Skip to main content

PDF files

You can use this version of the popular PDFLoader in web environments. By default, one document will be created for each page in the PDF file, you can change this behavior by setting the splitPages option to false.

Setup

npm install pdf-parse

Usage

import { WebPDFLoader } from "@langchain/community/document_loaders/web/pdf";

const blob = new Blob(); // e.g. from a file input

const loader = new WebPDFLoader(blob);

const docs = await loader.load();

console.log({ docs });

API Reference:

  • WebPDFLoader from @langchain/community/document_loaders/web/pdf

Usage, custom pdfjs build

By default we use the pdfjs build bundled with pdf-parse, which is compatible with most environments, including Node.js and modern browsers. If you want to use a more recent version of pdfjs-dist or if you want to use a custom build of pdfjs-dist, you can do so by providing a custom pdfjs function that returns a promise that resolves to the PDFJS object.

In the following example we use the "legacy" (see pdfjs docs) build of pdfjs-dist, which includes several polyfills not included in the default build.

npm install pdfjs-dist
import { WebPDFLoader } from "@langchain/community/document_loaders/web/pdf";

const blob = new Blob(); // e.g. from a file input

const loader = new WebPDFLoader(blob, {
// you may need to add `.then(m => m.default)` to the end of the import
pdfjs: () => import("pdfjs-dist/legacy/build/pdf.js"),
});

Eliminating extra spaces

PDFs come in many varieties, which makes reading them a challenge. The loader parses individual text elements and joins them together with a space by default, but if you are seeing excessive spaces, this may not be the desired behavior. In that case, you can override the separator with an empty string like this:

import { WebPDFLoader } from "@langchain/community/document_loaders/web/pdf";

const blob = new Blob(); // e.g. from a file input

const loader = new WebPDFLoader(blob, {
parsedItemSeparator: "",
});

Was this page helpful?


You can also leave detailed feedback on GitHub.