title	emoji	colorFrom	colorTo	sdk	short_description
DrugX	💊	blue	purple	docker	AI Platform for Preventing Dangerous Drug Mixes

💊 DrugX – AI Platform for Preventing Dangerous Drug Mixes

DrugX is an AI tool that checks medicine interactions to prevent dangerous side effects and keep patients safe.

It works by:

• Standardizing medicine names (so brand, generic, or even misspelled names are recognized).
• Checking for known interactions using trusted medical databases.
• Looking at real-world safety reports from the FDA’s adverse event system.
• Summarizing the results with AI into clear risk levels and safety notes.

This way, DrugX helps patients and clinicians avoid harmful drug combinations while staying simple to use and medically reliable.

🌐 Production Link

🚀 Try the Live Demo

📸 Screenshots

⚙️ Pre-requisites

Development Tools:

• Python 3.11.x (not 3.12+)
• uv package manager
• Docker Desktop
• Make:
  • Windows: winget install GnuWin32.Make
  • macOS: brew install make
  • Linux: sudo apt install make

🏗️ Project Structure

• The backend is built around specialized API clients (rxnorm, openfda, ddinter, and pubchem) that handle normalization, interaction lookups, and adverse event retrieval.
• The frontend is a Streamlit app that lets users enter medications, runs the analysis pipeline, and displays results with clear risk levels and safety notes.
• For data analysis, a Jupyter notebook (data/notebooks/data_exploration.ipynb) was used for EDA of the DDInter dataset, ensuring data quality and consistency before loading it into PostgreSQL.

🚀 Installation Instructions

1. Clone the Repository:

    git clone https://github.com/lisekarimi/drugx.git
    cd drugx

2. Set Up Environment:

   cp .env.example .env
   # fill in API keys and DB URI
   uv sync

3. Environment Variables Required:

Check .env.example — all required environment variables are listed there.

📚 Documentation

Detailed documentation for each component (RxNorm, DDInter, OpenFDA, PubChem, LLM, Monitoring, and testing) is available in the project Wiki.

▶️ Usage

Start services (app + database + jupyter):

make up

🧭 Architecture Overview

DrugX follows a modular, plug-and-play design:

• Workflow: Normalize drug names → check interactions (DDInter) → fetch adverse events (OpenFDA) → summarize with AI.
• Components: RxNorm, DDInter, OpenFDA, PubChem, LLM, and Monitoring.
• Extensible: New data sources or databases can be added easily without changing the core pipeline.

👉 Read the full architecture documentation

🧪 Unit Testing

Run tests:

make test

Other test commands are available in the Makefile.

👉 Read the full testing documentation

🛠️ Development

• Lint & Fix: make lint | make fix
• Hooks: make install-hooks
• Secrets: Gitleaks scans for sensitive data
• CI/CD: GitHub Actions runs build, lint, and tests
• DB: Use PostgreSQL in a cloud environment (e.g. Supabase, which offers free session pooling for concurrent access).

🚀 Deployment

Production Deployment

• Hugging Face Spaces: Deployment triggered manually via GitHub Actions when "deploy" is committed (see .github/workflows/deploy-hf.yml for details)

Kubernetes Deployment

This project includes Kubernetes deployment files for cloud deployment on GCP, AWS, or local clusters:

# Deploy to Kubernetes cluster
make k8s-build
make k8s-deploy
make k8s-url

The same YAML files can be used to deploy on Google Kubernetes Engine (GKE), Amazon EKS, or any Kubernetes cluster.

🛡️ Medical Safety Note

• DrugX is a clinical decision support tool, not a prescribing authority.
• No hallucinated mechanisms: only DDInter + curated rules.
• Disclaimers included when no interaction is found.