Sentinel

Sentinel: An AI-Powered PubMed Search Agent

Sentinel is a terminal-based AI application designed to streamline the process of locating relevant PubMed publications. Using a powerful local AI model, Sentinel accepts natural language search queries and transforms them into effective MeSH search strategies. This innovative approach simplifies the retrieval of research articles from PubMed via a user-friendly interface.

Key Features

• Natural Language Processing: Allows the user to input search queries in plain language for a seamless experience.
• AI-Generated Research Purpose: Utilizes the phi3.5 model connected with phidata to produce detailed and relevant research objectives.
• Comprehensive Search Execution: Accesses the PubMed API through Biopython, retrieving up to 250 top articles tailored to the search criteria, sorted by relevance and publication date.
• Efficient Data Management: Saves search results along with metadata into a relational SQL database for future reference.
• Export Functionality: Facilitates the export of search results and metadata to an Excel file, organized into user-friendly tabs.

Example Usage

To execute a new search from the terminal, use the command:

python main.py --query "efficacy of placebo injections in knee osteoarthritis patients between 2000 and 2025" --export  

In the absence of command line arguments, Sentinel will prompt for user input through an interactive menu, providing options to run a new search or view historical results.

Sample Output

When initiating a search query, the application will retrieve and display the number of relevant PubMed entries found, as exemplified below:

Found 250 PubMed IDs  
Retrieved 249 search results  
Data stored successfully in the database.  

Sentinel is a significant tool for researchers, providing a robust methodology to access and analyze scientific literature quickly and effectively.

For a visual overview, watch the demo here.

File Structure

sentinel/  
├── src/  
│   ├── __init__.py  
│   ├── agent.py  
│   ├── config.py  
│   ├── menu.py  
│   └── utils/  
│       ├── __init__.py  
│       ├── xlsx_export.py  
│       ├── database.py  
│       ├── extract_values.py  
│       └── pubmed_search.py  
├── tests/  
│   ├── __init__.py  
│   └── test_pubmed_search.py  
├── main.py  
├── requirements.txt  
└── setup.sh  

Requirements

The project relies on the following Python packages:

• phidata
• ollama
• sqlalchemy
• requests
• biopython
• openpyxl
• pandas
• python-dotenv

Configuration

Users must configure an .env file with necessary connection parameters for the Entrez API, including their email and API key.

Conclusion

Sentinel provides a state-of-the-art solution for accessing scholarly articles in the biomedical field, enhancing productivity and fostering insights with every search.