MathModelAgent

One-Minute Overview#

MathModelAgent is an intelligent agent system designed specifically for mathematical modeling competitions. It can automate the entire process from problem analysis to paper generation, reducing what would typically take 3 days down to just 1 hour. The system employs a multi-agent architecture with specialized roles (modeler, coder, writer), each configurable with different language models to produce complete, submission-ready papers.

Core Value: Drastically improve mathematical modeling efficiency by condensing a 3-day process into 1 hour while generating complete papers ready for submission.

Quick Start#

Installation Difficulty: Medium - Requires setup of Python, Node.js, and Redis environments, but offers Docker for simplified deployment

# Clone the project
git clone https://github.com/jihe520/MathModelAgent.git

# Docker Deployment (Recommended)
# After starting Docker container, access frontend: http://localhost:5173
# Backend API available at: http://localhost:8000

# Local Deployment
cd backend
pip install uv
uv sync
source .venv/bin/activate
uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload
cd frontend
pnpm install
pnpm run dev

Is this suitable for my needs?

• ✅ Mathematical modeling competition preparation: Quickly generate complete modeling papers
• ✅ Automated modeling workflow: Full automation from problem analysis to paper writing
• ❌ Need for original high-quality academic papers: Current generated content is for reference only
• ❌ Commercial use: Personal use is free, commercial use requires author permission

Core Capabilities#

1. Complete Intelligent Modeling Workflow#

• Automatically analyzes mathematical problems, performs modeling, writes code, corrects errors, and writes complete papers User Benefit: Reduces the traditional 3-day manual modeling process to just 1 hour

2. Multi-Agent Collaboration System#

• Includes specialized agents (modeler, coder, writer), each configurable with the most suitable language models User Benefit: Specialized divisions of labor improve modeling quality while supporting multiple language models for cost optimization

3. Code Interpreter Functionality#

• Supports local Jupyter interpreter with code saved as notebook for later editing
• Supports cloud code interpreter services like E2B and daytona User Benefit: Flexible code execution environments adapting to different use cases and configuration requirements

4. Multi-Model Support#

• Supports all major language models through litellm, allowing configuration of the most suitable model for each task User Benefit: Not locked to specific models, flexible model selection based on needs and cost

Technology Stack & Integration#

Development Languages: Python (backend), JavaScript/Node.js (frontend) Main Dependencies: FastAPI, uvicorn, Redis, Jupyter, litellm, pnpm Integration Method: API/Web interface/CLI

Ecosystem & Extensions#

• Plugins/Extensions: Supports prompt injection technology for individually configuring requirements and templates for each subtask
• Integration Capabilities: Planned integration with drawing tools (napki, draw.io, plantuml, svg, mermaid.js), web search tools, and RAG knowledge bases

Maintenance Status#

• Development Activity: Project is in experimental exploration phase with updates and maintenance based on author's availability
• Recent Updates: Has latest releases with ongoing feature additions and optimizations
• Community Response: Actively welcomes community contributions, encourages PRs and issues, with QQ group and Discord community support

Commercial & Licensing#

License: Custom License

• ✅ Commercial Use: Requires author permission for commercial applications
• ✅ Modification: Allowed but commercial use requires authorization
• ⚠️ Restrictions: Free for personal use, commercial use prohibited

Documentation & Learning Resources#

• Documentation Quality: Basic - Installation and usage guides provided, but documentation is still evolving as project is experimental
• Official Documentation: Deployment tutorials in project README
• Example Code: Demo video and notebook examples available, results stored in backend/project/work_dir/xxx/* directory