🔍 RAG & Fine-tuning | RAG与微调¶

🎯 Learning Objective | 学习目标：Learn to make AI use your private data to create a personalized AI assistant | 学会让AI使用你的私有数据，打造专属AI助手

🌟 Why Do We Need RAG and Fine-tuning? | 为什么需要RAG和微调？¶

An AI large model is like a smart graduate: AI大模型就像一个聪明的毕业生：

📚 Knowledgeable | 知识渊博 - But doesn't know your company's internal docs | 但不知道你公司的内部文档
🧠 Powerful | 能力强大 - But unfamiliar with your professional domain | 但不了解你的专业领域
💬 Good communicator | 善于交流 - But might give outdated information | 但可能给出过时信息

RAG and Fine-tuning are two methods to make AI "understand you"! RAG 和微调就是让AI"了解你"的两种方法！

🎯 RAG vs Fine-tuning | RAG对比微调¶

Comparison	RAG (Open-book Exam)	Fine-tuning (Professional Training)
对比项	RAG（开卷考试）	微调（专业培训）
🎯 Principle	Provide reference materials to AI	Retrain the AI
🎯 原理	给AI提供参考资料	重新训练AI
⏱️ Speed	Fast deployment	Requires training time
⏱️ 速度	快速部署	需要训练时间
💰 Cost	Low cost	Higher cost
💰 成本	低成本	较高成本
🔄 Updates	Real-time updates	Requires retraining
🔄 更新	实时更新	需要重新训练
📊 Use Cases	Knowledge Q&A	Style/capability changes
📊 适用	知识问答	风格/能力改变

📚 Chapter Contents | 本章内容¶

1️⃣ RAG Fundamentals | RAG基础 ¶

Introduction to Retrieval-Augmented Generation: 检索增强生成的入门：

📖 What is RAG | RAG是什么 - The open-book exam analogy | 开卷考试的比喻
🔍 How it works | 工作原理 - Retrieve → Augment → Generate | 检索→增强→生成
🛠️ Basic implementation | 基础实现 - Build a simple RAG hands-on | 动手搭建简单RAG

2️⃣ Advanced RAG Techniques | 高级RAG技术 ¶

Making RAG more powerful: 让RAG更强大：

🎯 Retrieval optimization | 检索优化 - Find the most relevant content | 找到最相关的内容
📊 Reranking | 重排序 - Improve answer quality | 提高答案质量
🔄 Hybrid retrieval | 混合检索 - Combine multiple methods | 结合多种方法

3️⃣ Fine-tuning Guide | 微调指南 ¶

Train your own AI: 训练你的专属AI：

📝 Data preparation | 数据准备 - Prepare training data | 准备训练数据
⚙️ LoRA technique | LoRA技术 - Low-cost fine-tuning | 低成本微调
🎓 Hands-on tutorial | 实战教程 - Step-by-step guide | 手把手教学

🎮 RAG Workflow Diagram | RAG工作流程图¶

User Question → 🔍 Vector Retrieval → 📄 Find Relevant Docs → 🤖 AI Generates Answer
用户问题    →    向量检索      →    找到相关文档    →    AI生成答案
    │                                                      │
    └───── Answer contains accurate private knowledge ─────┘
           答案包含准确的私有知识

💡 Scenario Selection Guide | 场景选择指南¶

Choose RAG when: | 选择 RAG 的场景：¶

✅ Need to access latest documents | 需要访问最新文档
✅ Data updates frequently | 数据经常更新
✅ Want answers with source citations | 希望答案有来源引用
✅ Need rapid deployment | 快速部署需求

Choose Fine-tuning when: | 选择微调的场景：¶

✅ Need specific output style | 需要特定的输出风格
✅ Domain-specific terminology understanding | 专业领域的术语理解
✅ Capability improvement for specific tasks | 特定任务的能力提升
✅ Data is relatively stable | 数据相对稳定

⏱️ Estimated Study Time | 预计学习时间¶

RAG Fundamentals | RAG基础：3-4 hours | 小时
Advanced RAG | 高级RAG：3-4 hours | 小时
Fine-tuning Guide | 微调指南：4-5 hours | 小时

Total | 总计：About 10-13 hours | 约 10-13 小时

💡 Pro Tip | 小贴士：We recommend learning RAG first - it's simpler and more practical! Fine-tuning is an advanced skill that you can learn later.

建议先学RAG，它更简单实用！微调是进阶技能，可以之后再学。