{ "title": "Agent 开发：从原理到生产级实践", "version": "1.0.0", "build_time": "2026-04-01T15:05:49.115Z", "total_chapters": 52, "total_word_count": 224429, "volumes": [ { "id": "vol1", "title": "认知篇", "chapters": [ { "id": "ch01", "title": "第1章：什么是AI Agent", "file": "ch01.json", "markdown": "ch01.md", "word_count": 7226, "key_concepts": [ "Agent的定义", "一个直观的定义", "学术视角的定义", "实践中的定义", "Agent的核心特征" ], "dependencies": [], "difficulty": "beginner" }, { "id": "ch02", "title": "第2章：Agent技术演进史", "file": "ch02.json", "markdown": "ch02.md", "word_count": 5290, "key_concepts": [ "引言：从\"计算\"到\"行动\"的范式转变", "ELIZA：对话式Agent的起点", "SHRDLU：自然语言理解的里程碑", "Terry Winograd的贡献与局限", "专家系统：知识即力量" ], "dependencies": [ "ch01" ], "difficulty": "beginner" }, { "id": "ch03", "title": "第3章：Agent编程范式", "file": "ch03.json", "markdown": "ch03.md", "word_count": 5098, "key_concepts": [ "从函数式到 Agent式：编程范式的演变", "范式演进的驱动力", "Agent范式的独特特征", "为什么现在需要Agent范式", "Agent编程的核心要素" ], "dependencies": [ "ch01" ], "difficulty": "beginner" } ] }, { "id": "vol2", "title": "基础篇", "chapters": [ { "id": "ch04", "title": "第4章：Agent 核心概念", "file": "ch04.json", "markdown": "ch04.md", "word_count": 3660, "key_concepts": [ "什么是 Agent", "从 Chatbot 到 Agent", "Agent 的直觉类比", "Agent 的能力层次", "Agent 架构模型" ], "dependencies": [ "ch01", "ch03" ], "difficulty": "beginner" }, { "id": "ch05", "title": "第5章：LLM 基础与 Prompt Engineering", "file": "ch05.json", "markdown": "ch05.md", "word_count": 3163, "key_concepts": [ "LLM 工作原理简述", "从输入到输出：LLM 的推理流程", "关键概念", "LLM 的能力边界", "Prompt 设计原则" ], "dependencies": [ "ch04" ], "difficulty": "beginner" }, { "id": "ch06", "title": "第6章：工具调用与 Function Calling", "file": "ch06.json", "markdown": "ch06.md", "word_count": 1988, "key_concepts": [ "Function Calling 机制", "什么是 Function Calling", "Function Calling vs 传统方式", "主流模型的 Function Calling 支持", "工具定义与描述" ], "dependencies": [ "ch04", "ch05" ], "difficulty": "beginner" }, { "id": "ch07", "title": "第7章：记忆与上下文管理", "file": "ch07.json", "markdown": "ch07.md", "word_count": 1884, "key_concepts": [ "短期记忆 vs 长期记忆", "记忆的层次模型", "各层记忆的实现", "短期记忆管理", "长期记忆管理" ], "dependencies": [ "ch04" ], "difficulty": "beginner" } ] }, { "id": "vol3", "title": "进阶篇", "chapters": [ { "id": "ch08", "title": "第8章：多Agent协作", "file": "ch08.json", "markdown": "ch08.md", "word_count": 2810, "key_concepts": [ "为什么需要多Agent？", "单Agent vs 多Agent", "何时选择多Agent", "多Agent架构模式", "层级式架构（Hierarchical）" ], "dependencies": [ "ch04", "ch06" ], "difficulty": "intermediate" }, { "id": "ch09", "title": "第9章：Agent推理与规划", "file": "ch09.json", "markdown": "ch09.md", "word_count": 2672, "key_concepts": [ "推理：Agent 的\"思考\"核心", "推理的层次模型", "ReAct：推理与行动交织", "ReAct 核心循环", "ReAct 完整实现" ], "dependencies": [ "ch04", "ch05" ], "difficulty": "intermediate" }, { "id": "ch10", "title": "第10章：Agent评估与优化", "file": "ch10.json", "markdown": "ch10.md", "word_count": 