JudgePeach/math-question-bank

What it solves

MathBank is a lightweight, semi-automated workbench for secondary school mathematics teachers to manage question banks and typeset exam papers. It eliminates the tedious manual work of digitizing math problems from various formats (PDF, Word, LaTeX) and the complex process of arranging them into professional, high-quality PDF exam papers that meet national exam standards.

How it works

The system runs locally on the user's computer and integrates several AI-driven workflows:

• Smart Ingestion: It uses a dual-strategy PDF parser (native vector extraction for speed and VLM-based OCR for images) and a structured Word converter to break down documents into individual questions.
• AI-Powered Content: It integrates LLMs (like DeepSeek) to automatically generate detailed solutions, teaching reflections, and even select questions based on difficulty and syllabus requirements.
• Visual Reconstruction: It can use VLMs to identify images in problems and optionally redraw them using TikZ code for high-fidelity geometric figures.
• Typesetting Engine: It provides a 1:1 A4 simulation canvas for intuitive drag-and-drop layout, which then renders into professional PDFs via a local LaTeX toolchain.

Who it’s for

Secondary school mathematics teachers who need to create professional exam papers and maintain a digital library of math problems without needing to be experts in LaTeX or complex software.

Highlights

• 1:1 A4 Simulation: A visual canvas for arranging papers with presets for national exam formats.
• Multi-format Import: Intelligent slicing of PDF, Word (.docx), Markdown, and LaTeX files.
• AI-Assisted Workflow: Automated question selection, solution generation, and geometric figure redrawing via TikZ.
• Syllabus Integration: Built-in support for major Chinese high school math curricula (PEP A/B, Jiangsu, and Shanghai versions).
• Local-First: Runs entirely on the user's machine for privacy and ease of use, with portable packages for Windows and macOS.