Dr. Artificial

Dr. Artificial is an AI-powered virtual doctor designed for the Dedalus Datathon Andalucia 2025. It leverages state-of-the-art (SOTA) large language models to provide insightful and data-driven responses in a healthcare environment, aiming to improve patient data analysis and clinical decision support.

Features

• Conversational AI: Capable of answering medical-related queries with high accuracy while minimizing hallucinations.
• Contextual Memory: Retains conversation history to support chained question-answering.
• Data Visualization: Generates graphical and non-textual responses using Mermaid.js.
• Actionable Insights: Recommends next steps based on knowledge base data.
• Scalable and Modular Architecture: Implements prompt engineering techniques and efficiently processes synthetic datasets.
• Efficient LLM Usage: Optimized for sustainable and cost-effective model inference using Amazon Bedrock via LiteLLM.

Project Structure

├── README.md  # Project documentation
├── requirements.txt  # Dependencies
├── datos/  # Synthetic patient datasets
│   ├── chats/  # Conversation history
│   ├── r_dataton/  # Data analysis scripts
│   └── users/  # User profiles
├── model/  # Core AI model
│   ├── api/  # API endpoints
│   ├── config/  # Model configurations
│   ├── data/  # Processed medical data
│   └── utils/  # Utility scripts
├── web/  # Web application interface
│   ├── app/  # Flask application
│   ├── static/  # CSS and JS files
│   ├── templates/  # HTML templates
│   └── run.py  # Web server entry point
└── logo/  # Branding assets

Prerequisites

• API Key: A valid API key for Amazon Bedrock is required.
• Model Connection: Connection may need to be adjusted to match the used model.

Installation

1. Clone the Repository:

   git clone https://github.com/yourusername/dr-artificial.git
   cd dr-artificial

2. Set Up a Virtual Environment (Optional but Recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows use: venv\Scripts\activate

3. Install Dependencies:

   pip install -r requirements.txt

Usage

1. Run the Web Application:

   python web/run.py

AI Model Details

• LLM Model: Claude 3.5 Sonnet via Amazon Bedrock (accessed through LiteLLM). Any model could be used.
• Embeddings: Amazon Titan Text Embeddings V2
• RAG (Retrieval-Augmented Generation): Utilized for efficient contextual retrieval and cost reduction.
• Response Format: Markdown-based with Mermaid.js for graphical representation.

Future Improvements

• Vector Database Storage: Transition from file-based storage to a vector database.
• AI-Powered Preprocessing: Automate data extraction from databases via queries leveraging other AIs, as DeepSeek.
• File & Image Attachments: Support for handling diverse document types.
• Automated Report Generation: Beyond simple assistance, full document generation.

Contributing

We welcome contributions! Please open an issue or submit a pull request for improvements.

License

This project was developed as part of the Dedalus Datathon Andalucia 2025 and is intended for educational and research purposes only. It is provided "as is" without warranty of any kind. Use at your own risk.