新增wiki命令文件
This commit is contained in:
25
wiki/concepts/Indexing.md
Normal file
25
wiki/concepts/Indexing.md
Normal file
@@ -0,0 +1,25 @@
|
||||
---
|
||||
title: "Indexing"
|
||||
type: concept
|
||||
tags: [RAG, 索引, 数据处理]
|
||||
---
|
||||
|
||||
## Definition
|
||||
Indexing是将外部文档切分并建立索引的过程,是RAG的第一阶段。
|
||||
|
||||
## Core Mechanism
|
||||
1. 文档加载(Load)
|
||||
2. 文档切分(Split):按段落、句子或Token切分
|
||||
3. 向量化(Embed):通过Embedding Model转为向量
|
||||
4. 存储(Store):存入Vector Store
|
||||
|
||||
## Key Properties
|
||||
- 切分策略影响检索质量
|
||||
- 受Context Window限制
|
||||
- 需要平衡粒度
|
||||
|
||||
## Connections
|
||||
- [[RAG]] ← 阶段1 ← [[Indexing]]
|
||||
- [[Retrieval]] ← 下游 ← [[Indexing]]
|
||||
- [[Embedding Vector]] ← 输出 ← [[Indexing]]
|
||||
- [[Vector Store]] ← 目标 ← [[Indexing]]
|
||||
Reference in New Issue
Block a user