Healthcare Assistant RAG Model

Overview

This project is a Healthcare Assistant built using a Retrieval-Augmented Generation (RAG) model. The system provides accurate and relevant healthcare advice based on user queries by retrieving information from the NHS A-Z website and augmenting it with a Large Language Model (LLM).

The project is developed as part of the MLH "Hack for Hackers" hackathon.

Features

• Comprehensive Healthcare Information: The system leverages the NHS A-Z data, which covers a wide range of medical conditions, treatments, self-care advice, and medicines.
• Retrieval-Augmented Generation (RAG): Combines a vector similarity search with an LLM to deliver personalized and accurate responses.
• Efficient Query Handling: For each query, the system retrieves the most relevant documents from the database to enhance the LLM's output.
• Scalable Backend: Powered by MongoDB Atlas for efficient storage and retrieval of embeddings and documents.

System Architecture

Components:

1. Frontend:

   • Accepts user queries and displays responses.
   • Communicates with the backend via REST API.

2. REST API:

   • Acts as a bridge between the frontend and the RAG engine.
   • Sends user queries to the RAG engine and returns the response to the frontend.

3. RAG Engine:

   • MongoDB Atlas:
     • Stores document embeddings and a vector search index.
     • Performs similarity searches to retrieve the top 5 relevant documents for each query.
   • Web Scraper:
     • Extracts data from the NHS A-Z website and stores it in the database.
   • LLM:
     • Processes the user query along with the retrieved documents to generate a biased, contextually enriched response.

Data Source

The primary data source is the NHS A-Z website, which provides comprehensive healthcare information. The data is scraped and stored in MongoDB Atlas as embedded documents to facilitate efficient similarity search.

Technologies Used

• Frontend: React, HTML, CSS
• Backend: Python, FastAPI, Langchain
• Database: MongoDB Atlas
• LLM: OpenAI GPT
• Web Scraper: Python (BeautifulSoup, Requests)
• Embedding Generation: OpenAI embeddings API

Installation and Setup

1. Clone the repository:

   git clone <repository-url>
   cd nhs-rag

2. Install dependencies for the backend:

   cd rest
   npm pip install -r requirements.txt

3. Install dependencies for the web scraper:

   cd scraper
   pip install -r requirements.txt

4. Set up environment variables: Create a .env file in the rest directory with the following details:

   MONGODB_ATLAS_CLUSTER_URI=<your-mongodb-atlas-uri>
   OPENAI_API_KEY=<your-openai-api-key>
   NHS_URL=https://www.nhs.uk/conditions/

5. Run the web scraper to populate the database:

   python scraper/scraper.py

6. Start the backend server:

   fastapi dev main.py

7. Run the frontend (optional):

   cd frontend
   npm install
   npm start

How It Works

1. User Query: The user submits a query through the frontend.
2. Document Retrieval: The REST API sends the query to the RAG engine, which performs a similarity search on the MongoDB database to fetch the top 5 relevant documents.
3. Augmented Response: The query and retrieved documents are sent to the LLM to generate a contextually enriched response.
4. Response Delivery: The response is sent back to the frontend and displayed to the user.

Future Improvements

• Enhance the web scraper to update data periodically from the NHS website.
• Add multi-language support for a wider audience.
• Incorporate additional healthcare datasets to improve the breadth of information.
• Optimize the embedding generation and similarity search processes for faster responses.

Project Members

• Ehtisham UL Hasan
• Muhammad Junaid
• Shamsher

Contributing

Contributions are welcome! Feel free to open issues or submit pull requests to improve the project.

License

This project is licensed under the MIT License.

Acknowledgments

• NHS A-Z Website for providing the data.
• MLH for organizing the "Hack for Hackers" hackathon.
• OpenAI for the LLM and embeddings API.

---

We hope this Healthcare Assistant helps users make informed healthcare decisions efficiently!