This repository contains the model-serving component for the REMLA project (Team 22).
It wraps the trained ML model (from model-training) in a RESTful API using Flask, enabling prediction requests via HTTP.

model-service/
├── app.py                     # Main Flask app exposing REST endpoints
├── requirements.txt           # Python dependencies
├── Dockerfile                 # Container definition for deployment
├── .github/workflows/         # CI/CD automation
│   └── release.yml            # Docker image build + tag + push
├── .gitignore
└── README.md

The Flask app exposes a REST endpoint to receive user text input and return a sentiment prediction.

• POST /predict
  • Request JSON: { "text": "the food was amazing" }
  • Response JSON: { "label": "positive" }

• Depends on lib-ml (installed via versioned Git tag)
  • Handles text preprocessing consistently with training
• lib-ml is installed via:

git+https://github.com/remla25-team22/[email protected]

• The service is packaged in a Docker container
• Uses multi-stage builds to keep image size small
• Versioned images are pushed to GitHub Container Registry
• URL of the model can be passed via MODEL_URL environment variable (if loaded remotely)