Glossary

A2A Protocol (Agent-to-Agent Protocol) is an open interoperability standard initiated by Google that...

Adapter is a small trainable module added to a pretrained neural network so the model can be adapted...

AES (Advanced Encryption Standard) is a symmetric block cipher algorithm adopted by the U.S. governm...

Agent Development Kit is a framework layer that provides reusable abstractions for building AI agent...

Agent Harness is the runtime control layer around an AI agent that constrains execution, manages too...

Agent Memory refers to the systems and mechanisms that enable AI agents to store, retrieve, and util...

Agent Runtime is the execution environment that manages an AI agent session, including state, tool c...

Agent Trajectory is the ordered record of an AI agent run, including observations, messages, decisio...

Agentic Workflow is a design pattern where AI agents autonomously plan, execute, and iterate on comp...

AGI (Artificial General Intelligence) is a theoretical form of artificial intelligence that possesse...

AI Agent is an autonomous software system powered by Large Language Models, implementing goal-orient...

AI App Builder is a category of no-code/low-code platforms that use large language models to generat...

AI Code Review is an automated code review technique that integrates artificial intelligence (specif...

AI Watermark is a technique for embedding imperceptible signals into AI-generated content (text, ima...

API (Application Programming Interface) is a set of rules, protocols, and tools that allows differen...

An API Gateway is a server component that acts as the single entry point for all client API requests...

Approval Gate is a policy checkpoint that pauses or blocks an AI action until a human reviewer, rule...

Artificial Intelligence (AI) is a branch of computer science that focuses on creating intelligent sy...

ASCII (American Standard Code for Information Interchange) is a character encoding standard that ass...

Aspect Ratio is the proportional relationship between the width and height of an image, video, or di...

Attention Mechanism is a neural network technique that enables models to dynamically focus on releva...

AutoGen is an open-source framework for developing Large Language Model (LLM) applications. Its core...

Autonomous Driving is a technology that enables vehicles to navigate and operate without human inter...

AWQ (Activation-aware Weight Quantization) is a weight-only quantization method that identifies and ...

Backpropagation is a fundamental algorithm for training artificial neural networks that efficiently ...

Barcode is a machine-readable representation of data using patterns of parallel lines (1D) or geomet...

Base64 is a binary-to-text encoding scheme that represents binary data in an ASCII string format by ...

Bearer Token is an access token type used in HTTP authentication where the client presents a token t...

Bi-Encoder is a retrieval model architecture that encodes queries and documents separately into embe...

Binary (also known as base-2) is a numeral system that uses only two symbols: 0 and 1. It is the fun...

BM25 is a probabilistic lexical ranking function that scores documents based on query term matches, ...

BMI is a numerical value calculated from a person's weight and height, used as a screening tool to c...

Callback Handler is a lifecycle hook component that receives events from model calls, tools, retriev...

Catastrophic Forgetting is the loss or degradation of previously learned capabilities when a model i...

CDN is a geographically distributed network of proxy servers and data centers that work together to ...

Chain Orchestration is a linear or mostly linear composition pattern where LLM application steps exe...

Chain-of-Thought (CoT) is a prompting technique that encourages large language models to break down ...

Chatbot is an artificial intelligence software application designed to simulate human-like conversat...

ChatModel is an application-layer abstraction for invoking conversational language models through st...

ChatTemplate is a reusable role-based message template that turns variables, instructions, examples,...

Checksum is a small-sized block of data derived from another block of digital data for the purpose o...

Chunk Overlap is the repeated text shared between adjacent document chunks so that information near ...

Chunk Size is the token, character, or structural length chosen for each document unit indexed in a ...

Chunking is the process of splitting long documents or data sources into smaller retrievable units t...

Citation is the attribution of an AI-generated claim to a specific source document, passage, URL, re...

CMYK is a subtractive color model used in color printing, where cyan, magenta, yellow, and black ink...

CNN (Convolutional Neural Network) is a class of deep neural networks specifically designed for proc...

Cold Start is the additional startup latency that occurs when a model service handles a request befo...

Computer Vision is a field of artificial intelligence that enables computers to interpret and unders...

Context Budget is the planned allocation of a model's limited context-window tokens across instructi...

Context Caching is the practice of reusing repeated prompt context or computed model state so an LLM...

Context Compression is the process of reducing the amount of context sent to an LLM while preserving...

Context Engineering is the process of precisely providing large models with the necessary and only n...

Context Precision is a RAG evaluation metric that measures how much of the retrieved context is rele...

Context Recall is a RAG evaluation metric that measures whether the retrieved context contains the e...

