title	Pharma Sidekick
emoji	📈
colorFrom	pink
colorTo	gray
sdk	gradio
sdk_version	6.5.1
app_file	app.py
pinned	false
license	mit
short_description	A pharmacist aid in dispensing and prescription analysis

🩺 Pharma Sidekick: AI Clinical Assistant

Pharma Sidekick is an intelligent AI agent designed to assist clinical pharmacists in validating prescriptions. By leveraging autonomous agentic workflows, it acts as a "second pair of eyes" to ensure patient safety and optimize pharmacotherapy.

🔗 Live Demo on Hugging Face

🚀 Features

Unlike standard chatbots, Sidekick doesn't just generate text. It uses a suite of specialized tools to perform clinical checks before forming a conclusion:

• Allergy & Interaction Checking: Screens for drug-drug interactions and cross-sensitivity with patient allergies.
• Dosing Verification: Validates dosages based on:
  • Pediatrics: Weight-based dosing for children.
  • Geriatrics: Beers Criteria and age-related adjustments.
  • Renal Function: Adjustments for reduced eGFR/CrCl.
• Pregnancy & Lactation: Safety profiles for pregnant or breastfeeding patients.
• Data Source: Utilizes openFDA data for authoritative drug information.
• Structured Assessment: Delivers a clear "Dispense" or "Do Not Dispense" verdict with clinical reasoning and recommendations.

🏗️ Technical Architecture

This project demonstrates Agentic AI engineering concepts:

• LangGraph: Orchestrates the cyclic graph workflow, managing state between the "Worker" (which uses tools) and the "Evaluator" (which ensures quality).
• OpenAI GPT-4o: Provides the reasoning and clinical judgment capabilities.
• Gradio: Powers the interactive web interface.
• openFDA: Source for regulatory drug data.

🛠️ Installation & Usage

Prerequisites

• Python 3.10+
• An OpenAI API Key

Steps

1. Clone the repository

   git clone https://github.com/codypharm/pharma-sidekick.git
   cd pharma-sidekick

2. Set up environment variables Create a .env file in the root directory:

   OPENAI_API_KEY=your_sk_key_here

3. Install dependencies This project uses uv for fast package management, but standard pip works too.

   Using uv:

   uv sync

   Using pip:

   pip install -r requirements.txt

4. Run the application

   # Using uv
   uv run app.py
   
   # Or using python directly
   python app.py

5. Access the UI Open your browser to http://127.0.0.1:7860.

🎓 Origin & Credits

This project was built as part of The Complete Agentic AI Engineering Course by Ed Donner. It adapts the core concepts of building autonomous agents to the domain of clinical pharmacy.

⚠️ Disclaimer

Prototype Status: This tool is a proof-of-concept prototype.

• Limitation: The hosted specific demo on Hugging Face (Free Tier) may timeout on complex queries.
• Clinical Use: This tool is for educational and demonstrative purposes only. It should not replace professional clinical judgment. Always verify outputs with official monographs.

📄 License

MIT