Docker & Containerization

To ensure consistent performance of the machine learning models and the Intelligence Engine across different environments, Soteria utilizes Docker for backend containerization. This approach encapsulates the Python 3.13+ runtime, complex ML dependencies (PyTorch, Scikit-Learn), and the Flask middleware into a single, deployable unit.

Intelligence Engine Containerization

The backend service, which handles AST analysis and vulnerability scoring, is designed to run as a containerized application. This is particularly important given the specific versions of data science libraries required for the Hybrid Stacking Model.

Prerequisites

Docker Desktop (or Docker Engine on Linux)
Backend Environment Variables: A .env file containing your production secrets.

Local Build and Execution

To replicate the production environment locally or prepare for a manual deployment, follow these steps:

Create a Dockerfile (if not already present) in the root or /backend directory:

FROM python:3.13-slim

# Set working directory
WORKDIR /app

# Install system dependencies for ML libraries
RUN apt-get update && apt-get install -y \
    build-essential \
    && rm -rf /var/lib/apt/lists/*

# Copy requirements first to leverage Docker cache
COPY backend/requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Copy source code
COPY . .

# Expose the Flask port
EXPOSE 5001

# Start the Intelligence Engine
CMD ["python", "middleware/app.py"]

Build the Image:
```
docker build -t soteria-backend .
```

Run the Container:

docker run -p 5001:5001 \
  -e SECRET_KEY="your_secure_key" \
  -e SQLALCHEMY_DATABASE_URI="sqlite:///kyber.db" \
  soteria-backend

Environment Configuration

The containerized Intelligence Engine relies on specific environment variables to manage security and database connections. Ensure these are injected during the Docker run command or via your CI/CD pipeline:

Deployment to Render

Soteria is pre-configured for deployment on Render using the "Web Service" type with Docker.

Link Repository: Connect your GitHub repository to Render.
Select Environment: Choose Docker as the runtime.
Specify Context: Ensure the Docker build context is set to the root of the project to allow the container to access both /middleware and /backend directories.
Model Persistence: Since the project uses serialized models (acidModel.pkl, acidModel_neural.pt), ensure these files are tracked in Git or downloaded into the container during the build process to enable immediate scanning capabilities.

Data Persistence & Volumes

The Intelligence Engine uses a local SQLite database (kyber.db) for user authentication and scan history.

For Development: Standard container storage is sufficient.
For Production: If you require persistent user data across container restarts on platforms like Render or AWS, you must mount a Persistent Disk/Volume to the directory containing kyber.db. Alternatively, update SQLALCHEMY_DATABASE_URI to point to a managed PostgreSQL instance.

Automated Retraining (Watchdog)

The watch_data.py process can be run as a sidecar container or a background process within the main container. It monitors the /backend/data directory for new training samples. When new malicious patterns are identified, the containerized pipeline automatically triggers trainerModel_Hybrid.py to update the "Structural DNA" matrix and retrain the classifier.