2077, "key_concepts": [ "为什么Agent评估如此困难", "评估的多维度挑战", "评估指标体系", "质量指标", "效率指标" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" }, { "id": "ch11", "title": "第11章：安全与对齐", "file": "ch11.json", "markdown": "ch11.md", "word_count": 2301, "key_concepts": [ "为什么安全是 Agent 的第一要务", "Agent 安全威胁全景", "安全等级定义", "Prompt 注入攻击与防御", "攻击类型" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" } ] }, { "id": "vol4", "title": "高级篇", "chapters": [ { "id": "ch12", "title": "第12章：RAG增强Agent", "file": "ch12.json", "markdown": "ch12.md", "word_count": 4980, "key_concepts": [ "概述", "LLM 的知识困境", "什么是 RAG", "为什么 Agent 需要 RAG", "RAG 的发展历程" ], "dependencies": [ "ch04", "ch07" ], "difficulty": "intermediate" }, { "id": "ch13", "title": "第13章：Agent与外部系统集成", "file": "ch13.json", "markdown": "ch13.md", "word_count": 2135, "key_concepts": [ "集成的必要性与挑战", "为什么Agent需要集成外部系统", "核心挑战", "集成架构总览", "REST API集成" ], "dependencies": [ "ch06" ], "difficulty": "intermediate" }, { "id": "ch14", "title": "第14章：多模态Agent", "file": "ch14.json", "markdown": "ch14.md", "word_count": 1810, "key_concepts": [ "多模态AI概述", "什么是多模态", "多模态Agent的核心价值", "视觉Agent", "图像理解" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" }, { "id": "ch15", "title": "第15章：Agent的可观测性", "file": "ch15.json", "markdown": "ch15.md", "word_count": 1854, "key_concepts": [ "可观测性的三支柱", "日志、指标、追踪", "Agent可观测性框架", "Agent系统可观测性挑战", "非确定性" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" } ] }, { "id": "vol5", "title": "专项篇", "chapters": [ { "id": "ch16", "title": "第16章：代码 Agent", "file": "ch16.json", "markdown": "ch16.md", "word_count": 3409, "key_concepts": [ "概述：当 AI 学会写代码", "代码 Agent 的能力光谱", "代码 Agent 与传统 IDE 辅助的本质区别", "代码 Agent 的核心架构", "架构总览" ], "dependencies": [ "ch06", "ch09" ], "difficulty": "intermediate" }, { "id": "ch17", "title": "第17章：数据分析 Agent", "file": "ch17.json", "markdown": "ch17.md", "word_count": 2466, "key_concepts": [ "概述：从\"写 SQL\"到\"说人话\"", "数据分析 Agent 的能力边界", "数据分析 Agent 的技术栈", "自然语言 SQL 查询引擎", "Schema 感知的 Prompt 工程" ], "dependencies": [ "ch06", "ch09" ], "difficulty": "intermediate" }, { "id": "ch18", "title": "第18章：内容创作 Agent", "file": "ch18.json", "markdown": "ch18.md", "word_count": 2326, "key_concepts": [ "概述：当 AI 学会\"写作\"", "内容创作 Agent 的能力光谱", "内容创作 Agent 的核心挑战", "文章写作与编辑引擎", "文章结构化管理器" ], "dependencies": [ "ch05", "ch06" ], "difficulty": "intermediate" }, { "id": "ch19", "title": "第19章：自动化工作流 Agent", "file": "ch19.json", "markdown": "ch19.md", "word_count": 1426, "key_concepts": [ "概述：从脚本到智能编排", "工作流 vs. 