ml-server

ML Take Home - backend

FastAPI Backend for the ML Take Home project.
This project is a simple REST API tha exposes a machine leaning service through a POST endpoint. The endpoint will allow a client to request a classification for a given image.

The project structure is as follows:

ml-server/
├── Pipfile
├── README.md
├── app
│   ├── app.py
│   ├── models
│   │   ├── response.py
│   │   └── schemas
│   │       └── classifier.py
│   ├── services
│   │   └── classifier.py
│   └── utils
│       └── strings.py
├── main.py
├── mlmodels
├── run.sh
├── tests
│   ├── conftest.py
│   ├── test_base.py
│   └── test_image.jpeg
└── train.py

• train.py: file that must contain everything related to the training of the machine learning model
• main.py: run the app with uvicorn
• mlmodels/: to store the trained machine learning model. The api must load the model from this path
• tests/: to test the api services
• app/: contains the app related files. This app follows an N-layer structure where the layers are:
  • models: contains the models and schemas used by the app.
  • services: contains the classifier service which exposes a function to classify an image
  • app.py: create the app and define the services as endpoints

Adding new functionalities

To add a new functionality, expose it as an endpoint. To do so, you must:

• register the endpoint in app/app.py
• create a new service in app/services
• import the service in app/app.py and call it in the endpoint definition
• if required, add any new model or schema in app/models

Quickstart

# Clone the codebase
git clone [email protected]:Monadical-SAS/ML-take-home.git
cd ML-take-home

# Init the server
cd ml-server 
pipenv install
pipenv shell
python3 main.py 

To run the tests

# Running the tests
cd tests
pytest