MatSciLLM

MatSciLLM is primarly an LLM-based application designed for material science research, enabling content extraction and analysis from PDFs along with access to the Materials Project Database.

Features

• User Interface - User-friendly chat interface with convenient PDF management.

• Specific Question - Extracts the most relevant content from all uploaded PDFs based on a provided query.

• Every PDF - Provides answers to the query for each PDF individually.

• Database Access - Retrieves all unique stable element combinations for a given material from Materials Project (expandable).

Setup

This repository includes Makefile for easy installation.

Prerequisite

• Ollama

1. Clone Repository

   git clone https://github.com/xandaaaa/MatSciLLM.git
   cd MatSciLLM

2. Add API Key
   Insert your Materials Project API key into the .env.example file. This key is required for accessing the Materials Project API.

3. Create .env File
   Copy .env.example to .env to set up environment variables.

   cp .env.example .env

4. Install dependencies
   Use the provided Makefile to install dependencies.

   Note: For Python, it’s recommended to create and activate a virtual environment before running this.

   make install

5. Run application
   Start the application using the Makefile, then open the application at the specified localhost address.

   make run

    

Notes:

• To stop the application, press Ctrl + C in the terminal.
• The default LLM model is llama3:latest.
• Alternatively, you can use a different LLM model by changing the model_name parameter in the following functions:
  • backend/dbMP.py → llm_extract_element(model_name=...)
  • backend/rag_chain.py → build_rag_chain(model_name=...)
  • backend/rag_summarizer.py → build_rag_summarizer(model_name=...)