脚本 vs. Agent", "工作流 Agent 的核心能力", "工作流引擎设计", "核心数据结构" ], "dependencies": [ "ch06", "ch09" ], "difficulty": "intermediate" } ] }, { "id": "vol6", "title": "附录", "chapters": [ { "id": "appendixA", "title": "附录A：Agent框架对比", "file": "appendixA.json", "markdown": "appendixA.md", "word_count": 7441, "key_concepts": [ "A.1 概述", "A.1.1 框架分类", "A.2 框架详细分析", "A.2.1 LangChain", "A.2.2 LangGraph" ], "dependencies": [], "difficulty": "beginner" }, { "id": "appendixB", "title": "附录B：常用工具与API速查", "file": "appendixB.json", "markdown": "appendixB.md", "word_count": 4633, "key_concepts": [ "B.1 大语言模型API", "B.1.1 OpenAI API", "B.1.2 Anthropic API (Claude)", "B.1.3 本地模型推理", "B.2 向量数据库" ], "dependencies": [], "difficulty": "beginner" }, { "id": "appendixC", "title": "附录C：Agent部署指南", "file": "appendixC.json", "markdown": "appendixC.md", "word_count": 2844, "key_concepts": [ "C.1 部署概览", "C.1.1 部署模式对比", "C.1.2 部署前检查清单", "C.2 Docker容器化部署", "C.2.1 Dockerfile最佳实践" ], "dependencies": [], "difficulty": "beginner" }, { "id": "appendixD", "title": "附录D：术语表", "file": "appendixD.json", "markdown": "appendixD.md", "word_count": 17072, "key_concepts": [ "使用说明", "A", "A/B Testing（A/B测试）", "Absorbing State（吸收状态）", "Action Space（动作空间）" ], "dependencies": [], "difficulty": "beginner" } ] }, { "id": "vol7", "title": "Agent编程技法", "chapters": [ { "id": "ch20", "title": "第20章：Prompt高级技巧", "file": "ch20.json", "markdown": "ch20.md", "word_count": 2419, "key_concepts": [ "引言：从\"提示词\"到\"编程接口\"", "本章学习目标", "Prompt版本管理", "为什么需要版本管理", "基于Git的Prompt版本管理" ], "dependencies": [ "ch05" ], "difficulty": "intermediate" }, { "id": "ch21", "title": "第21章：状态机与流程编排", "file": "ch21.json", "markdown": "ch21.md", "word_count": 2338, "key_concepts": [ "引言", "本章学习目标", "有限状态机（FSM）设计", "为什么Agent需要状态机", "状态机基础实现" ], "dependencies": [ "ch09" ], "difficulty": "intermediate" }, { "id": "ch22", "title": "第22章：错误处理与重试策略", "file": "ch22.json", "markdown": "ch22.md", "word_count": 1462, "key_concepts": [ "引言", "本章学习目标", "错误分类体系", "Agent错误分类", "错误严重等级" ], "dependencies": [ "ch06" ], "difficulty": "intermediate" }, { "id": "ch23", "title": "第23章：Agent测试方法", "file": "ch23.json", "markdown": "ch23.md", "word_count": 1576, "key_concepts": [ "引言", "本章学习目标", "Agent测试金字塔", "测试分层", "各层测试对比" ], "dependencies": [ "ch10" ], "difficulty": "intermediate" }, { "id": "ch24", "title": "第24章：性能调优", "file": "ch24.json", "markdown": "ch24.md", "word_count": 1699, "key_concepts": [ "Agent系统性能概述", "延迟来源分析", "性能优化决策树", "LLM推理优化", "模型选择策略" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" }, { "id": "ch25", "title": "第25章：调试与诊断", "file": "ch25.json", "markdown": "ch25.md", "word_count": 2032, "key_concepts": [ "Agent调试的挑战", "非确定性", "调试难度等级", "本地调试环境搭建", "Mock LLM" ], "dependencies": [ "ch15" ], "difficulty": "intermediate" } ] }, { "id": "vol8", "title": "实战案例集", "chapters": [ { "id": "ch26", "title": "第26章：智能客服系统", "file": "ch26.json", "markdown": "ch26.md", "word_count": 2309, "key_concepts": [ "需求分析与功能规划", "业务背景", "功能清单", "非功能需求", "架构设计" ], "dependencies": [ "ch06", "ch07", "ch08" ], "difficulty": "advanced" }, { "id": "ch27", "title": "第27章：代码审查助手", "file": "ch27.json", "markdown": "ch27.md", "word_count": 1769, "key_concepts": [ "需求分析与功能规划", "业务背景", "功能清单", "技术选型", "架构设计" ], "dependencies": [ "ch16" ], "difficulty": "advanced" }, { "id": "ch28", "title": "第28章：数据分析平台", "file": "ch28.json", "markdown": "ch28.md", "word_count": 26306, "key_concepts": [ "需求分析与功能规划", "业务背景", "功能清单", "非功能需求", "架构设计" ], "dependencies": [ "ch17" ], "difficulty": "advanced" }, { "id": "ch29", "title": "第29章：内容生成流水线", "file": "ch29.json", "markdown": "ch29.md", "word_count": 2279, "key_concepts": [ "需求分析与功能规划", "业务背景", "功能清单", "非功能需求", "架构设计" ], "dependencies": [ "ch18" ], "difficulty": "advanced" }, { "id": "ch30", "title": "第30章：多Agent协作项目管理系统", "file": "ch30.json", "markdown": "ch30.md", "word_count": 2412, "key_concepts": [ "需求分析与功能规划", "业务背景", "功能清单", "非功能需求", "架构设计" ], "dependencies": [ "ch08", "ch32" ], "difficulty": "advanced" } ] }, { "id": "vol9", "title": "Agent设计模式", "chapters": [ { "id": "ch31", "title": "第31章 单Agent模式", "file": "ch31.json", "markdown": "ch31.md", "word_count": 7964, "key_concepts": [ "概述", "ReAct 模式（Reasoning + Acting）", "意图", "动机", "结构" ], "dependencies": [ "ch04" ], "difficulty": "advanced" }, { "id": "ch32", "title": "第32章 多Agent编排模式", "file": "ch32.json", "markdown": "ch32.md", "word_count": 5062, "key_concepts": [ "概述", "Orchestrator 模式（中心编排）", "意图", "动机", "结构" ], "dependencies": [ "ch08" ], "difficulty": "advanced" }, { "id": "ch33", "title": "第33章 人机协作模式", "file": "ch33.json", "markdown": "ch33.md", "word_count": 4272, "key_concepts": [ "概述", "Human-in-the-Loop 模式", "意图", "动机", "结构" ], "dependencies": [ "ch08" ], "difficulty": "advanced" }, { "id": "ch34", "title": "第34章：工具使用模式", "file": "ch34.json", "markdown": "ch34.md", "word_count": 1781, "key_concepts": [ "工具抽象的本质", "从函数到工具", "Function Calling 标准接口", "基础工具模式", "单一工具调用" ], "dependencies": [ "ch06" ], "difficulty": "advanced" }, { "id": "ch35", "title": "第35章：企业级设计模式", "file": "ch35.json", "markdown": "ch35.md", "word_count": 1286, "key_concepts": [ "企业级Agent的特殊需求", "Agent Gateway模式", "API网关设计", "路由与负载均衡", "Agent Registry与发现" ], "dependencies": [ "ch31", "ch32", "ch33", "ch34" ], "difficulty": "advanced" } ] }, { "id": "vol10", "title": "生产级Agent平台", "chapters": [ { "id": "ch36", "title": "第36章 生产环境架构设计", "file": "ch36.json", "markdown": "ch36.md", "word_count": 3648, "key_concepts": [ "概述：从原型到生产的跨越", "全局架构总览", "微服务架构下的 Agent 服务", "服务拆分原则", "服务间通信" ], "dependencies": [ "ch35" ], "difficulty": "advanced" }, { "id": "ch37", "title": "第37章 可扩展性与高可用", "file": "ch37.json", "markdown": "ch37.md", "word_count": 2636, "key_concepts": [ "概述：为什么高可用对 Agent 系统尤其重要", "可用性目标定义", "水平扩展策略", "无状态设计原则", "有状态服务的特殊处理" ], "dependencies": [ "ch36" ], "difficulty": "advanced" }, { "id": "ch38", "title": "第38章 监控与告警体系", "file": "ch38.json", "markdown": "ch38.md", "word_count": 2174, "key_concepts": [ "概述：可观测性三支柱", "指标体系设计", "指标分类", "核心 Agent 指标定义", "指标采集中间件" ], "dependencies": [ "ch15", "ch36" ], "difficulty": "advanced" }, { "id": "ch39", "title": "第39章 安全与权限管理", "file": "ch39.json", "markdown": "ch39.md", "word_count": 1921, "key_concepts": [ "概述：Agent 系统的独特安全挑战", "身份认证", "认证方案选型", "JWT 实现", "API Key 管理" ], "dependencies": [ "ch11", "ch36" ], "difficulty": "advanced" }, { "id": "ch40", "title": "第40章：CI/CD与版本管理", "file": "ch40.json", "markdown": "ch40.md", "word_count": 1182, "key_concepts": [ "Agent系统的CI/CD特殊性", "Prompt版本管理", "Prompt版本化存储", "Prompt变更检测", "自动化评估Pipeline" ], "dependencies": [ "ch36" ], "difficulty": "advanced" } ] }, { "id": "vol11", "title": "生态与跨平台", "chapters": [ { "id": "ch41", "title": "第41章：Agent 生态全景", "file": "ch41.json", "markdown": "ch41.md", "word_count": 8313, "key_concepts": [ "生态全景：从模型到应用的技术栈", "技术栈分层模型", "层间关系与依赖分析", "模型层生态：LLM 提供商全景", "国际头部厂商" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" }, { "id": "ch42", "title": "第42章：跨平台部署", "file": "ch42.json", "markdown": "ch42.md", "word_count": 6055, "key_concepts": [ "概述：为什么需要跨平台", "Agent 产品跨平台的必要性", "Agent 跨平台的特殊挑战", "跨平台架构总览", "Web 端部署" ], "dependencies": [ "ch36" ], "difficulty": "intermediate" }, { "id": "ch43", "title": "第43章：开源社区与贡献", "file": "ch43.json", "markdown": "ch43.md", "word_count": 6246, "key_concepts": [ "概述：开源对 Agent 生态的重要性", "开源：Agent 生态的基石", "Agent 开源项目分类全景", "参与开源的价值", "主要开源项目深度分析" ], "dependencies": [ "ch04" ], "difficulty": "intermediate" }, { "id": "ch44", "title": "第44章：商业化实践", "file": "ch44.json", "markdown": "ch44.md", "word_count": 6400, "key_concepts": [ "概述：Agent 商业化的机遇与挑战", "Agent 产品的商业机遇", "Agent 商业化的核心挑战", "Agent 商业化的三个阶段", "产品形态分析" ], "dependencies": [ "ch36" ], "difficulty": "intermediate" } ] }, { "id": "vol12", "title": "Agent编程的未来", "chapters": [ { "id": "ch45", "title": "第45章：Agent原生应用", "file": "ch45.json", "markdown": "ch45.md", "word_count": 7080, "key_concepts": [ "引言：从\"AI增强\"到\"Agent原生\"的范式跃迁", "什么是Agent原生应用", "定义", "核心特征", "发展阶段" ], "dependencies": [ "ch04" ], "difficulty": "beginner" }, { "id": "ch46", "title": "第46章：AI Agent与AGI", "file": "ch46.json", "markdown": "ch46.md", "word_count": 8294, "key_concepts": [ "引言：Agent是通向AGI的桥梁吗？", "当前Agent能力边界", "能力评估框架", "当前Agent的能力图谱", "能力边界的深层原因" ], "dependencies": [ "ch04" ], "difficulty": "beginner" }, { "id": "ch47", "title": "第47章：伦理与社会影响", "file": "ch47.json", "markdown": "ch47.md", "word_count": 5487, "key_concepts": [ "引言：技术无善恶，但使用者有责任", "就业影响与技能转型", "被影响的工作类别", "历史的启示", "技能转型路径" ], "dependencies": [ "ch11" ], "difficulty": "beginner" }, { "id": "ch48", "title": "第48章：展望与行动指南", "file": "ch48.json", "markdown": "ch48.md", "word_count": 