💊 Prescripto - Medical Assistant & Prescription Translator

Prescripto is an innovative web application designed to translate, validate, and schedule medical treatment plans. Starting from a simple photo of a prescription (no matter how illegible the doctor's handwriting is), it extracts data using a hybrid AI system, validates it against official databases, and automatically synchronizes administration alarms directly into the patient's Google Calendar.

🏆 Project Context: This project was born within the AI Ideation Hackathon (Innovation Labs Cluj). Subsequently, the prototype (MVP) code was completely refactorized into a robust production architecture, object-oriented (OOP), being prepared for scalability.

---

📑 Table of Contents

1. Main Features
2. Architecture and Technologies
3. Test Materials and Business Documentation
4. Complete Installation and Setup Guide
5. How to Use (User Flow)
6. Project Structure
7. Roadmap and Future Improvements

---

✨ Main Features

• 📸 Hybrid OCR and Scanning: Uses Google Cloud Vision to extract raw text from medical prescriptions, receipts, or hospital discharge summaries with maximum accuracy.
• 🤖 AI Analysis with Cross-Validation (RAG): The Google Gemini 1.5 engine analyzes raw text and intelligently correlates it with:
  • The Official Disease Nomenclature (PDF)
  • The complete Medications List (CSV)
  • Result: Eliminates AI hallucinations and returns only medications that physically exist in pharmacies.
• 💊 Simplified Leaflets (For Everyone): Generates clear explanations, free of excessive medical jargon, regarding the purpose of the treatment, side effects, and administration advice.
• ✅ Validation Interface: The user maintains absolute control. They can edit concentrations and manually choose generic alternatives available in pharmacies, based on INN (International Nonproprietary Name) / DCI.
• 📅 Google Calendar Synchronization: Automatically creates recurring events and alarms (pop-up reminders 10 minutes before), sending direct invitations to the patient's phone (via a Service Account).

---

🏗️ Architecture and Technologies

The application complies with Clean Architecture, Separation of Concerns, and SOLID principles, being logically divided for easy testing and maintenance:

• Models (models/): We use Pydantic for the strict definition of data schemas (e.g., PrescriptionData, ConfirmedMedication). It ensures Type Safety and automatic validation of the JSON structures returned by the AI.
• Services (services/): Completely isolates the business logic:
  • ai_service.py: Manages prompts and communication with Google Vision and Gemini.
  • data_service.py: Manages queries in local files (Pandas for CSV, PyPDF2 for PDF).
  • calendar_service.py: Manages authentication and payloads for the Google Calendar API.
• UI Interface (ui/): Modular Streamlit components (sidebar.py, views.py, components.py) that turn main.py into a simple traffic controller.

---

📊 Test Materials and Business Documentation

For a complete evaluation of the project, in the /docs folder (or directly in the repository) you will find:

1. Lean Canvas (LeanCanvas.jpeg): The business model created at the Hackathon, detailing the problem, solution, customer segment, and monetization model.
2. Pitch Presentation (prezentare_prescripto_pitch.pdf): The presentation (Pitch) delivered to the jury at Innovation Labs. It contains the identified problem, proposed solution, market analysis, and product development strategy.
3. Test Image (RetetaTest.jpg): A complete medical prescription ready to be uploaded into the application to test the OCR + AI flow.
4. Databases:
   • Disease codes nomenclature (in /data).
   • INN/DCI medications database (in /data).

---

🚀 Complete Installation and Setup Guide

Prerequisites

• Python 3.9 or a newer version installed on your system.
• A Google Cloud Console account with the following APIs enabled:
  • Google Cloud Vision API
  • Google Calendar API
• O cheie API Google Gemini validă (obținută de pe Google AI Studio).

Step 1: Cloning and environment preparation

1. Clone this repository locally:

   git clone [https://github.com/username-ul-tau/prescripto.git](https://github.com/username-ul-tau/prescripto.git)
   cd prescripto

2. Create and activate an isolated Python virtual environment:

   # For Windows
   python -m venv .venv
   .venv\Scripts\activate
   
   # For macOS/Linux
   python3 -m venv .venv
   source .venv/bin/activate

3. Install the required packages:

   python -m pip install --upgrade pip
   pip install -r requirements.txt
   

Step 2: Configuring Environment Variables (Secrets)

1. Create a file named exactly .env in the root of the project.
2. Add your Gemini key:

   GEMINI_API_KEY="AIzaSy...your-key..."

Step 3: Configuring Google Calendar (Service Account)

1. In the Google Cloud Console, create a Service Account.
2. Generate a private key for this account and download it as a JSON file.
3. Rename the downloaded file to credentials.json and place it in the root folder of the project (next to main.py and .env).

Step 4: Starting the Application

1. Run the command below in your terminal to start the web server:

   streamlit run main.py

2. The application will be available in the browser at: http://localhost:8501.

---

🎯 How to Use (User Flow)

1. Scan (Upload): Drag and drop the test image (RetetaTest.jpeg) over the upload area in the interface.
2. Process (AI Magic): Press the "Process Prescription with AI" button. Track the status in real time: Vision reads the text, Gemini extracts the medical logic, and Pandas searches for alternatives in the database.
3. Validate (Human in the Loop): Review the generated tabs. If the AI identified Noliprel 5mg, but you bought Noliprel 10mg, use the drop-down menu to select the alternative purchased from the pharmacy. You can read a summary of the leaflet generated by the AI on the spot.
4. Schedule (Calendar Synchronization): Enter your email address (e.g., ion.popescu@gmail.com) in the final section and press "Add Alarms".
5. Completion: Verify the entered email address. You will receive official invitations from the assistant robot. Once accepted, the alarms will activate on your phone.