AI Engineering Lab: Glossary
Every technical term in the program, in plain language. Terms are grouped by theme and ordered for reading; within each group, earlier terms build on later ones' prerequisites. Each entry ends with where the program teaches it in depth.
How to use this file: don't read it cover to cover. When a week uses a word you don't know, jump to it (your editor's search is your friend), read the one-paragraph definition, and get back to work.
课程里每一个技术术语，都用平实的语言解释。术语按主题分组、并按阅读顺序排列；在同一组内，前面的术语会用到后面术语的前提。每条结尾都标明课程在哪里深入讲它。 怎么用这份文件：不要从头读到尾。当某一周用到一个你不认识的词，直接跳过去（用编辑器的搜索就行），读完那段定义，然后回去继续干活。
Part of AI Engineering Lab · Developed by Zorost Intelligence AI Lab · zorost.com
AI Engineering Lab 的一部分 · 由 Zorost Intelligence AI Lab 开发 · zorost.com
Foundations: software & data
API (Application Programming Interface): A way for one program to ask another program to do something and get an answer back, over a fixed contract. When you "call a model," your code is making an API request. Week 1; KB-01.
API（应用程序接口）：一个程序请另一个程序做事并拿到结果的方式，基于一份固定的约定。当你「调用模型」时，你的代码就是在发一个 API 请求。第 1 周；KB-01。
CLI (Command-Line Interface): A program you drive by typing commands into a terminal instead of clicking buttons. Most AI engineering tools (git, Ollama, Claude Code) are CLIs. Week 1.
CLI（命令行界面）：通过在终端里输入命令来操作、而不是点按钮的程序。大多数 AI 工程工具（git、Ollama、Claude Code）都是 CLI。第 1 周。
Git / GitHub: Git is version control: it records every change to your files so you can undo, compare, and collaborate. GitHub is a website that hosts git repositories so others can see and copy them. This program lives on GitHub. Week 1.
Git / GitHub：Git 是版本控制，它记录你文件的每一次改动，好让你可以撤销、对比和协作。GitHub 是一个托管 git 仓库的网站，让别人能看到并复制它们。本课程就放在 GitHub 上。第 1 周。
IDE (Integrated Development Environment): A code editor with extras: file browser, terminal, debugger, and extensions in one window. The program uses VS Code; Cursor is an AI-native IDE. Week 1; skills/cursor.
IDE（集成开发环境）：带附加功能的代码编辑器——文件浏览器、终端、调试器和扩展都在同一个窗口里。本课程使用 VS Code；Cursor 是一个 AI 原生的 IDE。第 1 周；skills/cursor。
Jupyter / notebook: An interactive document that mixes code cells, their output, and written explanation. The program's hands-on work all happens in notebooks (.ipynb files) so you can run one cell at a time and see what happened. Week 1.
Jupyter / notebook：一种交互式文档，把代码单元格、它们的输出和文字说明混在一起。本课程的动手工作全都在 notebook（.ipynb 文件）里完成，这样你可以一次跑一个单元格、马上看到发生了什么。第 1 周。
Library / package: Pre-written code you import instead of writing from scratch. pandas is a data library; torch is a deep-learning library. Week 1.
库 / 包：别人预先写好的代码，你直接 import 而不是从零写。pandas 是数据处理库，torch 是深度学习库。第 1 周。
Python: The programming language of AI engineering, chosen for its readable syntax and unmatched ecosystem of AI/data libraries. Taught from zero in Week 1.
Python：AI 工程使用的编程语言，因为它语法可读、且拥有无可比拟的 AI 与数据生态。第 1 周从零教起。
Repository ("repo"): A folder of code tracked by git. This program is a repository. Week 1.
仓库（repo）：一个由 git 跟踪的代码文件夹。本课程就是一个仓库。第 1 周。
SQL (Structured Query Language): The language for asking questions of tabular databases: SELECT column FROM table WHERE condition. AI engineering runs on data, and data still speaks SQL. Week 2; KB-02.
SQL（结构化查询语言）：用来向表格式数据库提问的语言：SELECT column FROM table WHERE condition。AI 工程建立在数据之上，而数据至今仍然说 SQL。第 2 周；KB-02。
Terminal / shell: The text window where you type commands directly to the operating system (Terminal on macOS, PowerShell/Windows Terminal on Windows). Week 1.
