Language: English | 中文
MinerU Web is a web application for document parsing and result review. It provides file upload, asynchronous parsing, original file preview, Markdown preview, PDF source tracing, Popo-enhanced result comparison, and Markdown export. The current version targets MinerU 3.4.4. The business backend now calls the official MinerU HTTP service instead of depending on MinerU internal Python APIs.
Current release: v3.4.4.
- FastAPI backend and Vue 3 frontend
- Email/password login with user-scoped files, statistics, and settings
- Redis-backed asynchronous parsing queue
- Configurable worker concurrency for matching MinerU API capacity
- MinIO/S3 storage for source files, Markdown outputs, and images
- Parse service health checks, with MinerU API connection and version information on the settings page
- Upload support for PDF, Office documents, and common image formats
- Original file preview for PDF, Office, images, and text
- PDF source tracing with bbox highlights, block linking, type filters, and table comparison
- Markdown preview for raw Markdown, page-based Markdown, Popo-enhanced Markdown, and OCR/Popo comparison
- MinerU task status and progress visibility on the files page
- Single-file and batch export for Markdown, page Markdown, and Popo Markdown
- Optional MinerU-Popo postprocessing; Popo failures do not block the base parsing result
- Support for official MinerU 3.4.4 backend options
- Multi-architecture images for the business backend, worker, and frontend
- Linux server deployment with
mineru-routerfor unified multi-GPU scheduling - macOS Apple Silicon deployment with MinerU API on the host and business services in Docker
Create the environment file:
cp .env.example .envEdit .env and set a MinIO endpoint that can be reached by both the browser and containers, for example:
MINIO_ENDPOINT=SERVER_IP:9000
WORKER_REPLICAS=1
WORKER_CONCURRENCY=1Linux / server deployment:
docker compose --env-file .env -f docker-compose.yml up -dmacOS Apple Silicon deployment:
docker compose --env-file .env -f docker-compose.mac.yml up -d --buildAfter startup, open:
- Web:
http://SERVER_IP:8088 - Backend API:
http://SERVER_IP:8000 - MinerU router:
http://SERVER_IP:8002 - MinIO console:
http://SERVER_IP:9001
Register an email account on the first visit to start using the app.
For Linux server deployment, macOS Apple Silicon setup, model downloads, MinerU Router, multi-GPU scheduling, MinIO endpoint configuration, and verification commands, see Deployment Guide.
Home - system overview and quick actions
Upload - drag and drop documents into the parsing queue
Files - upload and manage multiple document formats
Preview - parse and inspect document content
Table tracing - filter table blocks and compare the PDF region with the Markdown table
Settings - backend selection and parse service status
mineru-web/
├── backend/ # FastAPI backend, worker, database models, and tests
├── frontend/ # Vue 3 frontend
├── docs/
│ └── deployment.md # Deployment guide
├── docker-compose.yml # Linux / server deployment
├── docker-compose.mac.yml # macOS host MinerU API deployment
└── README.md
Common environment variables are listed in .env.example. Full details are in the Deployment Guide.
Frequently used options:
MINIO_ENDPOINT: MinIO address reachable by both browser and containers.WORKER_REPLICAS/WORKER_CONCURRENCY: worker replica count and per-worker concurrency.MINERU_API_USE_ASYNC_TASKS: enable the MinerU/tasksasynchronous API.POPO_ENABLED: enable MinerU-Popo postprocessing.
Verification commands are listed in the Deployment Guide.
Common checks:
cd backend
uv run pytest tests -vcd frontend
npm run buildThis project version follows the compatible MinerU version. Current version v3.4.4 corresponds to MinerU 3.4.4:
lpdswing/mineru-web-frontend:v3.4.4lpdswing/mineru-web-backend:v3.4.4lpdswing/mineru-web-mineru-api:v3.4.4
Use tag v3.4.4 when publishing the GitHub Release. After the release is published, .github/workflows/docker-build.yml builds and pushes Docker images with the same release tag. If only MinerU Web changes while the compatible MinerU version stays the same, use a suffix such as v3.4.4-web.1.
- Adapted to MinerU 3.4.4
- Updated the MinerU API image build to install
mineru[core]==3.4.4 - Bumped project release images to
v3.4.4 - Kept the existing HTTP API client and page-aware Markdown patch; their upstream integration points are unchanged from MinerU 3.4.0
- Adapted to MinerU 3.4.0
- Updated the MinerU API image build to install
mineru[core]==3.4.0 - Bumped project release images to
v3.4.0
- Adapted to MinerU 3.3.1
- Updated the MinerU API image build to install
mineru[core]==3.3.1 - Updated official backend options to
vlm-engineandhybrid-engine - Kept legacy
vlm-auto-engineandhybrid-auto-enginesettings compatible - Added explicit hybrid
effort=highfor MinerU 3.3.1 requests, configurable withMINERU_API_HYBRID_EFFORT
- Adapted to the official MinerU 3.2.3 HTTP API
- Switched parsing to sidecar / router mode
- Preserved MinIO/S3 transfer for images and Markdown artifacts
- Removed MinerU internal dependencies from the business backend and worker
- Added parse service status to the settings page
- Added official MinerU 3.2.3 backend parameter support
- Added email login and user-scoped data isolation
- Added worker concurrency configuration and multi-GPU router deployment notes
- Added optional MinerU-Popo postprocessing, preview, and export
- Enhanced PDF source tracing with bbox highlights, source maps, type filters, and table comparison
- Added original file preview for Office, image, and text files
- Added parse progress visibility backed by MinerU task state
- Added batch delete and batch export
- Delete now cleans up parsed MinIO artifacts together with the source file
This project is licensed under AGPL-3.0. See LICENSE for details.