mathmodelagent automates the end-to-end mathematical modeling workflow and produces a submission-ready, formatted paper. It generates typeset PDF output from a mathematical modeling problem description.
Project overview
The project combines human-in-the-loop decision points, composable task-oriented skills, and 17 Typst paper templates to turn a modeling problem description into a submission-ready PDF paper.
Project type
AI Agent · RAG
Use cases
Documents & Office · Knowledge Q&A
Deployment
Refer to project documentation
License
License pending
Best for
Creators, researchers, and developers who want to automate the mathematical modeling process and produce a formatted paper ready for submission
Users who prefer to review and guide key decisions during an automated workflow rather than rely on a fully autonomous pipeline
Key capabilities
Pauses at key nodes to wait for user approval with 6 decision actions (confirm, edit, regenerate, ask, skip, abort).
Provides 17 Typst paper templates for mainstream competitions to automatically generate typeset PDF papers.
Project distillation driven by Skills that can be run independently to perform specific tasks like analysis or paper writing.
Limitations and risks
The project is in an experimental iterative demo phase with many bugs. Adopters should expect instability during use.
The AI-generated output is currently insufficient to win mathematical modeling competitions directly.
Getting started
Clone the repository, then start the backend and frontend services. The setup difficulty is medium because it requires installing and configuring Redis as well as setting API keys in configuration files. The project offers a WebUI and a CLI for interaction.
Configure LLM API keys through the WebUI or a .env file. The project supports any LLM via LiteLLM.