终端 / shell：直接向操作系统输入命令的文本窗口（macOS 上是 Terminal，Windows 上是 PowerShell 或 Windows Terminal）。第 1 周。
Virtual environment: An isolated folder of Python packages for one project, so one project's library versions can't break another's. You make one in Week 1 and never think about it again. Week 1.
虚拟环境：为某个项目单独隔离出来的一个 Python 包文件夹，这样一个项目的库版本不会搞坏另一个项目。你在第 1 周建好它，之后就不用再管了。第 1 周。
Machine learning & deep learning
Model: A file of learned numbers (parameters) plus the code shape that uses them. Given input, it produces output: an ETA, a category, or the next token. Week 3.
模型：一个装着学到的数字（参数）的文件，加上使用这些参数的代码形态。给定输入，它产出输出：一个预计到达时间、一个类别，或者下一个 token。第 3 周。
Training: The process where a model's parameters are automatically adjusted to reduce its error on example data. Training is expensive; using a trained model (inference) is cheap. Week 3 to 4; KB-02.
训练：自动调整模型参数、以降低它在样例数据上误差的过程。训练很贵；使用一个已训练好的模型（推理）很便宜。第 3 到 4 周；KB-02。
Inference: Using a trained model to make predictions on new input. Most of AI engineering is inference-side: prompting, serving, evaluating. Week 3; KB-02.
推理：用训练好的模型对新输入做预测。AI 工程的大部分工作都在推理这一侧：写提示词、部署服务、评测。第 3 周；KB-02。
Feature: An input variable you hand to a model: distance, carrier, day of week. Feature engineering is crafting inputs the model can actually learn from. Week 3; Week 23.
特征：你交给模型的一个输入变量：距离、承运商、星期几。特征工程就是精心构造出模型真正能学得动的输入。第 3 周；第 23 周。
Label / target: The answer column in training data: the actual ETA, the true on-time flag. Models learn by comparing predictions to labels. Week 3.
标签 / 目标：训练数据里的答案列：实际的到达时间、真实的准时与否。模型通过把预测与标签做对比来学习。第 3 周。
Regression vs. classification: Regression predicts a number (ETA in hours); classification predicts a category (on-time vs. delayed). Week 3 builds one of each.
回归与分类：回归预测一个数值（以小时计的到达时间）；分类预测一个类别（准时还是延误）。第 3 周会各做一个。
Train/test split: Holding out data the model never saw during training, to measure how it performs on new cases, the only performance that matters. Week 3.
训练/测试集划分：留出一部分模型在训练中从未见过的数据，用来衡量它在新样本上的表现——这是唯一真正重要的表现。第 3 周。
Overfitting: When a model memorizes the training data's noise and performs worse on new data. The central failure mode of ML; the test split is how you detect it. Week 3 to 4; KB-02.
过拟合：模型把训练数据里的噪声也背下来了，于是在新数据上表现更差。这是机器学习的核心失效模式；测试集就是发现它的办法。第 3 到 4 周；KB-02。
Neural network: A model made of layers of simple units that multiply inputs by learned weights and pass them through a non-linearity. Depth (many layers) lets it learn complex patterns. Week 4.
神经网络：一种由多层简单单元构成的模型，这些单元把输入乘以学到的权重，再经过一个非线性变换。深度（多层堆叠）让它能学到复杂模式。第 4 周。
PyTorch: The dominant open-source deep-learning framework. Week 4 builds a small neural network in it so the "deep learning" in every LLM stops being magic. Week 4.
PyTorch：主流的开源深度学习框架。第 4 周会用它搭一个小神经网络，让每个 LLM 里的「深度学习」不再显得神秘。第 4 周。
Autograd / gradient descent: The math that makes training work: compute how each parameter contributed to the error (gradients), then nudge every parameter downhill. PyTorch's autograd does the calculus for you. Week 4; KB-02.
