Skip to content

pdf-diagram-extract

Extract diagram and graph entities from PDFs as structured JSON — no LLM, pure code.

PyPI version Python 3.10+ License: MIT

What does it do?

pdf-diagram-extract takes a PDF file and extracts the geometric entities inside diagrams and graphs — shapes, edges, connections, and chart elements — returning everything as structured JSON.

Supported Diagram Types

Type Detection Method
Flowcharts Rectangles + diamonds + directed arrows
Network diagrams Circles/ellipses with interconnecting edges
Org charts Tree-structured rectangles
Bar charts Aligned rectangles with varying heights
Pie charts Circle with radial line segments
Line charts Evenly-spaced connected data points
Scatter plots Clusters of small unconnected dots

Key Features

  • 🔍 Vector extraction — parses native PDF drawing primitives (PyMuPDF)
  • 🖼️ Raster fallback — OpenCV contour detection for scanned PDFs
  • 🔗 Relationship resolution — links edges to shapes via spatial proximity
  • 📊 Chart detection — heuristic identification of common chart types
  • 📦 JSON output — fully serializable structured results
  • ⚡ 3 dependencies only — PyMuPDF, OpenCV-headless, NumPy

Quick Start

pip install pdf-diagram-extract

import diagramextract
import json

result = diagramextract.extract("document.pdf")
print(json.dumps(result.to_dict(), indent=2))

Get started → API Reference →