Context Window is the maximum number of tokens that a large language model can process in a single i...

Continuous Batching is an LLM serving technique that dynamically groups active requests during infer...

CORS (Cross-Origin Resource Sharing) is a security mechanism that allows web browsers to make reques...

CrewAI is an open-source Python framework specifically designed to coordinate multiple autonomous Ag...

Cron is a time-based job scheduler in Unix-like operating systems that enables users to schedule com...

Cross-Encoder is a ranking model architecture that jointly encodes a query and a candidate document ...

CSS (Cascading Style Sheets) is a stylesheet language used to describe the presentation of HTML and ...

CSV (Comma-Separated Values) is a plain text file format that stores tabular data using commas to se...

Data URL is a URI scheme that allows embedding small data items inline in web documents as if they w...

Dataset Curation is the process of selecting, cleaning, organizing, labeling, deduplicating, and val...

Decode Phase is the LLM inference stage that generates output one token at a time using the KV cache...

Deep Learning is a subset of machine learning that uses artificial neural networks with multiple lay...

Dense Retrieval is a semantic search method that represents queries and documents as dense embedding...

Diff is a comparison technique that identifies and displays the differences between two sets of data...

Diffusion Model is a class of generative deep learning models that learn to generate data by gradual...

Distillation (Knowledge Distillation) is a model compression technique where a smaller 'student' mod...

DiT (Diffusion Transformer) is a generative model architecture that replaces the traditional U-Net b...

DNS (Domain Name System) is a hierarchical and distributed naming system that translates human-reada...

Document Loader is an ingestion component that reads raw content from files, web pages, object stora...

Document Transformer is a pipeline component that cleans, splits, enriches, filters, or restructures...

DPO (Direct Preference Optimization) is a simplified approach to aligning language models with human...

Eino is an open-source Go framework in the CloudWeGo ecosystem for building production-oriented LLM ...

Embedding is a technique in machine learning that transforms discrete data such as words, sentences,...

Emoji is a standardized set of pictographic symbols and icons encoded in Unicode, used in digital co...

Encryption is the process of converting plaintext data into an unreadable format (ciphertext) using ...

Epoch time refers to the reference point (January 1, 1970, 00:00:00 UTC) from which Unix timestamps ...

Escape Character is a character that invokes an alternative interpretation of subsequent characters ...

EXIF is a standard that specifies the formats for images, sound, and ancillary tags used by digital ...

Few-Shot Learning is a machine learning paradigm where models learn to perform tasks from only a sma...

Fine-tuning is a transfer learning technique that adapts a pre-trained machine learning model to a s...

Function Calling is a capability of large language models that allows them to generate structured ou...

GAN (Generative Adversarial Network) is a class of deep learning models consisting of two neural net...

Generative AI is a category of artificial intelligence systems capable of creating new content—inclu...

Geohash is a geocoding system that encodes geographic coordinates (latitude and longitude) into a sh...

GIF (Graphics Interchange Format) is a bitmap image format developed by CompuServe in 1987 that supp...

Glob is a pattern matching syntax used to specify sets of filenames or paths using wildcard characte...

Go (also known as Golang) is a statically typed, compiled programming language designed at Google. I...

Golden Dataset is a curated set of trusted examples used as a stable reference for evaluating model,...

GPT (Generative Pre-trained Transformer) is a family of large language models developed by OpenAI th...

Gradient Descent is a first-order iterative optimization algorithm used to find the minimum of a fun...

Graph Orchestration is a control-flow pattern that represents AI application logic as nodes and edge...

GraphRAG (Graph Retrieval-Augmented Generation) is an advanced AI retrieval architecture. It uses LL...

Grounding is the practice of tying an AI system's outputs to retrieved evidence, verified sources, t...

GRPO is a reinforcement learning optimization method for language models that uses relative rewards ...

Guardrails are safety mechanisms and constraints implemented in AI systems to prevent harmful, inapp...

GUID (Globally Unique Identifier) is Microsoft's implementation of the UUID standard, a 128-bit iden...

Gzip is a file compression format and software application that uses the DEFLATE algorithm to reduce...

Hallucination is a phenomenon in AI systems where models generate information that appears confident...

Hash is a mathematical function that converts input data of any size into a fixed-size string of cha...

Hex Color is a six-digit alphanumeric representation of colors used in web design and digital graphi...

Hexadecimal (also known as hex or base-16) is a positional numeral system that uses 16 distinct symb...

HMAC (Hash-based Message Authentication Code) is a cryptographic technique that combines a secret ke...

HSL Color (Hue, Saturation, Lightness) is a cylindrical color model that represents colors in a way ...