自动微分 / 梯度下降：让训练跑起来的数学：先算出每个参数对误差的贡献（梯度），再让每个参数往低处挪一点。PyTorch 的 autograd 替你完成微积分。第 4 周；KB-02。
Embedding: A dense vector of numbers representing meaning, so similar things land near each other in vector space. Used inside LLMs and, separately, to power semantic search. Week 5; KB-03.
嵌入（embedding）：一个表示语义的稠密数值向量，让相似的东西在向量空间里彼此靠近。它既用在 LLM 内部，也单独用来支撑语义检索。第 5 周；KB-03。
LLM core
LLM (Large Language Model): A very large neural network trained to predict the next token in text. That single capability, scaled, produces summarization, extraction, reasoning, and dialogue. Week 5; KB-03.
LLM（大语言模型）：一个规模极大的神经网络，训练目标是预测文本中的下一个 token。就是这一项能力在规模放大后，产生了摘要、抽取、推理和对话。第 5 周；KB-03。
Token: The chunk of text a model actually reads, roughly ¾ of an English word. Costs, context limits, and speed are all measured in tokens. Week 5; KB-03.
Token：模型实际读取的文本片段，大约相当于四分之三个英文单词。成本、上下文上限和速度都以 token 计量。第 5 周；KB-03。
Tokenizer: The component that cuts text into tokens and maps each to an integer ID. Different models tokenize differently, so always measure with the actual model's tokenizer. Week 5.
分词器：把文本切成 token 并把每个 token 映射为整数 ID 的组件。不同模型的分词方式不同，所以始终要用真实模型的分词器来测量。第 5 周。
Transformer: The neural architecture behind every modern LLM: a stack of blocks, each mixing self-attention (tokens gathering information from each other) with a small feed-forward network. Week 5; KB-03.
Transformer：所有现代 LLM 背后的神经网络架构：一叠同样的块，每块把自注意力（token 之间互相收集信息）与一个小型前馈网络混合起来。第 5 周；KB-03。
Attention / self-attention: The mechanism that lets each token weigh which other tokens matter to it ("it" in a sentence attending back to its noun). The Q·K·V projection dance is explained visually in Week 5. Week 5; KB-03.
注意力 / 自注意力：让每个 token 衡量其他哪些 token 对自己重要的机制（句子里的「它」会回头关注它所指的名词）。Q·K·V 投影这套动作在第 5 周有可视化讲解。第 5 周；KB-03。
Context window: The maximum number of tokens a model can consider at once: your instructions, the conversation, retrieved documents, and the model's own output all compete for it. Managing it is context engineering. Week 6; KB-04.
上下文窗口：模型一次能考虑的最大 token 数：你的指令、对话、检索到的文档，以及模型自己的输出，都要争抢这个空间。管理它就是上下文工程。第 6 周；KB-04。
KV cache: A serving optimization that stores each token's key/value vectors so the model doesn't recompute attention for the whole prompt on every generated token. Why the first token is slow and the rest stream fast. Week 5; KB-03.
KV 缓存：一种服务端的优化，把每个 token 的 key/value 向量存起来，这样模型在生成每一个新 token 时不必为整个提示词重算注意力。这就是为什么第一个 token 慢、后面会流式地快。第 5 周；KB-03。
Temperature: A sampling dial: low temperature makes output more predictable (pick the most likely token); high temperature makes it more varied. Extraction tasks run near zero. Week 5 to 6.
温度（temperature）：一个采样旋钮：温度低会让输出更可预测（挑最可能的 token）；温度高会让输出更多样。抽取类任务通常设在接近零。第 5 到 6 周。
Hallucination: When a model states something false with full confidence, because its job is producing plausible text, not verified fact. The engineering answers are retrieval (give it the facts), structured output, and evals. Week 6 to 7; KB-03.
幻觉：模型信誓旦旦地说出错误内容，因为它的本职是生成看起来合理的文本，而不是经过核实的事实。工程上的应对是检索（把事实喂给它）、结构化输出和评测。第 6 到 7 周；KB-03。
Prompt: Everything you put in front of the model: system instructions, examples, the user's request, retrieved evidence. Week 6.
提示词：你放在模型面前的一切：系统指令、示例、用户的请求、检索到的证据。第 6 周。
System prompt: The standing instructions sent with every request: role, policy, output format, guardrails. It is a policy layer, not a security layer. Week 6; KB-04.
