Decoder. plain-English AI glossary

Document Loader

▲ Rising

Also called Document Loader

The plumbing that pulls raw files in and turns PDFs, web pages, and docs into clean text a pipeline can use.

Think of it like

The intake desk at a clinic — whatever form you arrive in, it gets your details onto the one standard chart everyone else reads.

Example

A loader ingests a folder of PDFs, Word docs, and HTML pages, strips headers and footers, and emits plain text ready for chunking.

How it actually works

Before anything gets chunked or embedded, raw sources have to be read and normalized — and real documents are messy. Loaders parse each format, pull out text (and sometimes tables and metadata), and drop boilerplate. It’s unglamorous but consequential: a loader that mangles a PDF’s columns or drops a table poisons every downstream step. Garbage in here is garbage in the whole pipeline.

For product teams

The first mile of any RAG system — flaky loaders are a top hidden cause of bad answers.

For engineers

Parse heterogeneous sources into normalized text plus metadata; the quality floor for chunking and embedding downstream.

Chunking — Its output feeds this step.
RAG — The start of a RAG pipeline.
Grounding — Clean context helps grounding.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome