OpenDataLoader PDF — AI-Ready Document Parsing

Overview

OpenDataLoader PDF is an open-source (Apache-2.0) PDF parser. A Java engine reads the PDF and writes Markdown, JSON, HTML, plain text, an annotated PDF or a Tagged PDF; the Python and Node.js packages bundle the JAR and start it for you. The default mode is deterministic, runs on CPU and sends nothing to a network. An optional hybrid mode routes hard pages (borderless tables, scans, formulas, charts) to a local Docling-based server.

Every JSON element carries its type, page number and bounding box, so an answer in a RAG pipeline can point back to the exact place in the source file.

Instructions

Step 1: Install

Java 11 or newer must be on PATH — the packages do not ship a JVM. Without it the CLI stops with 'java' command not found.

java -version                               # must print 11 or higher
sudo apt install default-jre-headless       # Debian/Ubuntu (Java 11, 17 or 21 depending on the release)
brew install --cask temurin                 # macOS