HTML (HyperText Markup Language) is the standard markup language for creating web pages and web appl...

HTML Entity (HTML Character Entity) is a string that begins with an ampersand (&) and ends with a se...

HTTP (Hypertext Transfer Protocol) is the foundation protocol of the World Wide Web, defining how me...

HTTPS (Hypertext Transfer Protocol Secure) is the secure version of HTTP that encrypts all communica...

Human-in-the-Loop is a control pattern where human review, approval, correction, or escalation is in...

Hybrid Search is a technique in information retrieval and RAG (Retrieval-Augmented Generation) syste...

HyDE is a retrieval technique that asks a language model to generate a hypothetical document or answ...

Image Compression is the process of reducing the file size of digital images by removing redundant o...

In-Context Learning (ICL) is the ability of large language models to learn and adapt to new tasks fr...

Indexer is a pipeline component that writes processed documents, chunks, embeddings, metadata, or sp...

Inference (in machine learning) is the process of using a trained model to make predictions or gener...

Instruction Tuning is a supervised fine-tuning approach that trains a language model on diverse inst...

IP Address (Internet Protocol Address) is a unique numerical identifier assigned to every device con...

ISO 8601 is an international standard for representing dates and times in a clear, unambiguous forma...

Jailbreaking, in the context of Artificial Intelligence, refers to an advanced adversarial prompting...

JavaScript is a high-level, interpreted programming language that enables interactive web pages and ...

JPEG (Joint Photographic Experts Group) is a commonly used lossy compression method for digital imag...

JSON (JavaScript Object Notation) is a lightweight, text-based data interchange format that is easy ...

JSON Mode is an LLM API or decoding feature that constrains model responses so they are valid JSON o...

JSON Schema is a vocabulary that allows you to annotate and validate JSON documents. It describes th...

JSONPath is a query language for JSON documents that enables extracting and filtering data from comp...

JWT (JSON Web Token) is a compact, URL-safe means of representing claims to be transferred between t...

Knowledge Graph is a structured representation of real-world entities and their relationships, organ...

KTO is a preference tuning method that optimizes a language model using examples labeled as desirabl...

KV Cache (Key-Value Cache) is an optimization technique used in Transformer-based model inference th...

LangGraph is a Python/JS framework for building Stateful, Multi-Actor applications. As an extension ...

Latency is the elapsed time between a request and a response or milestone in an AI system, such as f...

Lint is a static code analysis warning or error generated by a linter to flag style violations, susp...

Linting is the process of using automated tools to analyze source code for potential errors, bugs, s...

LLM (Large Language Model) is a type of artificial intelligence model trained on massive amounts of ...

LLM-as-Judge is an evaluation technique that uses a large language model to assess, score, or compar...

LoRA (Low-Rank Adaptation) is a parameter-efficient fine-tuning technique that adapts large pre-trai...

LoRA Rank is the low-rank dimension used in LoRA adapters, controlling how much trainable capacity i...

Lorem Ipsum is a pseudo-Latin placeholder text commonly used in graphic design, web development, and...

Lost in the Middle is the tendency of language models to use information near the beginning or end o...

Machine Learning (ML) is a subset of artificial intelligence that enables computer systems to automa...

Markdown is a lightweight markup language created by John Gruber in 2004 for formatting plain text d...

MCP (Model Context Protocol) is an open protocol standard introduced by Anthropic in 2024, enabling ...

MCP App (Model Context Protocol Application) is an AI-native application built on the MCP standard t...

MCP Client is the protocol component, usually managed by an MCP Host, that establishes and maintains...

MCP Gateway is an intermediary layer that centralizes discovery, routing, authentication, authorizat...

MCP Host is the AI application or agent environment that owns the user session, presents MCP capabil...

MCP Prompt is a reusable prompt template exposed by an MCP Server so MCP clients and hosts can prese...

MCP Resource is a read-oriented context primitive exposed by an MCP Server, representing data such a...

MCP Sampling is a Model Context Protocol capability that allows an MCP Server to request language-mo...

MCP Server is a process or service that implements the Model Context Protocol and exposes capabiliti...

MCP Tool is an executable capability exposed by an MCP Server, described with a name, natural-langua...

MD5 (Message-Digest Algorithm 5) is a widely used cryptographic hash function that produces a 128-bi...

Metadata Filtering is the practice of restricting retrieval results using structured attributes atta...

MIME Type is a standardized identifier used to indicate the nature and format of a file or data, con...

Minification is the process of removing all unnecessary characters from source code without changing...

Mixture of Experts (MoE) is a neural network architecture that uses multiple specialized sub-network...

