checkpaper¶

AI论文验证智能体系统 - 基于Python和FastAPI的论文验证agent

项目信息
仓库	https://github.com/Maicarons/checkpaper
主要语言	Python
Stars	0
类型	原创项目
许可证	—
最近更新	2026-07-28

CheckPaper - AI Paper Verification Agent¶

About¶

CheckPaper is an AI-powered academic paper verification agent system. It automatically detects various issues in academic papers — including formatting non-compliance, citation errors, reference fraud, and data integrity problems — helping researchers improve the quality and reliability of their work.

Features¶

Format Checking — Validates heading hierarchy, numbering consistency, font uniformity, page layout, and TOC accuracy
Figure & Table Reference Check — Cross-references all figure/table definitions against in-text citations; detects orphan and uncited references
Citation Integrity Check — Validates that every in-text citation matches a reference list entry; detects duplicates and missing references
Data Source Verification — Verifies data source authenticity and accessibility
Data Processing Verification — Runs GRIM/SPRITE statistical consistency tests; validates p-values and confidence intervals
Reference Verification — Verifies references against Crossref and Semantic Scholar APIs; checks DOI validity and detects potentially fraudulent citations

Supported Formats¶

Format	Extensions	Parser
PDF	`.pdf`	PyMuPDF
Word	`.docx`, `.doc`	python-docx
LaTeX	`.tex`, `.latex`	pylatexenc
BibTeX	`.bib`	bibtexparser

Quick Start¶

Prerequisites¶

Python 3.11+
Node.js 18+ (for frontend development)
Docker & Docker Compose (optional)

Installation¶

## 1. Clone the repository
git clone https://github.com/Maicarons/checkpaper.git
cd checkpaper

## 2. Configure environment variables
cp .env.example .env
## Edit .env with your OpenAI API key and settings

## 3. Install backend dependencies
pip install uv
uv sync

## 4. Start the backend server
uvicorn backend.app.main:app --reload --port 9031

## 5. Install and start the frontend
cd frontend
npm install
npm start

The application will be available at: - Frontend: http://localhost:9032 - Backend API: http://localhost:9031 - Swagger UI: http://localhost:9031/docs

Docker Deployment¶

## Start all services
docker-compose up -d

## Production with MySQL
docker-compose --profile production up -d

Project Structure¶

checkpaper/
├── backend/                    # Backend services
│   ├── app/
│   │   ├── api/v1/endpoints/   # API route handlers
│   │   ├── core/               # Configuration, security, DB
│   │   ├── services/           # Business logic
│   │   │   ├── document/       # Document parsing service
│   │   │   ├── validation/     # Validation orchestration
│   │   │   ├── agent/          # AI agent service
│   │   │   └── report/         # Report generation
│   │   ├── schemas/            # Pydantic data models
│   │   ├── prompts/            # LLM prompt templates
│   │   └── main.py             # FastAPI application entry
│   ├── mcp_server/             # MCP tool server
│   │   ├── tools/              # Tool implementations
│   │   └── server.py           # MCP server entry
│   └── tests/                  # Test files
├── frontend/                   # React frontend application
│   └── src/
│       ├── pages/              # Page components
│       ├── components/         # Reusable components
│       ├── services/           # API client
│       └── hooks/              # Custom React hooks
├── docs/                       # VitePress documentation
├── docker-compose.yml          # Docker configuration
├── pyproject.toml              # Python project configuration
└── .env.example                # Environment variable template

Tech Stack¶

Component	Technology
Backend	Python 3.11+ · FastAPI · SQLModel
Frontend	React 18 · TypeScript · Ant Design
AI Agent	OpenAI Agents SDK · MCP Protocol
Document Parsing	PyMuPDF · python-docx · pylatexenc
Reference Verification	Crossref API · Semantic Scholar API
Database	SQLite (dev) · MySQL (production)
Deployment	Docker · Docker Compose

API Reference¶

Endpoint	Method	Description
`/api/v1/documents/upload`	POST	Upload a paper document
`/api/v1/documents/`	GET	List uploaded documents
`/api/v1/documents/{id}`	GET	Get document details
`/api/v1/documents/{id}/parse`	POST	Parse document content
`/api/v1/validation/start`	POST	Start a validation task
`/api/v1/validation/tasks/{id}/results`	GET	Get validation results
`/api/v1/validation/types`	GET	List validation types
`/api/v1/reports/`	GET	List reports
`/api/v1/reports/{id}/download`	GET	Download report (md/pdf/html)
`/health`	GET	Health check

Full API documentation available at http://localhost:9031/docs (Swagger UI) when the backend is running.

User documentation (VitePress) is available at http://localhost:9032/docs/ when the frontend is running. Start it with: cd docs && npm install && npm run dev

Validation Types¶

Type	Description
`format`	Format and structure checking
`figure_table`	Figure and table reference validation
`citation`	Citation integrity checking
`data_source`	Data source verification
`data_processing`	Statistical consistency verification
`reference`	Reference authenticity verification

Documentation¶

Full documentation is built with VitePress:

cd docs
npm install
npm run dev     # Development server
npm run build   # Production build

English: docs/guide/introduction.md
中文: docs/zh/guide/introduction.md

Configuration¶

Key environment variables (see .env.example for all options):

Variable	Description	Default
`OPENAI_API_KEY`	OpenAI API key	—
`OPENAI_BASE_URL`	API base URL	`http://192.168.56.1:8990`
`DATABASE_URL`	Database connection	`sqlite:///./checkpaper.db`
`DEBUG`	Debug mode	`false`
`SECRET_KEY`	Application secret	`change-me-in-production`

Testing¶

## Backend tests
pytest

## Backend with coverage
pytest --cov=backend/app --cov-report=html

## Frontend tests
cd frontend && npm test

## Linting
ruff check . && mypy .

Contributing¶

We welcome contributions! Please see CONTRIBUTING.md for details, or read the Developer Guide.

License¶

This project is licensed under the GNU Affero General Public License v3.0 — see LICENSE for details.

Acknowledgments¶

Built with FastAPI, OpenAI, PyMuPDF, Ant Design, and the MCP Protocol.

中文文档