Acronyms and Terminologies

This section provides definitions and explanations for key terms and acronyms used throughout the GLChat to help users, both technical and non-technical, to understand the platform’s components, features, and configuration settings more easily. Refer to this page whenever you encounter unfamiliar terminology in the documentation.

Term/Acronym	Description
A2A (Agent-to-Agent)	A system where one AI agent can work together with another, sharing tasks or passing information to complete complex actions more effectively.
API (Application Programming Interface)	A way for different software systems to talk to each other and exchange information or functions. It helps apps work together behind the scenes.
API Key	A unique code used to securely connect one system or app to another. It works like a password for software, allowing access while keeping things protected.
Augment	Enhancing the system's ability by adding extra information or support from other tools or data.
Batch Size	The number of items processed by the system at once, like documents or data chunks.
Bearer Token	A secure access key used to verify a user's identity when making system requests.
Blacklist	A list of blocked or restricted websites that are given special permission to access something.
Cache	Temporary storage that helps speed up repeated tasks or data retrieval.
Char (Character)	Each letter, number, space, or symbol in a text.
Chunker	A tool that breaks long text into smaller pieces (or “chunks”) to help the AI process the content more effectively and stay within its limits.
Docproc (Document Processing)	Systems that read, understand, and extract data from documents.
DPO (Document Processing Orchestrator)	A process that helps prepare and organize documents so they can be used as reliable sources of information in the system.
Email Domain	The part of an email address that comes after the “@” symbol (e.g., in [e-mail], the domain is example.com).
Formatter	A tool that changes how information is displayed or structured, so it looks cleaner or follows a standard.
Graph RAG Indexer	A smart indexing tool that uses a graph (network-like structure) to help AI find and retrieve better answers.
HTTP (HyperText Transfer Protocol)	The standard way data is sent and received over the web.
Index Name	The name of the vector index where the knowledge base content is stored. It’s auto-generated by default but can be customized if needed.
Knowledge Base	A collection of information or documents that the chatbot can refer to when answering questions.
Kwargs (Keyword Arguments)	Extra options or settings you can include when using a feature, by naming them clearly. It helps the system understand exactly what you want to change or customize.
LLM (Large Language Model)	A type of Language Model that has been trained with a huge amount of text data, making it much more powerful and capable. It can handle complex instructions, answer detailed questions, and hold natural conversations.
LLM Labs	A platform used to test, manage, and connect AI models (like LLMs) to other parts of the system, such as document storage or search tools.
LM (Language Model)	An AI model that can understand and generate text, like answering questions or continuing a sentence. It can be small or large depending on how much it has been trained.
LM Invoker	A component that takes your prompt (question or instruction) and sends it to the AI model (LM), then returns the response.
LMRP (Language Model Request Processor)	A module that helps the AI prepare your question, send it to the language model, and clean up the answer before showing it to you. It ensures the whole process, from asking to answering, is smooth and accurate.
MCP (Model Context Protocol)	A system that allows the chatbot to use extra tools, like custom code or functions, to perform more advanced tasks.
Model	In this context, model refers to a language model, the part of the AI that has learned to understand and respond to human language based on large amounts of training data.
Multimodal	A system that can understand and process more than one type of input, such as text, images, or audio.
OAuth (Open Authorization)	A secure way to connect one app to another without sharing your password. For example, an admin can approve access through a permission screen to link GLChat with services (e.g. Google Workspace).
PII (Personally Identifiable Information)	Sensitive data like names, emails, phone numbers, etc., which can be anonymized.
PII Anonymization	The process of hiding or replacing personal information (like names, phone numbers, or ID numbers) to protect user privacy before the data is processed or stored.
Pipeline	A step-by-step process the system follows to handle a chat or a document, from receiving input to giving a response.
Preset	A saved configuration that controls how the system behaves, including settings like supported file types, privacy rules, and search options.
Prompt	The message or instruction you type to the chatbot, or the internal guide the system uses to decide how to respond.
RAG (Retrieval-Augmented Generation)	A method that enhances LLM responses with relevant external information from a knowledge base.
RAGO (Retrieval-Augmented Generation Orchestrator)	A GLChat pipeline type used to configure how information is retrieved and presented during chat interactions.
Rerank	Re-sorting search results to show the most relevant ones at the top.
SSE (Server-Sent Events)	A way for the system to send real-time updates from the server to the user.
SSO (Single-Sign On)	A login method that lets users access multiple systems using one account, usually their work email.
Threshold	A set limit used by the system to decide whether something is accepted or rejected.
Timeout	The maximum time allowed for an AI or tool to respond. If it takes too long, the system will stop waiting and cancel the request.
Transport Type	The method used to deliver responses from the system to the user (e.g., HTTP or SSE).
Vector Store	A type of database where document meanings (not just words) are stored in a way the AI can search by understanding the context. It helps the system find the most relevant answers quickly.
Vector Weight	A value showing how important a piece of information is when comparing or searching using AI.
Vectorizer	A tool that converts text into a numeric format (called vectors) so that the AI can understand and work with the content mathematically.
Whitelist	A list of allowed websites that are given special permission to access something.