👻 GhostCite

This project was vibe coded. Built rapidly with AI-assisted development — expect rough edges, have fun, and contribute if you spot improvements!

GhostCite is a local-first web app that checks whether the references in an academic PDF are real. Upload a document, and GhostCite will extract every reference it finds, look each one up on Crossref, and tell you which ones appear to be genuine and which might be fabricated or "hallucinated".

✨ Features

• PDF & DOCX upload — drag-and-drop or select a file (up to 20 MB) through a simple web UI.
• Automatic reference extraction — detects the reference/bibliography section and parses individual entries, handling numbered, Harvard, and bracketed citation styles.
• Crossref verification — queries the Crossref API for each reference and classifies it as ✅ Verified, ⚠️ Partial Match, or ❌ Not Found.
• Summary report — download a standalone summary PDF listing every reference and its verification status.
• Fully local & private — runs entirely on your machine via Docker. Your documents are never sent to a third-party service (only individual reference strings are queried against Crossref's public API).

🚀 Quick Start

Prerequisites

• Docker and Docker Compose installed on your machine.

Run with one command

./start.sh

This builds the Docker image and starts the container. Once ready, open your browser at:

👉 http://localhost:8000

How to use

1. Open http://localhost:8000 in your browser.
2. Upload a PDF or DOCX file containing academic references.
3. Wait for GhostCite to extract and verify the references (this may take a moment depending on the number of references).
4. Review the results table showing each reference and its verification status.
5. Download the summary report PDF listing each reference and its status.

Manual Docker commands

# Build and start in background
docker compose up --build -d

# View logs
docker compose logs -f

# Stop the app
docker compose down

🛠 Tech Stack

• Python 3.12 with Flask and Gunicorn
• pypdf and ReportLab for PDF reading and report generation
• python-docx for DOCX support
• Crossref REST API for reference verification
• Docker for containerised deployment

Contact

For questions or queries about GhostCite, contact Dan at daniel.leightley@kcl.ac.uk.

📄 License

This project is provided as-is.