系统提示词：随每个请求一起发送的常设指令：角色、策略、输出格式、护栏。它是一层策略，不是一层安全防护。第 6 周；KB-04。
Few-shot prompting: Including 1 to 3 worked input→output examples in the prompt so the model copies the pattern. One good example beats three mediocre ones. Week 6.
少样本提示（few-shot）：在提示词里放进 1 到 3 个完整的输入→输出示例，让模型照着模仿这个模式。一个好例子胜过三个平庸的例子。第 6 周。
Chain-of-thought: Asking the model to reason step by step before answering. Helps multi-step problems; costs tokens on every call. Week 6; KB-04.
思维链：让模型先一步步推理再回答。对多步问题有帮助；每次调用都会多花 token。第 6 周；KB-04。
Structured output: Constraining the model's answer to a schema (usually JSON) so code, not a human, can consume it. Remember: a schema constrains shape, never truth. Week 6; KB-04.
结构化输出：把模型的回答约束到一个 schema（通常是 JSON），好让代码而不是人来消费它。记住：schema 只能约束形状，永远不能约束真假。第 6 周；KB-04。
Prompt injection: An attack where instructions hidden in content the system reads (a web page, a document field) hijack the model. OWASP's #1 LLM risk; defenses live outside the prompt. Week 6; KB-04.
提示词注入：一种攻击：藏在系统所读内容（一个网页、某个文档字段）里的指令劫持了模型。这是 OWASP 列出的头号 LLM 风险；防御手段在提示词之外。第 6 周；KB-04。
Prompt caching: A provider feature that reuses the processed prefix of repeated requests, cutting cost and latency when your system prompt and context are stable. Week 6; KB-04.
提示词缓存：某些供应商提供的能力，把重复请求中已处理过的前缀复用起来，当你的系统提示词和上下文稳定时，能显著降低成本和延迟。第 6 周；KB-04。
RAG (Retrieval-Augmented Generation): Retrieve relevant documents, put them in the context window, and ask the model to answer from them, with citations. The default cure for hallucination and stale knowledge. Week 7; KB-05.
RAG（检索增强生成）：检索出相关文档，放进上下文窗口，再让模型根据它们作答、并给出来源引用。这是应对幻觉和知识过时的默认解法。第 7 周；KB-05。
Vector database: A database that stores embedding vectors and finds the nearest neighbors to a query vector fast. The retrieval half of RAG. Week 7; KB-05.
向量数据库：存储嵌入向量，并能快速找出与查询向量最近邻的数据库。它是 RAG 中「检索」的那一半。第 7 周；KB-05。
Chunking: Splitting documents into pieces small enough to embed and retrieve usefully. Chunk size and overlap are among the highest-leverage RAG decisions. Week 7.
分块（chunking）：把文档切成立即能被有效嵌入和检索的小块。块大小和重叠度是 RAG 里影响最大的决定之一。第 7 周。
Reranking: A second-pass model that re-orders the vector search's top candidates by true relevance. Cheap accuracy for RAG pipelines. Week 7; KB-05.
重排序（reranking）：用一个二次处理的模型，按真实相关性重新排列向量检索的前几名候选。为 RAG 流水线以很低成本换到准确率。第 7 周；KB-05。
Knowledge graph: Data stored as entities and relationships (shipment, carried_by→ carrier), enabling multi-hop questions vector search can't answer. Week 7 builds one.
知识图谱：以实体和关系的形式存储数据（shipment ——carried_by→ carrier），从而支持向量检索答不了的多跳问题。第 7 周会构建一个。
Model engineering
Open-weight model: A model whose parameter files you can download and run yourself (Llama, Qwen, Mistral, DeepSeek…). "Open source" in common speech, though licenses vary. Week 8; KB-06.
开放权重模型：参数文件可以下载、自行运行的模型（Llama、Qwen、Mistral、DeepSeek……）。俗称「开源」，不过各家的许可证并不一样。第 8 周；KB-06。
Foundation model: A large model trained on broad data that you adapt to tasks, by prompting, RAG, or fine-tuning, rather than training from scratch. Week 8; KB-06.
