Fanchouille / generic_api

Dockerized deployment of Scikit model with FastAPI

Generic API

This a dockerized API to expose a scikit model with fastAPI (dummy model with Iris dataset).

Development

Anaconda local environment support

Install Anaconda local environment as below:

./install.sh

Set Anaconda local environment as ${PWD}/.conda as Python interpreter of the project (using PyCharm)

Structure

API endpoints

main.py

Schemas of inputs and outputs

schemas.py

Custom preprocessing and get_prediction function

processing.py

Pickled or joblib scikit model

models/

Configuration files

config/

What do I need to do ?

Clone repo

Change inputs format in schemas.py (& outputs if needed)

Change processing code if needed.

Train & pickle (with joblib) a scikit model and store it in app/model folder. Be sure to use the same version of scikit during training & inference.

Build & run docker image (see below)

Build & Run :

Build image :

• cd generic_api
• docker build -t generic_api .

Run image :

remove -d to keep CLI attached

docker run -d -p 80:80 generic_api

Access :

Your API documentation is available at http://127.0.0.1/docs

Deployment :

See DEPLOYMENT.md for simple Azure deployment.