molandev/mineru-converter-api

By molandev

•Updated 7 months ago

Mineru HTTP API (port 10996) for PDF conversion to Markdown (CPU-only mode)

Image

Machine learning & AI

Developer tools

587

Overview Tags

molandev/mineru-converter-api repository overview

⁠MinerU PDF Converter Service - Overview

⁠📦 Source Code

GitHub: https://github.com/molandev/markdown-converter⁠
Gitee: https://gitee.com/molandev/markdown-converter⁠

⁠⚠️ Important Notice

This image runs in CPU-only mode with pipeline backend.

For GPU acceleration, please use the official MinerU image⁠.

CPU Mode: Suitable for moderate workloads, no GPU required

Pipeline Backend: Optimized for general document processing

GPU Mode: Visit official repository for high-performance GPU support

This image only supports PDF files.

Does NOT support .docx or .doc Word documents

Does NOT support .pptx, .xlsx, or other Office formats

For other document types, use our other converters:

Document Type Recommended Image
DOCX, PPTX, XLSX → Markdown molandev/markitdown-converter-api:latest
DOC → DOCX/PDF molandev/libreoffice-converter-api:latest

Document Type	Recommended Image
DOCX, PPTX, XLSX → Markdown	`molandev/markitdown-converter-api:latest`
DOC → DOCX/PDF	`molandev/libreoffice-converter-api:latest`

⁠🚀 What Is It?

A high-precision PDF parsing service powered by MinerU with advanced OCR capabilities. Built on GraalVM Native Image for cloud-native deployment, it transforms complex PDF documents into structured Markdown or JSON format with exceptional accuracy.

⁠✨ Core Features

⁠🔥 Superior PDF Parsing

OCR Engine: Extract text from scanned documents and images
Layout Preservation: Maintains document structure and formatting
Table Detection: Intelligent table extraction with proper formatting
Image Extraction: Preserves images with organized output

⁠🛡️ Production-Ready Reliability

Smart Concurrency Control: Semaphore-based rate limiting prevents system overload
Auto-Cleanup Mechanism: Scheduled task automatically purges temporary files
Graceful Degradation: Queuing mechanism with configurable timeouts

⁠🎯 Flexible Output Formats

Markdown: Clean, LLM-friendly text output
JSON: Structured data with full metadata
ZIP Archive: All output files in one package (md, json, pdf variants, images)

⁠📦 Output Files (ZIP Format)

File	Description
`result.md`	Extracted Markdown content
`result_content_list.json`	Content structure list
`result_middle.json`	Intermediate processing data
`result_model.json`	Model output data
`result_layout.pdf`	Layout analysis PDF
`result_origin.pdf`	Original PDF copy
`result_span.pdf`	Span annotation PDF
`images/`	Extracted images folder

⁠🎮 Quick Start

⁠Pull from Docker Hub

docker pull molandev/mineru-converter-api:latest

⁠Run in Seconds

docker run -d \
  --name mineru-converter \
  -p 10996:10996 \
  -v /data/temp:/app/temp \
  -e CONVERTER_MAX_CONCURRENT=5 \
  -e CONVERTER_RETENTION_MINUTES=30 \
  molandev/mineru-converter-api:latest

⁠📚 API Documentation

Access the complete API guide at: http://localhost:10996/help

⁠Convert Your First PDF

# Get Markdown (default)
curl -F "[email protected]" \
     http://localhost:10996/convert/mineru/upload \
     -o output.md

# Get JSON
curl -F "[email protected]" \
     "http://localhost:10996/convert/mineru/upload?outputFormat=json" \
     -o output.json

# Get all files (ZIP)
curl -F "[email protected]" \
     "http://localhost:10996/convert/mineru/upload?outputFormat=zip" \
     -o output.zip

# Or provide a URL
curl -X POST "http://localhost:10996/convert/mineru/url?fileUrl=https://example.com/doc.pdf&outputFormat=zip" \
     -o output.zip

⁠⚙️ Configuration

Parameter	Default	Description
`CONVERTER_MAX_CONCURRENT`	5	Max parallel conversions (MinerU is resource-intensive)
`CONVERTER_RETENTION_MINUTES`	30	Temp file retention
`CONVERTER_SCHEDULE_MINUTES`	5	Cleanup interval
`SERVER_PORT`	10996	Service port

⁠🔧 Advanced Usage

⁠Resource-Constrained Environment

docker run -e CONVERTER_MAX_CONCURRENT=2 \
           -e CONVERTER_RETENTION_MINUTES=60 \
           molandev/mineru-converter-api:latest

⁠High-Performance Setup

docker run -e CONVERTER_MAX_CONCURRENT=10 \
           --memory=8g \
           --cpus=4 \
           molandev/mineru-converter-api:latest

⁠💡 Why Choose This?

✅ OCR-Powered - Extract text from scanned PDFs with high accuracy
✅ Multiple Output Formats - Markdown, JSON, or complete ZIP archive
✅ Resource-Efficient - GraalVM Native Image for minimal footprint
✅ Production-Ready - Smart concurrency control & auto-cleanup
✅ Cloud-Native - Docker-first design, Kubernetes-ready

⁠🎯 Perfect For

AI Knowledge Base Pipelines: Convert PDFs to LLM-friendly Markdown
Document Digitization: OCR for scanned documents and archives
Data Extraction: Structured JSON output for downstream processing
Research & Analysis: Extract tables, figures, and structured content

⁠📊 API Reference

⁠Upload Endpoint

POST /convert/mineru/upload

Parameter	Type	Required	Description
file	File	Yes	PDF file to convert
outputFormat	String	No	Output format: `md` (default), `json`, `zip`

⁠URL Endpoint

POST /convert/mineru/url

Parameter	Type	Required	Description
fileUrl	String	Yes	Remote PDF URL
outputFormat	String	No	Output format: `md` (default), `json`, `zip`

Built with Spring Boot 3.5.6 + GraalVM 25 + MinerU

Precision PDF parsing with OCR intelligence. 📊

Tag summary

Recent tags

Content type

Image

Digest

sha256:d266fe868…

Size

7.1 GB

Last updated

7 months ago

docker pull molandev/mineru-converter-api