MI-Agent

An agentic workflow for materials-informatics (MI) engineers, built with LangGraph and powered by OpenAI models. MI-Agent codifies the end-to-end MI pipeline—data loading, merging, feature selection, EDA, AutoML baselining, hyperparameter tuning, and executive reporting—into reusable nodes orchestrated as a LangGraph. LangSmith integration tracks and visualizes your graph executions. The result? MI workflows that run in seconds instead of hours, boosting your productivity by an order of magnitude.

🚀 Why MI-Agent?

• Agentic LangGraph design lets you hit “play” on a full MI pipeline
• 10× faster: eliminate boilerplate and manual scripting
• Extensible nodes: swap in your own extractors, metrics, or plots
• LangSmith-backed for graph tracking, versioning, and observability
• Production-ready: versionable, testable, pip-installable

🛠️ Prerequisites

• Conda (Miniconda or Anaconda)
• Python 3.10
• OpenAI API key
• LangSmith API key

Installation via pip

1. Create & activate a conda environment

   conda create -n mi_agent python=3.10 -y
   conda activate mi_agent
   

2. Install via pip

   pip install materials_informatics_agent
   

3. Configure your API keys

   MI-Agent will automatically look for a file named .env in your current working directory (or any parent) and load any keys it finds.

   In the folder where you’ll run the CLI (or in any ancestor), create a file called .env containing:

   OPENAI_API_KEY=sk-…
   LANGCHAIN_API_KEY=lsv2_…
   

4. Prepare your problem file

   MI-Agent requires a .txt file (an example is provided in the sample_problem.txt in the project root of the source code) which contains:

   • your problem description

   • relative paths to your CSV(s), including any folder prefix (e.g. data/sample_data.csv)

   Example problem.txt:

   You are tasked with predicting alloy strength from composition data...
   
   - data/sample_data_1.csv: Contains experimental results...
   - data/sample_data_2.csv: Contains formulation recipes...
   

5. Run the agent

   Now, start the mi_agent pipeline as below:

   mi_agent --problem-file <path/to/problem.txt> --output-dir <path/to/output_dir>
   

   MI-Agent will:

   • Identify & load the CSV(s) listed in the problem file
   • Merge files if needed
   • Select target & features
   • Propose & execute EDA
   • Save all generated code (*.py) for EDA analysis and images (*.png) generated during EDA into <output_dir>
   • Run multiple ML models, select top 5, tune hyperparameters, and choose the best model
   • Generate and save a 5-page technical summary into <output_dir>
   • Log all reasoning steps to LangSmith