基础模型：在广泛数据上训练出来的大型模型，你通过提示、RAG 或微调把它适配到具体任务，而不是从零训练。第 8 周；KB-06。
Ollama / llama.cpp / MLX: The three local-inference stacks the program uses: Ollama for convenience, llama.cpp for control and GGUF, MLX for Apple Silicon. Week 8; skills/ollama-llamacpp; KB-08.
Ollama / llama.cpp / MLX：本课程使用的三套本地推理栈：Ollama 图方便，llama.cpp 图可控和 GGUF，MLX 面向 Apple Silicon。第 8 周；skills/ollama-llamacpp；KB-08。
GGUF: The file format llama.cpp uses for quantized models, one file you can download and run anywhere. Week 8 to 9.
GGUF：llama.cpp 用于量化模型的文件格式，一个文件下载下来就能在任何地方运行。第 8 到 9 周。
Quantization: Storing model weights at lower precision (16-bit → 8/6/5/4-bit) to shrink memory and speed up inference, at a small quality cost. Q4_K_M is the classic sweet spot. Week 9; KB-06.
量化：用更低的精度（16 位 → 8/6/5/4 位）存储模型权重，以缩小内存占用、加快推理，代价是少量质量损失。Q4_K_M 是经典的最佳平衡点。第 9 周；KB-06。
VRAM: GPU memory. The hard wall for local models: the model's weights, the KV cache, and the runtime must all fit. Week 8 to 9 teach you to size it before downloading.
VRAM：显存。本地模型的硬墙：模型的权重、KV 缓存和运行时必须全部装得下。第 8 到 9 周会教你在下载之前先算清楚需要多大。
vLLM: A high-throughput open-source serving engine for LLMs (paged attention, continuous batching). What you graduate to when one user becomes fifty. Week 9; KB-08.
vLLM：面向 LLM 的高吞吐开源服务引擎（分页注意力、连续批处理）。当你的用户从一个变成五十个时，就该升级到它。第 9 周；KB-08。
Fine-tuning: Continuing a model's training on your data to change its behavior: tone, format, domain vocabulary. Week 10; KB-06.
微调：在你自己数据上继续训练模型，以改变它的行为：语气、格式、领域词汇。第 10 周；KB-06。
SFT (Supervised Fine-Tuning): Fine-tuning on input→ideal-output examples. The first rung of behavior change. Week 10.
SFT（监督微调）：用「输入→理想输出」的样例做微调。这是改变行为的第一级台阶。第 10 周。
LoRA (Low-Rank Adaptation): Fine-tuning a small set of added matrices instead of all parameters: cheap, fast, and the resulting "adapter" file is tiny. Week 10; KB-06.
LoRA（低秩适配）：只微调一小批额外加进去的矩阵，而不是全部参数：便宜、快，产出的「适配器」文件很小。第 10 周；KB-06。
DPO (Direct Preference Optimization): Training on pairs of "better vs. worse" answers to align behavior with preferences, without a separate reward model. Week 10.
DPO（直接偏好优化）：用「较好 vs. 较差」成对的回答来训练，让行为对齐偏好，而不需要单独训练一个奖励模型。第 10 周。
Distillation: Training a small model to imitate a big model's outputs, trading capability for speed and cost. KB-06.
蒸馏：训练一个小模型去模仿大模型的输出，用能力换取速度和成本。KB-06。
Eval (evaluation): A repeatable measurement of a model or system's quality on a fixed set of cases, producing a score. The discipline that separates AI engineers from demo builders. Week 11; KB-07.
Eval（评测）：在固定的一组用例上，对模型或系统的质量做可重复的测量，产出一个分数。正是这门功夫把 AI 工程师和「只能做演示的人」区分开。第 11 周；KB-07。
Golden set: The curated set of inputs with known-good answers that an eval scores against. Fix the golden set before touching the prompt. Week 6, 11.
黄金集：一份经过挑选的输入集合，配有已知的正确回答，评测就以它为基准打分。在动提示词之前，先把黄金集定下来。第 6、11 周。
LLM-as-judge: Using a strong model to grade another system's outputs against a rubric, when correctness is fuzzy (summaries, answers). Cheap and scalable, but must be calibrated against human grades. Week 11; KB-07.