7462, "key_concepts": [ "引言：未来已来，只是分布不均", "技术趋势总结", "十大趋势", "三个层次的影响", "开发者学习路径" ], "dependencies": [ "ch04" ], "difficulty": "beginner" } ] } ], "skills": [ { "id": "first-agent", "title": "构建你的第一个Agent", "description": "从零搭建一个能感知、推理、执行工具调用的最小Agent系统", "difficulty": "beginner", "chapters": [ "ch01", "ch04", "ch06" ] }, { "id": "prompt-engineering", "title": "Prompt Engineering 实战", "description": "系统掌握面向Agent的Prompt设计技巧", "difficulty": "beginner", "chapters": [ "ch05", "ch20" ] }, { "id": "rag-agent", "title": "构建RAG增强Agent", "description": "实现检索增强生成，让Agent拥有外部知识库", "difficulty": "intermediate", "chapters": [ "ch04", "ch07", "ch12" ] }, { "id": "multi-agent", "title": "多Agent协作系统", "description": "设计和实现多Agent编排方案", "difficulty": "intermediate", "chapters": [ "ch08", "ch32" ] }, { "id": "agent-testing", "title": "Agent测试与评估", "description": "建立Agent质量保障体系", "difficulty": "intermediate", "chapters": [ "ch10", "ch23" ] }, { "id": "production-deploy", "title": "生产级Agent部署", "description": "从开发到生产的完整链路", "difficulty": "advanced", "chapters": [ "ch36", "ch37", "ch38", "ch40" ] }, { "id": "enterprise-patterns", "title": "企业级设计模式", "description": "大规模Agent系统的架构模式和最佳实践", "difficulty": "advanced", "chapters": [ "ch35", "ch39" ] } ], "dependency_graph": { "ch01": [], "ch02": [ "ch01" ], "ch03": [ "ch01" ], "ch04": [ "ch01", "ch03" ], "ch05": [ "ch04" ], "ch06": [ "ch04", "ch05" ], "ch07": [ "ch04" ], "ch08": [ "ch04", "ch06" ], "ch09": [ "ch04", "ch05" ], "ch10": [ "ch04" ], "ch11": [ "ch04" ], "ch12": [ "ch04", "ch07" ], "ch13": [ "ch06" ], "ch14": [ "ch04" ], "ch15": [ "ch04" ], "ch16": [ "ch06", "ch09" ], "ch17": [ "ch06", "ch09" ], "ch18": [ "ch05", "ch06" ], "ch19": [ "ch06", "ch09" ], "ch20": [ "ch05" ], "ch21": [ "ch09" ], "ch22": [ "ch06" ], "ch23": [ "ch10" ], "ch24": [ "ch04" ], "ch25": [ "ch15" ], "ch26": [ "ch06", "ch07", "ch08" ], "ch27": [ "ch16" ], "ch28": [ "ch17" ], "ch29": [ "ch18" ], "ch30": [ "ch08", "ch32" ], "ch31": [ "ch04" ], "ch32": [ "ch08" ], "ch33": [ "ch08" ], "ch34": [ "ch06" ], "ch35": [ "ch31", "ch32", "ch33", "ch34" ], "ch36": [ "ch35" ], "ch37": [ "ch36" ], "ch38": [ "ch15", "ch36" ], "ch39": [ "ch11", "ch36" ], "ch40": [ "ch36" ], "ch41": [ "ch04" ], "ch42": [ "ch36" ], "ch43": [ "ch04" ], "ch44": [ "ch36" ], "ch45": [ "ch04" ], "ch46": [ "ch04" ], "ch47": [ "ch11" ], "ch48": [ "ch04" ] }, "agent_instructions": { "protocol": "HTTP GET, JSON/Markdown response, no auth required", "base_url": "/claude-code-book/agent-api/", "recommended_flow": [ "GET manifest.json → 获取全书索引", "Check dependency_graph → 确定学习顺序", "GET chapters/{id}.json → 按需加载章节", "Use key_takeaways for quick assessment", "GET skills/{id}.md → 安装技能书" ], "token_tips": [ "manifest.json 约 3K tokens，建议缓存", "单章 JSON 约 5-15K tokens", "先读 key_takeaways 再决定是否加载全文" ] } }