backend/README.md

DermaDetect Backend API (DermaVision)

This directory contains the robust Django backend for the DermaDetect application. It handles advanced AI inference, LLM-based clinical report generation, and dynamic PDF exporting.

Key Features

1. AI Skin Analysis (ONNX): Uses a compiled ONNX Computer Vision model (dermavision.onnx) to detect diseases like Vasculitis, Cellulitis, and Bullous Disease with high accuracy.
2. Multithreaded GradCAM Heatmaps: Generates visual saliency maps (occlusion sensitivity) utilizing multi-threading to pinpoint exactly where the AI is looking on the lesion, in under a second.
3. Groq LLM Integration: Uses the blazing-fast Groq API (Llama3-8b-8192) to ingest the visual AI's findings and the patient's symptoms, dynamically generating a structured Clinical Evaluation, Prescribed Therapy Regimen, and Patient Handout.
4. Professional PDF Reports: Uses reportlab to instantly compile side-by-side diagnostic images, clinical notes, and a digital verified signature into a highly polished, branded PDF complete with a verifiable QR code.

Setup Instructions

1. Environment Setup

It is highly recommended to use a Python virtual environment.

# Create a virtual environment
python -m venv derma

# Activate the virtual environment (Windows)
derma\Scripts\activate

# Activate the virtual environment (Mac/Linux)
source derma/bin/activate

2. Install Dependencies

Install the required Python packages:

pip install -r requirements.txt

3. Environment Variables

Create a .env file in the root of the backend folder and add your API keys. Use .env.example as a template:

GROQ_API_KEY=your_groq_api_key_here

4. Database Migrations

Apply the standard Django migrations to set up the SQLite database:

python manage.py migrate

5. Run the Server

Start the Django development server:

python manage.py runserver

The API will run at http://127.0.0.1:8000/.

API Endpoints

• POST /api/analyze/: Accepts a base64 encoded image and symptoms. Returns primary findings, confidence, multithreaded GradCAM heatmap, and the full Groq-powered clinical report.
• POST /pdf/: Accepts patient demographics, images, and clinical text to generate and return a base64 encoded PDF clinical assessment report.