LLM 当裁判：在正确性比较模糊的场景（摘要、回答）里，用一个强模型按评分细则给另一个系统的输出打分。便宜且可扩展，但必须用人工评分做校准。第 11 周；KB-07。
Error analysis: Reading real failures by hand, clustering them into categories, and fixing the largest category first. The highest-ROI habit in the program. Week 11; KB-07.
误差分析：人工阅读真实失败案例，把它们归成类别，先修最大的那一类。本课程中回报率最高的习惯。第 11 周；KB-07。
Harnesses, loops & agents
Harness (coding-agent harness): The software wrapper that lets an LLM act on your computer: read files, run commands, edit code, with permissions, rules files, and context management around it. Claude Code, Cursor, OpenCode, and the DeepSeek Harness are the four the program compares. Week 12; KB-09; reference/skills
Harness（编程 agent 的 harness）：让 LLM 能在你电脑上行动的那层软件外壳：读文件、跑命令、改代码，并且围绕它配有权限、规则文件和上下文管理。Claude Code、Cursor、OpenCode 和 DeepSeek Harness 是本课程对比的四种。第 12 周；KB-09；reference/skills。
Rules file: The persistent instruction file a harness reads every session (CLAUDE.md, AGENTS.md, .cursor/rules): project conventions, commands, and boundaries. Week 12; reference/skills
规则文件：harness 每次会话都会读取的常驻指令文件（CLAUDE.md、AGENTS.md、.cursor/rules）：项目约定、命令和边界。第 12 周；reference/skills。
Subagent: A spawned helper agent with its own context window, given a narrow job (research, review) so the main agent's window stays clean. Week 12; KB-09.
子代理：被派生出来的助手 agent，拥有自己的上下文窗口，被赋予一件很窄的工作（调研、评审），好让主 agent 的上下文保持干净。第 12 周；KB-09。
Verifier: Anything that checks the agent's work automatically: tests, linters, type checks, evals. Closing the loop means the agent runs its verifier itself. Week 13.
验证器：任何能自动检查 agent 工作的东西：测试、linter、类型检查、评测。闭环的意思就是让 agent 自己跑它的验证器。第 13 周。
Spec-driven development: Writing the specification first and steering the agent to it, so quality is judged against an explicit contract instead of vibes. Week 13; KB-01.
规格驱动开发：先写规格，再让 agent 照着规格做，这样质量是拿一份明确的契约来衡量，而不是凭感觉。第 13 周；KB-01。
Agent: A model in a loop with tools: observe → think → act → observe…, until done or stopped. Agents handle tasks where the steps aren't known in advance. Week 14; KB-10.
Agent：一个在循环里带工具运行的模型：观察 → 思考 → 行动 → 再观察……直到完成或被停下。Agent 处理那些步骤事先未知的任务。第 14 周；KB-10。
ReAct: The foundational agent pattern: interleave reasoning ("what should I do next?") with actions (tool calls) in one loop. Week 14 builds one from scratch.
ReAct：最基础的 agent 模式：把推理（「下一步该做什么？」）与行动（工具调用）交错在同一个循环里。第 14 周从零构建一个。
Tool (function calling): A function the model can ask the harness to run: track_shipment(id). The model emits structured intent; your code executes it and returns the result. Week 14; KB-10.
工具（函数调用）：模型可以请求 harness 去运行的函数：track_shipment(id)。模型输出结构化的意图，你的代码执行它并返回结果。第 14 周；KB-10。
Guardrails: The hard limits around an agent: max steps, cost caps, refusal rules, human approval for irreversible actions. Week 14; KB-10.
护栏：围绕 agent 的硬性限制：最大步数、成本上限、拒答规则、不可逆操作需要人工批准。第 14 周；KB-10。
Trace / observability: A recorded log of every step an agent took, prompts, tool calls, outputs, costs, so failures are debuggable and evaluable. Week 17; KB-07, KB-10.
