pmatigakis / classifier

A Flask web application that exposes a scikit-learn classifier over an REST API endpoint

A Flask web application that exposes a scikit-learn classifier over an REST API endpoint

This application is used by TopicAxis API to classify the contents of a web page. The scikit-learn model that is used can be found at the trained-models repository. See the documentation for instructions on how to use this model.

Installation

Download the source code and install the latest stable version using pip.

git clone https://github.com/pmatigakis/classifier.git
cd classifier
git fetch --all
git checkout master
pip install .

Configuration

A settings.py file is required in the folder where the classifier will be started. See the example settings file in the configuration folder for information about what configuration settings are available. Most of the configuration variables in the example settings file can be modified using environment variables. Rename configuration/.env.template to configuration/.env and change what is required.

Running

Go into the folder with the settings file and start the server.

classifier-server

The classification endpoint is at http://<HOST>:<PORT>/api/v1/predict/<CLASSIFIER>. To run a prediction execute a POST request to that endpoint. For example to run a prediction using the demo Iris classifier execute the following request.

curl -X POST -H "Content-Type: application/json" -d '{"data": [[4.8,3.0,1.4,0.1]]}' http://192.168.1.103:8022/api/v1/predict/iris

The response status code should be 200 and the response should look like this

{
    "results": ["Iris-setosa"]
}