Build maintainable AI applications from requirements and structured outputs through retrieval, evaluation, safety, and release.
AI application development
Build maintainable AI applications from requirements and structured outputs through retrieval, evaluation, safety, and release.
7 chapters3 casesPublished
LEARNING PATH
Tutorial chapters
Actionable chapters arranged in a useful learning sequence.
chapters · PublishedAI 应用需求拆解:先定义边界与失败
把模糊的 AI 需求拆成输入、输出、证据、人工介入和不可做事项。
Read chapter chapters · Published结构化输出与工具调用:把模型接入业务代码使用 schema 校验模型输出,隔离工具权限和业务副作用。
Read chapter chapters · Published应用内 RAG:检索、引用与权限过滤把检索结果变成可追溯上下文,避免把无权限文档送给模型。
Read chapter chapters · PublishedAI 应用测试:事实、行为与安全三层回归建立固定样本和程序断言,持续发现答案、工具轨迹和权限回归。
Read chapter chapters · Published安全与隐私:输入过滤、最小权限与审计在模型调用前后建立敏感信息、提示注入和越权防线。
Read chapter chapters · Published生产运维:延迟、成本、缓存与人工接管把模型、检索和工具调用纳入可观测、可限流、可回滚的运行流程。
Read chapter chapters · PublishedAI 应用流式输出与部分失败:让用户看见真实状态设计可取消的流式响应、超时和部分失败状态,避免把等待伪装成成功。
Read chapterPRACTICE
Practical cases
Reusable notes built around tools, workflows, and real tasks.
