edgeparse-core

High-performance PDF-to-structured-data extraction engine.

edgeparse-core implements a 20-stage processing pipeline that extracts text, tables, images, and semantic structure from PDF documents and produces structured output in Markdown, JSON, HTML, or plain text.

Usage

use edgeparse_core::{convert, api::config::ProcessingConfig};
use std::path::Path;

let config = ProcessingConfig::default();
let doc = convert(Path::new("report.pdf"), &config)?;

println!("Pages: {}", doc.number_of_pages);
for element in &doc.kids {
    // process extracted content elements
}

Output Formats

Generate output in multiple formats using the output modules:

use edgeparse_core::output;

let markdown = output::markdown::to_markdown(&doc)?;
let json = output::legacy_json::to_legacy_json_string(&doc, "report")?;
let html = output::html::to_html(&doc)?;
let text = output::text::to_text(&doc)?;

Features

• Tagged PDF support — uses PDF structure tree for semantic extraction
• Table detection — border-based and cluster detection methods
• Reading order — XY-Cut++ algorithm for correct reading order
• Image extraction — embedded or external image output
• Content safety — filters hidden text, off-page content, tiny text
• PII sanitization — optional personal data redaction
• Multi-column layout — automatic column detection and ordering

Feature Flags

License

Apache-2.0