Mock data is artificially generated or simulated data used for testing, development, and demonstrati...

Model Alignment is the process of training AI systems to behave in accordance with human values, int...

Model Serving is the production practice of deploying machine learning or language models behind API...

Multi-Agent Systems are AI architectures where multiple autonomous agents collaborate, communicate, ...

Multimodal AI refers to artificial intelligence systems that can process, understand, and generate c...

NanoID is a tiny, secure, URL-friendly unique string ID generator for JavaScript. It generates compa...

Neural Network is a computational model inspired by the biological neural networks in the human brai...

NLP (Natural Language Processing) is a branch of artificial intelligence that focuses on enabling co...

OAuth (Open Authorization) is an open standard authorization protocol that allows users to grant thi...

Ollama is an open-source framework for running, building, and sharing Large Language Models (LLMs) o...

OpenTelemetry is an open-source observability framework that provides a unified set of APIs, SDKs, a...

ORPO is a preference optimization method that combines supervised learning on chosen responses with ...

Overfitting is a modeling error in machine learning that occurs when a model learns the training dat...

PagedAttention is an LLM serving technique that manages key-value cache memory in fixed-size blocks,...

Password is a secret string of characters used to authenticate a user's identity and grant access to...

PDF (Portable Document Format) is a file format developed by Adobe in 1992 to present documents cons...

PEFT (Parameter-Efficient Fine-Tuning) is a family of techniques that adapt large pre-trained models...

Planner-Executor is an AI agent architecture that separates high-level task decomposition and strate...

PNG (Portable Network Graphics) is a raster graphics file format that supports lossless data compres...

PPO is a reinforcement learning algorithm that updates a policy while limiting how far each update m...

Preference Data is training data that records which model responses are preferred, ranked, rejected,...

Prefill is the LLM inference phase that processes the full input prompt in parallel and produces the...

Prettify is the process of formatting code or data to make it more readable by adding proper indenta...

Prompt is a natural language input or instruction given to an AI model to guide its response, servin...

Prompt CI/CD is the application of continuous integration and deployment practices to prompt, templa...

Prompt Engineering is the practice of designing, refining, and optimizing input instructions (prompt...

Prompt Injection is a cybersecurity attack specifically targeting applications built on Large Langua...

Prompt Regression Test is an evaluation that checks whether a prompt or related LLM application chan...

Prompt Template is a reusable prompt structure that combines fixed instructions with variable inputs...

Prompt Versioning is the practice of tracking, reviewing, testing, and releasing changes to prompts ...

QLoRA (Quantized Low-Rank Adaptation) is an efficient fine-tuning technique that combines 4-bit quan...

QR Code (Quick Response Code) is a two-dimensional matrix barcode that can store various types of da...

Quantization is a model compression technique that reduces the precision of neural network weights a...

Query Rewriting is the process of transforming a user's original question into one or more clearer, ...

Query String is the part of a URL that contains data to be passed to web applications, appearing aft...

RAG (Retrieval-Augmented Generation) is an AI architecture that enhances large language model output...

Rate limiting is a technique used to control the number of requests a client can make to an API or s...

Reciprocal Rank Fusion is a rank aggregation method that combines multiple retrieval result lists by...

Recommendation System is an information filtering system that predicts and suggests items or content...

Red Teaming is a structured adversarial testing methodology where security experts deliberately atte...

Regex (Regular Expression) is a sequence of characters that defines a search pattern, primarily used...

Reinforcement Learning (RL) is a type of machine learning where an intelligent agent learns to make ...

Reranking is an advanced stage in information retrieval and RAG (Retrieval-Augmented Generation) pip...

REST (Representational State Transfer) is an architectural style for designing networked application...

Retriever is a query-to-context component that receives a user or agent query and returns relevant d...

Reward Model is a model trained to assign scores to candidate responses based on preference data or ...

RGB Color (Red Green Blue) is an additive color model where red, green, and blue light are combined ...

RGBA Color is an extension of the RGB color model that adds an alpha channel for opacity control, al...

RLHF (Reinforcement Learning from Human Feedback) is a training technique that aligns large language...

RNN (Recurrent Neural Network) is a class of neural networks designed to process sequential data by ...

Router Agent is an AI agent or routing component that classifies a request and delegates it to the m...

RSA (Rivest-Shamir-Adleman) is an asymmetric cryptographic algorithm that uses a pair of keys - a pu...

Semantic Search is an information retrieval technique that understands the meaning and intent behind...

Sentiment analysis is a natural language processing (NLP) technique used to identify and extract sub...

Serialization is the process of converting complex data structures or objects into a format that can...

