AI Glossary
Understand core AI concepts, from basics to practice. Plain-language explanations of trending terms like Agent, RAG, MCP, and more.
Agent
The essence of an Agent is iterative execution — perceive → plan → act → observe → replan, repeating until the task is complete or a termination condition is met.
Read MoreChain of Thought (CoT)
CoT (Chain of Thought) is a method that guides models to展开 reasoning step by step, making intermediate judgments explicit to reduce skipping and guesswork.
Read MoreComputer Use
Computer Use is the ability for an AI model to recognize interface elements by taking screenshots and simulate mouse clicks and keyboard input to control a computer.
Read MoreContext Window
Context Window is the maximum number of tokens a model can process in a single request, including input content, conversation history, system prompts — with space typically reserved for output as well.
Read MoreCosine Similarity
Cosine Similarity measures how close two vectors are in "direction," with values closer to 1 indicating greater similarity.
Read MoreEmbedding
The essence of Embedding is converting text into a sequence of numbers (vectors), so that semantically similar content has close distances in the numeric space.
Read MoreEmbedding Model
An Embedding Model is a model specifically designed to convert text into vectors, and it determines "how good the converted vectors are.
Read MoreFine-tuning
Fine-tuning is the technique of continuing to train an already-trained large model on domain-specific or task-specific data, so the model performs better in that scenario.
Read MoreFunction Schema
Function Schema is a structured description of a tool's capabilities, serving as the "parameter contract" when the model calls a tool.
Read MoreHarness Engineering
Harness Engineering is an Agent development methodology with the core idea of "harnessing Agent behavior through carefully designed systems, enabling maximum efficiency within boundaries.
Read MoreMCP (Model Context Protocol)
MCP (Model Context Protocol) is an open protocol that enables AI applications to connect to external capabilities in a unified way.
Read MoreMemory (Agent Memory)
Memory is an Agent's ability to save and reuse information, enabling it to maintain coherence across multiple interactions and complex tasks.
Read MoreOpenClaw
OpenClaw is a local-first personal AI assistant that runs on your own device and communicates with you through multiple channels.
Read MorePrompt
A Prompt is the input content given to the model, and the direct basis for the model to understand the task — it's not just "one sentence of instruction," but a task description that simultaneously includes goals, context, materials, constraints, and output requirements.
Read MorePrompt Engineering
Prompt Engineering is the process of designing, testing, comparing, and iteratively optimizing Prompts around model output effectiveness.
Read MoreRAG
RAG (Retrieval Augmented Generation) is an architecture of "retrieve external materials first, then generate answers based on those materials.
Read MoreReAct
ReAct (Reasoning + Acting) is a working mode that makes models alternate between "think, act, observe" cycles.
Read MoreRetrieval
Retrieval is the process of finding the most relevant information from large amounts of data for the current question.
Read MoreSKILL (Agent Skill)
SKILL is a specific capability unit that an Agent can execute, defining "what it can do" and "how to do it.
Read MoreStructured Output
Structured Output is the ability to have models return results in a predefined structure, such as JSON, object fields, or fixed enumerated values, rather than free-form text.
Read MoreSystem Prompt
System Prompt is the global rule layer set for the model before a conversation starts, defining role, goals, boundaries, and default behavior.
Read MoreTemperature
Temperature controls the randomness of model output.
Read MoreToken
A Token is the smallest unit of measurement when a model processes text, not equal to "character" or "word," but rather text fragments segmented by a tokenizer.
Read MoreTool Calling
Tool Calling is the ability that enables models to convert "what should I do" into "which tool should I call, what parameters should I pass.
Read MoreTop-p
Top-p limits the sampling candidate range, selecting only from the smallest set of words whose cumulative probability reaches the threshold.
Read MoreTransformer Architecture
Transformer is a deep learning architecture proposed by Google in 2017 that uses "Self-Attention Mechanism" to enable models to simultaneously process relationships between any positions in a sequence.
Read MoreVector Database
A Vector Database is a database specifically designed to store high-dimensional vectors and support semantic similarity retrieval.
Read MoreZero-shot / Few-shot
Zero-shot and Few-shot are ways to complete tasks through Prompt without training the model — essentially "learning through context" (In-context Learning).
Read More