dSupertramp / PubGPT

Extract relevant informations in PubMed papers in a simple way.

PubGPT 💉📄

PubGPT is a tool that can extract lots of relevant informations inside PubMed papers in a very simple way:

• Abstract
• Full text (when available)
• Genes
• Diseases
• Associations between genes and diseases (powered by LLMs and LangChain)
• MeSH terms
• Other terms

Local PDF parsing architecture

Install

To install everything, you need poetry. First of all, create a virtual environment with the command python3 -m venv venv_name and activate it with source venv_name\bin\activate.

After that, you can install poetry with the command pip install poetry and then run poetry install.

Run the app

To run the webapp, use the command:

cd pubgpt/
streamlit run app.py

License

PubGPT is licensed under the MIT License. See the LICENSE file for more details.

Todo

• Improve code
• Add sidebar where you can select the LLM and define credentials
• Add the extraction of more informations
• Create Python package
• Create documentation
• Add a fine-tuned model
• Improve web interface
• Add docstrings
• Add more LLMs for associations extraction
• Add webapp
• Add LangChain for PDFs