Macaulay2 AI Agent with Claude

I know everyone using Macaulay2 is smarter than enough to set these things up. But it would be nice if this can save you a couple minutes from typing them yourself.

Features

• Natural language interface for Macaulay2
• Powered by Claude Sonnet 4.5 (or your favorite LLM)
• Gröbner bases, Hilbert series, resolutions
• LangChain-based agent architecture

Prerequisites

• macOS 10.15+
• Python 3.8+
• Homebrew
• Anthropic API key with billing enabled

Quick Start

# Install Macaulay2
brew tap macaulay2/tap
brew install macaulay2

# Clone and setup
git clone https://github.com/Retieun/macaulay2-AI-agent-tutorial.git
cd macaulay2-agent
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Configure API key
echo 'ANTHROPIC_API_KEY=sk-ant-api03-your-key-here' > .env

# Run
python3 agent.py

Installation Details

1. Install Homebrew (if needed)

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

2. Install Macaulay2

brew tap macaulay2/tap
brew install macaulay2
M2 --version  # Verify installation

3. Setup Project

mkdir macaulay2-agent && cd macaulay2-agent
python3 -m venv venv
source venv/bin/activate
pip install anthropic langchain langchain-anthropic python-dotenv

4. Get API Key

1. Sign up at https://console.anthropic.com/
2. Add billing (minimum $5)
3. Create API key
4. Add to .env file:

echo 'ANTHROPIC_API_KEY=sk-ant-api03-your-key-here' > .env
chmod 600 .env

Usage

source venv/bin/activate
python3 agent.py

Example Queries

• Compute Gröbner basis of ideal(x^2-y, xy-1) in Q[x,y]
• Dimension of variety x^2+y^2+z^2-1?
• Hilbert series of Q[x,y,z]/(x^2, y^2, z^2)
• Minimal free resolution of ideal(x^2, xy, y^2)

Type quit to exit.

Project Structure

macaulay2-agent/
├── agent.py                 # Main agent
├── macaulay2_executor.py    # M2 code executor
├── m2_tool.py              # LangChain tool wrapper
├── requirements.txt         # Dependencies
├── .env                    # API key (gitignored)
└── README.md               # This file

Architecture

User Query → Claude Agent → Macaulay2Tool → Executor → M2 → Results → Claude → User

Cost

• ~$0.05-$0.20 per query
• $5 minimum credit
• Monitor: https://console.anthropic.com/settings/usage

Using Other LLMs

The agent is designed to work with Claude but can be adapted for other LLMs:

OpenAI (GPT-4)

1. Install: pip install langchain-openai openai
2. Update .env:

OPENAI_API_KEY=sk-proj-your-key-here

3. Modify agent.py:

from langchain_openai import ChatOpenAI  # Replace import

# In __init__, replace:
self.llm = ChatOpenAI(model="gpt-4", temperature=temperature, max_tokens=4096)

Google Gemini

1. Install: pip install langchain-google-genai google-generativeai
2. Update .env:

GOOGLE_API_KEY=your-key-here

3. Modify agent.py:

from langchain_google_genai import ChatGoogleGenerativeAI

# In __init__, replace:
self.llm = ChatGoogleGenerativeAI(model="gemini-pro", temperature=temperature)

Ollama (Local/Free)

1. Install Ollama: https://ollama.ai
2. Pull a model: ollama pull llama2
3. Install: pip install langchain-ollama
4. Modify agent.py:

from langchain_ollama import ChatOllama

# In __init__, replace:
self.llm = ChatOllama(model="llama2", temperature=temperature)
# Remove API key requirement

Note: Claude Sonnet 4 performs best for mathematical reasoning. Other models may require prompt adjustments for optimal results.

Troubleshooting

Issue	Solution
M2: command not found	brew install macaulay2
API key not found	Check .env in project root
Module not found	pip install -r requirements.txt
Timeout errors	Increase timeout in Macaulay2Tool(timeout=60)

Security

• .env is gitignored by default
• Never commit API keys
• Use .env.example for templates

License

MIT

Links

• Macaulay2
• Claude API
• LangChain