Perovskite MultiAI Agent (No Longer update due to confilct issue)

These code are written during my undergraduate. So, my code skills might not be clean and formal as now. Appologize for that.

I will no longer update since my year work has been put to n-author place. Sorry about that. But you are welcome for talk.

Schematic of Perovskite Muti-AI Agent

Functionality of this system

Please note that since the Unimol (mol predictor) and Synthesis advisor tool has large model weight. So, you can directlt fork the repository from Unimol and RetroExplainer. This repo is pre-released.

• Literature Data Extractor

  Helps pull useful information from research papers, such as material properties or experimental details, to support molecule evaluation.

• Retrieval-Augmented Generation (RAG)

  Improves molecule suggestions by referencing external sources, making the model’s output more informed and grounded.

• SAMGPT

  A molecule generator that can take structure constraints and desired properties into account. Useful for designing molecules like SAMs (self-assembled monolayers).

• UniMol

  A model used to predict properties of molecules, helping to screen and evaluate generated candidates before further steps.

• Supplier Search Tool (via MolPort API)

  Finds available molecules from commercial suppliers, along with their prices and other purchasing information.

• Molecule Utilities

  Offers tools for drawing molecules and converting between IUPAC names and SMILES strings:

  IUPAC $\rightleftharpoons$ SMILES

• Retrosynthesis Planner

  Suggests possible synthetic routes for the generated molecules, to help check how they might be made in practice.

How to use Perovskite Multi-AI Agent

The AI_agents jupyter notebook provides some examples of invoking the agent. You can change the API in the notebook.

PSC dataset collected from literatures

The all_PSC_dataset contains parameters extracted from approximately 319 papers (ACS, Springer, Elsevier, Weily, etc.)