整理文件路径:Technical→AI/
This commit is contained in:
42
wiki/sources/rag-cong-ru-men-dao-jing-tong.md
Normal file
42
wiki/sources/rag-cong-ru-men-dao-jing-tong.md
Normal file
@@ -0,0 +1,42 @@
|
||||
---
|
||||
title: "RAG从入门到精通系列1:基础RAG"
|
||||
type: source
|
||||
tags: [rag, LLM, retrieval, vector-database]
|
||||
date: 2025-12-18
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/AI/RAG从入门到精通系列1:基础RAG.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:RAG基础概念和实现流程
|
||||
- 问题域:LLM缺乏最新和私有数据
|
||||
- 方法/机制:索引→检索→生成的三步流程
|
||||
- 结论/价值:RAG是连接LLM与外部数据源的通用方法
|
||||
|
||||
## Key Claims
|
||||
- RAG = Retrieval Augmented Generation,检索增强生成
|
||||
- Indexing:将文档切分并转换为Embedding向量存入向量数据库
|
||||
- Retrieval:根据问题语义向量检索相关文档块
|
||||
- Generation:将问题和相关文档输入LLM生成答案
|
||||
- Embedding Model的Context Window有限(512-8192 token),需切分文档
|
||||
|
||||
## Key Concepts
|
||||
- [[RAG]]:检索增强生成
|
||||
- [[Embedding]]:将文本转换为数值向量的技术
|
||||
- [[向量数据库]]:存储和检索向量表示的数据库
|
||||
- [[文档切分]]:将长文档分割成符合Embedding窗口的块
|
||||
- [[Context Window]]:模型能接受的上下文长度限制
|
||||
|
||||
## Key Entities
|
||||
- [[LangChain]]:RAG实现框架
|
||||
- [[Qdrant]]:向量数据库
|
||||
- [[BAAI]]:Embedding模型提供商
|
||||
|
||||
## Connections
|
||||
- [[LLM]] ← uses ← [[RAG]]
|
||||
- [[RAG]] ← includes ← [[索引]]
|
||||
- [[RAG]] ← includes ← [[检索]]
|
||||
- [[RAG]] ← includes ← [[生成]]
|
||||
|
||||
## Contradictions
|
||||
Reference in New Issue
Block a user