SmartCIF

An intelligent workflow system for processing CIF (Crystallographic Information File) files of Metal-Organic Frameworks (MOFs), featuring automated structure analysis, solvent identification, and literature retrieval.

Overview

This project provides an automated pipeline for processing MOF crystal structures, including:

• CIF File Processing: Convert structures to P1 space group and primitive cells
• CSD Database Integration: Retrieve structure information from Cambridge Structural Database
• Cluster Analysis: Identify and analyze atomic clusters in crystal structures
• Solvent Identification & Removal: Automatically detect and remove solvent molecules
• Literature Retrieval: Download and parse related research papers
• Batch Processing: Parallel processing of multiple CIF files

Key Features

• Multi-Agent Architecture: Specialized agents for different tasks (CSD info, cluster analysis, solvent identification, etc.)
• LLM-Powered Analysis: Uses DeepSeek LLM for intelligent structure analysis and decision-making
• Automated Workflow: End-to-end processing from raw CIF files to cleaned structures
• Parallel Processing: Efficient batch processing with multiprocessing support
• PDF Integration: Automatic paper download and content extraction

Requirements

• Python 3.8+
• DeepSeek API key (configured in config.py)
• CSD Python API (optional, for CSD database access)
• Required Python packages (see environment.yml)

Quick Start

1. Configure API keys in src/config.py:

   DEEPSEEK_API_KEY = "your_api_key"
   UNPAYWALL_EMAIL = "your_email"

2. Set input/output paths in config.py:

   INPUT_FOLDER = "path/to/cif/files"
   P1_OUTPUT_FOLDER = "path/to/p1/output"
   FINAL_OUTPUT_FOLDER = "path/to/final/output"

3. Run single file processing:

   python src/run_workflow.py

4. Run batch processing:

   python src/batch_process_parallel.py

License

Contact

sopranos@sjtu.edu.cn