Glossary
RAG, evals, agents, embeddings, guardrails — defined in plain English, with the part most glossaries skip: what it means when you actually run it in production.
Guardrails are validation and policy checks placed around an LLM — on inputs, outputs, and tool calls — that constrain behavior, block unsafe actions, and enforce format or content rules the model cannot be trusted to follow on its own.
Embeddings are numeric vectors that represent the meaning of text (or images, code, audio) so that semantically similar items land close together in vector space, enabling similarity search.
LLM evals are automated tests that score a model or pipeline's outputs against expected behavior on a fixed dataset, so you can measure whether a change to prompts, models, or retrieval makes the system better or worse.
A context window is the maximum amount of text, measured in tokens, that an LLM can process in a single request — covering the prompt, retrieved context, conversation history, and the model's own output.
A multi-agent system uses several LLM agents, each with a focused role or toolset, that coordinate — sequentially or in parallel — to accomplish a task that a single agent would handle less reliably.
A vector database is a store optimized for indexing high-dimensional embedding vectors and running fast approximate nearest-neighbor search to retrieve the most semantically similar items to a query.
An agentic workflow is a predetermined sequence of LLM calls and tool uses, orchestrated by your code, where the path is fixed in advance — unlike an agent, which decides its own next step at runtime.
An AI agent is an LLM operating in a loop with access to tools, where the model decides which action to take next based on prior results, continuing until it reaches a goal or a stopping condition.
A hallucination is when an LLM produces confident, fluent output that is factually wrong or unsupported by its input — a consequence of models generating plausible text rather than retrieving verified facts.
Chunking is the process of splitting source documents into smaller passages before embedding them, so retrieval can return focused, self-contained units of text rather than whole documents.
Context engineering is the practice of deliberately curating what an LLM sees in its context window — instructions, retrieved data, tools, and history — to maximize relevance and reliability while staying within the token budget.
Fine-tuning is the process of further-training a pretrained model on a curated dataset of examples to specialize its behavior, style, or format, baking the new pattern into the model's weights.
LLM-as-a-judge is an evaluation technique that uses a language model to score or compare the outputs of another model against a rubric, enabling automated grading of open-ended responses that exact-match tests cannot handle.
Reranking is a second retrieval stage that takes an initial set of candidate documents and reorders them by relevance to the query using a more precise model, so the strongest passages reach the LLM's context first.
Retrieval-augmented generation (RAG) is a pattern where a system retrieves relevant documents from your own corpus at query time and passes them to an LLM as context, so the model answers from your data rather than its training memory.
The Model Context Protocol (MCP) is an open standard that defines how LLM applications connect to external tools and data sources through a uniform interface, so a tool built once can be reused across any MCP-compatible client.
Tool calling (or function calling) is the mechanism by which an LLM emits a structured request to invoke a function you defined, receives the result, and continues reasoning — letting the model take actions and fetch live data.
We use cookies for analytics and ads measurement (incl. the Meta Pixel). Privacy policy.