Trace / 可观测性：把 agent 走过的每一步都记录下来，提示词、工具调用、输出、成本——这样失败才可调试、可评测。第 17 周；KB-07、KB-10。
LangGraph: A framework that models agents as explicit state graphs: nodes (steps), edges (transitions), checkpoints (resumability). Week 15; KB-10.
LangGraph：一个把 agent 建模为显式状态图的框架：节点（步骤）、边（转移）、检查点（可恢复）。第 15 周；KB-10。
Multi-agent system: Splitting work across specialized agents (router, tracker, refunds) coordinated by an orchestrator. A measured trade, not a default. Week 16; KB-10.
多 agent 系统：把工作拆给多个专职 agent（路由、追踪、退款），由编排者协调。这是一次经过权衡的取舍，不是默认选项。第 16 周；KB-10。
MCP (Model Context Protocol): An open standard for exposing tools and data to AI agents: one server, many compatible clients. Week 16 builds an MCP server for ZoroLogistics. Week 16; KB-11.
MCP（Model Context Protocol）：一个把工具和数据暴露给 AI agent 的开放标准：一个服务器，多种兼容的客户端。第 16 周会为 ZoroLogistics 构建一个 MCP 服务器。第 16 周；KB-11。
A2A (Agent-to-Agent): Google's protocol for agents delegating tasks to other agents across boundaries; complements MCP's agent-to-tools focus. KB-11.
A2A（Agent-to-Agent）：Google 提出的协议，让 agent 跨边界把任务委派给其他 agent；与 MCP 侧重「agent 到工具」形成互补。KB-11。
OpenClaw: An open-source personal-agent framework the program uses in Week 17 to run a 24/7 assistant wired into the Week-16 MCP server. Week 17; agents/openclaw.
OpenClaw：一个开源的个人 agent 框架，第 17 周用它跑一个 7×24 的助手，并接上第 16 周做的 MCP 服务器。第 17 周；agents/openclaw。
Router / OpenRouter: A service that fronts hundreds of models behind one API key, so you can swap models, compare cost/quality, and use :free variants. Week 12; skills/openrouter.
路由器 / OpenRouter：用一个 API key 把几百个模型摆在身后的服务，让你可以随意换模型、比较成本与质量，还能用 :free 变体。第 12 周；skills/openrouter。
Cloud platforms & Databricks
Azure AI Foundry: Microsoft's platform for building, evaluating, and deploying AI apps and agents, with a model catalog and enterprise governance. Week 18; KB-12.
Azure AI Foundry：微软用于构建、评测和部署 AI 应用与 agent 的平台，带有模型目录和企业级治理。第 18 周；KB-12。
Google Vertex AI: Google Cloud's ML/GenAI platform: Gemini models, Agent Builder, evaluation services, and MLOps. Week 19; KB-12.
Google Vertex AI：Google Cloud 的机器学习/生成式 AI 平台：Gemini 模型、Agent Builder、评测服务和 MLOps。第 19 周；KB-12。
AWS Bedrock: Amazon's managed service offering foundation models (Claude, Llama, Nova…) via one API, plus Knowledge Bases, Agents, and Guardrails. Week 20; KB-12.
AWS Bedrock：亚马逊的托管服务，通过一个 API 提供基础模型（Claude、Llama、Nova……），另有 Knowledge Bases、Agents 和 Guardrails。第 20 周；KB-12。
Serverless endpoint: Model hosting where you pay per request and never manage servers; the opposite of provisioning a GPU instance yourself. Week 18 to 20.
无服务器端点：按请求付费、完全不用自己管服务器的模型托管方式；与你自己开通一台 GPU 实例正好相反。第 18 到 20 周。
Lakehouse: A data platform combining a data lake's cheap open storage with a data warehouse's reliability and SQL performance. Databricks' core idea. Week 21; KB-13.
湖仓（Lakehouse）：把数据湖廉价的开放存储与数据仓库的可靠性和 SQL 性能结合在一起的数据平台。Databricks 的核心主张。第 21 周；KB-13。
Delta Lake: The open storage layer (tables on object storage with ACID transactions, time travel, schema enforcement) under the lakehouse. Week 21.
