Skip to main content
Home/Glossary

AI & SaaS Glossary

90 key AI, machine learning, and SaaS terms explained in plain language.

C

CAC

CAC(客户获取成本)是指企业为获取一位新客户所花费的平均总成本。

CDN

A CDN (Content Delivery Network) is a network of edge nodes spread across the globe that serves static content — images, video, scripts — and even dynamic requests from locations near the user, reducing latency, offloading origin servers, and improving availability.

Chain of Thought

思维链是一种让AI模型在给出最终答案前,先展示其逐步推理过程的技术,就像人类在解决问题时会先列出思考步骤一样。

Churn Rate

流失率是指在一段时间内,停止使用某个产品或服务的客户所占的百分比。

CI/CD

CI/CD 是一种软件开发方法,它通过自动化代码的构建、测试和部署流程,帮助团队更频繁、更可靠地发布软件更新。

Computer Vision

计算机视觉是让计算机像人一样‘看’和理解图像和视频内容的技术。

Context Window

上下文窗口是指AI模型(如聊天机器人)在生成回复时,能够同时考虑和记住的对话或文本的最大长度。

Conversion Rate

转化率是指完成特定目标动作(例如购买、注册或点击)的访问者占总访问者的百分比。它衡量的是营销或用户体验将访客转化为客户或用户的效率。

Caching

Caching stores the result of a computation or request close to where it will be used so identical future requests can reuse it. On the web, caches exist at the browser, CDN edge, reverse proxy, application, and database layers — making it the highest-leverage performance optimization available.

R

RAG

RAG(检索增强生成)是一种让AI模型在回答问题时,能够从外部知识库中查找并引用最新、准确信息的技术。

Reinforcement Learning

强化学习是一种让AI通过试错来学习的方法,就像训练宠物一样,做对了给予奖励,做错了给予惩罚,从而学会完成特定任务。

REST API

REST API 是一种让不同软件系统通过互联网进行通信的标准方式,它使用常见的网络请求(如获取、创建、更新或删除数据)来交换信息。

Retention Rate

Retention rate is the percentage of users or customers who continue to use a product or service over a specific period of time, indicating how well a business keeps its audience engaged.

RLHF

RLHF(人类反馈强化学习)是一种训练AI模型的方法,它通过收集人类对模型输出的偏好反馈,来引导模型学习生成更符合人类价值观和期望的回复。

Reverse Proxy

A reverse proxy is an intermediary deployed on the server side: client requests hit the proxy first, which forwards them to backend services and returns the responses. Clients only ever see the proxy's address; the backend architecture stays completely hidden. Nginx, Caddy, HAProxy, and most cloud load balancers are reverse proxies.

Rate Limiting

Rate limiting protects services by capping the number of requests allowed in a time window: excess requests are rejected or queued. It prevents traffic spikes from overwhelming a system, and also powers anti-abuse and API metering. In high-availability design it is the core trade of sacrificing some requests to keep the whole service up.