Files
curso-langchain/slides/pages/rag/16h2.md
T

876 B

layout
layout
default

Splitters por tipo de contenido

RecursiveCharacterTextSplitter.fromLanguage() usa separadores adaptados a cada formato

Markdown — respeta encabezados y listas

import { RecursiveCharacterTextSplitter } from "@langchain/textsplitters";

const mdSplitter = RecursiveCharacterTextSplitter.fromLanguage("markdown", {
  chunkSize: 1000, chunkOverlap: 200,
});

HTML — respeta estructura de etiquetas

const htmlSplitter = RecursiveCharacterTextSplitter.fromLanguage("html", {
  chunkSize: 1000, chunkOverlap: 200,
});

Código JS/TS — respeta funciones y clases

const codeSplitter = RecursiveCharacterTextSplitter.fromLanguage("js", {
  chunkSize: 1500, chunkOverlap: 200,
});

Otros lenguajes soportados: python, java, go, rust, ruby, php...