Delta Lake：湖仓之下的开放存储层（对象存储上的表，具备 ACID 事务、时间旅行、schema 强制）。第 21 周。
Unity Catalog: Databricks' governance layer: one place for data, model, and permission management with lineage. Week 21; platforms/databricks/01.
Unity Catalog：Databricks 的治理层：数据、模型和权限管理集中在一处，并带血缘。第 21 周；platforms/databricks/01。
Medallion architecture: Organizing data in bronze (raw) → silver (cleaned) → gold (business-ready) tables. Week 21 to 22.
Medallion 架构：把数据组织成 bronze（原始）→ silver（清洗后）→ gold（可供业务使用）三层表。第 21 到 22 周。
PySpark: The Python API for Apache Spark, the distributed compute engine that processes data across a cluster. Week 22.
PySpark：Apache Spark 的 Python API，这个分布式计算引擎能在集群上处理数据。第 22 周。
Lakeflow (DLT & Jobs): Databricks' declarative pipeline and orchestration tooling: declare the tables; it handles dependencies, retries, and monitoring. Week 22.
Lakeflow（DLT 与 Jobs）：Databricks 的声明式流水线与编排工具：你只声明表，依赖、重试和监控它来管。第 22 周。
MLflow: The open-source experiment-tracking and model-registry system: every training run logged with parameters, metrics, and artifacts. Week 23.
MLflow：开源的实验跟踪与模型注册系统：每次训练都记下参数、指标和产物。第 23 周。
Feature store: A governed home for computed features so training and serving use identical definitions. Week 23.
特征存储：特征经过治理的统一存放处，让训练和线上服务使用完全一致的定义。第 23 周。
Model serving: Putting a model behind an endpoint that applications call. Databricks Model Serving, Foundry endpoints, and Vertex endpoints are the managed versions. Week 23; KB-12.
模型服务：把模型放在一个端点后面，供应用调用。Databricks Model Serving、Foundry 端点和 Vertex 端点是其托管版本。第 23 周；KB-12。
Vector Search / AI Search: Databricks' managed vector database for RAG, governed by Unity Catalog. Week 23.
Vector Search / AI Search：Databricks 面向 RAG 的托管向量数据库，由 Unity Catalog 治理。第 23 周。
Genie: Databricks' natural-language interface over governed data: ask in English, get a governed SQL answer. Week 23.
Genie：Databricks 在受治理数据之上的自然语言界面：用英文提问，得到受治理的 SQL 回答。第 23 周。
Agent Bricks: Databricks' tooling for building and evaluating agents on platform data. Week 23 to 24.
Agent Bricks：Databricks 用于在平台数据上构建和评测 agent 的工具。第 23 到 24 周。
DABs (Databricks Asset Bundles): Infrastructure-as-code for Databricks: the whole project, jobs, pipelines, endpoints, declared in YAML and deployed with one command. Week 24.
DABs（Databricks Asset Bundles）：Databricks 的基础设施即代码：整个项目——作业、流水线、端点——都用 YAML 声明，一条命令完成部署。第 24 周。
CI/CD (Continuous Integration / Continuous Deployment): Automatically testing and deploying code on every change, so production is never a manual leap of faith. Week 24.
CI/CD（持续集成 / 持续交付）：每次改动都自动测试并部署代码，让上线永远不是一次靠人赌运气的跳跃。第 24 周。
FinOps: The discipline of measuring and controlling cloud/AI spend: budgets, alerts, cost-per-query dashboards. Week 24; platforms/databricks/17.
FinOps：衡量并控制云与 AI 支出的功夫：预算、告警、每次查询成本的看板。第 24 周；platforms/databricks/17。
Governance: The controls that make data and AI usable safely at scale: permissions, lineage, audit, quality rules. The thread that runs through Week 21 to 24.
治理：让数据和 AI 能在规模上被安全使用的那些控制手段：权限、血缘、审计、质量规则。这是贯穿第 21 到 24 周的一条主线。
Missing a term? That's a bug, open an issue or add it (see .github/CONTRIBUTING.md).
少了某个术语？那是个 bug——去开一个 issue，或者把它加上（见 .github/CONTRIBUTING.md）。
© 2026 Zorost Intelligence LLC · zorost.com
© 2026 Zorost Intelligence LLC · zorost.com