SFT is a supervised training stage that fine-tunes a pretrained language model on curated prompt-res...

SHA-256 (Secure Hash Algorithm 256-bit) is a cryptographic hash function that produces a 256-bit (32...

A skill is a modular, reusable capability that extends an AI agent's functionality, enabling it to p...

Slug is a URL-friendly version of a string, typically derived from a title or name, that uses only l...

Small Language Model (SLM) is a category of language models with significantly fewer parameters than...

Sparse Retrieval is a lexical search method that represents queries and documents with sparse term-w...

Speculative Decoding is an LLM inference technique where a faster draft model proposes multiple cand...

Speech Recognition is a technology that enables computers to identify and convert spoken language in...

SQL (Structured Query Language) is a standardized programming language designed for managing and man...

SSE Transport is an HTTP-based Model Context Protocol transport pattern that uses Server-Sent Events...

SSL/TLS (Secure Sockets Layer / Transport Layer Security) are cryptographic protocols designed to pr...

STDIO Transport is a local Model Context Protocol transport that exchanges protocol messages between...

Structured Output is the practice of making an LLM return data in a predictable machine-readable for...

Supervised learning is a machine learning paradigm where models learn from labeled training data, us...

Supervisor Agent is a coordinating agent that assigns tasks, monitors progress, resolves conflicts, ...

SVG (Scalable Vector Graphics) is an XML-based vector image format for two-dimensional graphics that...

Synthetic Data is artificially generated data that mimics the statistical properties and patterns of...

System Prompt is the high-priority instruction layer that defines an LLM application's persistent ro...

Temperature is a hyperparameter in large language models that controls the randomness and creativity...

Tensor Parallelism is a model parallelism strategy that splits large neural network tensors and thei...

Test-Time Compute is a technique where AI models allocate additional computational resources during ...

Text-to-Image is an artificial intelligence technology that generates visual images from natural lan...

Text-to-Video is an AI generation technique that creates video content from natural language descrip...

Throughput is the amount of work a serving system completes per unit of time, such as requests per s...

A timestamp is a sequence of characters or encoded information identifying when a certain event occu...

Timezone is a region of the globe that observes a uniform standard time for legal, commercial, and s...

Token is the fundamental unit of text that Large Language Models (LLMs) process, representing a piec...

Tokenizer is the component that converts text into the token IDs a language model can process and de...

Tokens per Second is a throughput metric that measures how many output tokens an LLM generates per s...

TOML is a configuration file format designed to be easy to read due to its obvious semantics, mappin...

Tool Use is the capability of AI systems, particularly large language models, to interact with exter...

Training Data is the collection of labeled or unlabeled samples used to teach machine learning model...

Transformer is a deep learning architecture introduced in the landmark paper 'Attention Is All You N...

TTFT is the latency from sending an LLM request until the first generated token is received by the c...

TypeScript is a strongly typed programming language that builds on JavaScript by adding optional sta...

Unicode is a universal character encoding standard that assigns a unique number (code point) to ever...

Unix Timestamp (also known as Unix Epoch Time or POSIX time) is a system for tracking time as a runn...

Unsupervised Learning is a machine learning paradigm where algorithms discover hidden patterns, stru...

URI is a string of characters that identifies a resource on the internet or within a system, providi...

URL is a specific type of URI that provides the complete address to locate a resource on the interne...

URL Encoding (Percent-Encoding) is a mechanism for encoding information in a Uniform Resource Identi...

UTF-8 (8-bit Unicode Transformation Format) is a variable-width character encoding that can represen...

UUID (Universally Unique Identifier) is a 128-bit identifier that is guaranteed to be unique across ...

VAE (Variational Autoencoder) is a generative deep learning model that combines neural network autoe...

A vector database is a specialized database designed to store, index, and query high-dimensional vec...

vLLM is an open-source LLM serving engine designed for high-throughput inference with efficient KV c...

VLM (Vision-Language Model) is a multimodal AI model that can jointly process and reason over both v...

WebLLM is an open-source project developed by the MLC-AI team, aimed at bringing Large Language Mode...

WebP is a modern image format developed by Google that provides superior lossless and lossy compress...

Whitespace refers to characters that represent horizontal or vertical space in text but are typicall...

Workflow Orchestration is the structured coordination of tasks, dependencies, data mappings, executi...

XML (Extensible Markup Language) is a markup language and file format for storing, transmitting, and...

XPath is a query language for selecting nodes and computing values from XML documents using path exp...

YAML (YAML Ain't Markup Language) is a human-readable data serialization language commonly used for ...

Zero-shot learning is a machine learning paradigm where models perform tasks without any task-specif...