAI provides the necessary components to build powerful RAG flows.

[Image of Retrieval-Augmented Generation (RAG) pipeline architecture]

Hallucination

A “hallucination” occurs when an LLM generates information that is incorrect, fabricated, or not based on the provided input data.

AI Tool

Within an AI Agent, a “Tool” is a resource that the agent can decide to use to fulfill a task. A tool can be another component (like a calculator or a web search engine), an external API, or even another NappAI Flow.

Embedding

An “Embedding” is a numerical representation of data (text, images, etc.) in the form of a vector. AI uses these vectors to understand complex relationships and similarities between different pieces of information.

LangChain

LangChain is a very popular AI development framework that simplifies the creation of applications with Large Language Models. NappAI integrates many concepts and capabilities of LangChain to offer powerful and flexible components such as Agents.

Memory

Memory allows an AI Agent to remember the context of past interactions in a conversation. This is crucial for maintaining coherent dialogues without needing to resend the entire history with every new message. In NappAI, Agents are the components designed to use memory natively.

Large Language Model (LLM)

An LLM is an AI model trained on enormous amounts of text data to understand and generate human language. Examples of LLMs you can use in NappAI are the Gemini family of models from Google, GPT from OpenAI, etc.

Re-ranking (Reclasificación)

In a search or RAG system, “re-ranking” is a technique used to take an initial list of retrieved documents and reorder it to place the most relevant results at the beginning, thus improving the quality of the information delivered to the LLM.

Base64

Base64 is a method for encoding data and converting it into a plain text format.

  • Analogy: Imagine you want to send a liquid (like water) by mail. You can’t simply pour the water into the envelope. First, you have to put the water in a sealed bottle (encode it in Base64). The bottle is now a solid object that can be easily sent. Upon arrival, the recipient opens the bottle and retrieves the water.
  • In NappAI: We use Base64 to “bottle” files (which are binary data) and send them securely across the internet and between flow components.

Binary

Binary data is the most fundamental form of information a computer can understand, composed entirely of 1s and 0s. Every file on your computer—a PDF, an image, a video—is, at its lowest level, a binary file.

  • Analogy: If a text document is the cooked meal, binary data is the raw and exact ingredients that compose it.
  • In NappAI: When a component “processes” a file, it is working with this binary data to read, interpret, or modify its content.