app.py

import pickle
import json
from flask import Flask, request
import pandas as pd


FEATURES = pickle.load(open("churn/models/features.pk", "rb"))

model = pickle.load(open("churn/models/model.pk", "rb"))
column_equivalence = pickle.load(open("churn/models/column_equivalence.pk", "rb"))

# create the Flask app
app = Flask(__name__)

def convert_numerical(features):
    output = []
    for i, feat in enumerate(features):
        if i in column_equivalence:
            output.append(column_equivalence[i][feat])
        else:
            try:
                output.append(pd.to_numeric(feat))
            except:
                output.append(0)
    return output

@app.route('/query')
def query_example():
    features = convert_numerical(request.args.get('feats').split(','))
    response = {
        'response': [int(x) for x in model.predict([features])]
    }
    return json.dumps(response)

if __name__ == '__main__':
    # run app in debug mode on port 3001
    app.run(debug=True, port=3001)