UI_README.md

Medical Q&A Bot - Web UI

This is a user-friendly web interface for the Health Query Classifier & Research Retriever system.

Features

✨ Clean, Modern Interface - Built with Gradio for an intuitive user experience

🎯 Query Classification - Automatically triages queries as medical or administrative with confidence scores

📚 Intelligent Retrieval - Retrieves relevant medical research from PubMed and Miriad databases

🔍 Dual View Modes - View results in formatted HTML or raw JSON

⚙️ Customizable Settings - Adjust number of results and toggle reranker for better accuracy

Quick Start

1. Install Dependencies

Make sure you have the updated requirements installed:

pip install -r requirements.txt

This will install Gradio along with all other dependencies.

2. Prepare Data

If you haven't already, build the corpora:

python -m adapters.build_corpora

3. Launch the Web UI

python app.py

The interface will be available at: http://127.0.0.1:7860

Using the Interface

1. Enter Your Query - Type your health-related question in the text box
2. Adjust Settings (optional):
   • Number of Results: Control how many documents to retrieve (1-50)
   • Use Reranker: Enable for more accurate results (slower)
3. Click "Analyze Query" or press Enter
4. View Results:
   • Classification: See how your query was categorized
   • Formatted View: Readable cards with document information
   • JSON View: Raw data for technical analysis

Example Queries

Try these example queries to see the system in action:

• "I'm having a really bad rash on my hands. Is there anything stronger than aquaphor I can use?"
• "I'm traveling to South America soon. Do I need to get any vaccines before I go?"
• "worst headache of my life with fever and stiff neck"
• "Hey is there any way I can get an appointment in the next month?"

Configuration Options

Sharing Your Interface

To create a public shareable link (72 hours), modify app.py:

demo.launch(
    share=True,  # Creates a public link
    server_name="127.0.0.1",
    server_port=7860,
)

Custom Port

Change the port if 7860 is already in use:

demo.launch(
    server_port=8080,  # Use your preferred port
)

Authentication

Add password protection:

demo.launch(
    auth=("username", "password"),  # Simple auth
)

Architecture

The UI integrates with your existing codebase:

User Query → Gradio Interface → Classifier → Retriever → Results Display

• Frontend: Gradio (Python-based web framework)
• Classification: Uses your trained classifier model
• Retrieval: Hybrid search (BM25 + Dense embeddings + RRF)
• Reranking: Optional cross-encoder reranker

Troubleshooting

Models Not Loading

Ensure you have the classifier checkpoint and data files:

ls -la data/corpora/

Port Already in Use

Change the port number in app.py or kill the process using port 7860.

Gradio Import Error

Make sure Gradio is installed:

pip install gradio

No Medical Documents Found

Verify that corpora files exist in data/corpora/:

• medical_qa.jsonl
• miriad_text.jsonl
• unidoc_qa.jsonl

Run the build script if missing:

python -m adapters.build_corpora

Advanced Features

API Mode

Gradio automatically creates a REST API alongside the web UI. Access the API docs at:

http://127.0.0.1:7860/docs

Embedding the Interface

You can embed the Gradio interface in other web applications using iframes:

<iframe src="http://127.0.0.1:7860" width="100%" height="800px"></iframe>

Team

• David Gray
• Tarak Jha
• Sravani Segireddy
• Riley Millikan
• Kent R. Spillner

License

See the main README.md for project license information.