server.py

from bottle import route, run, request, response, post
from sklearn.externals import joblib
import json
import pandas

model = joblib.load('model.pkl')


@route('/predict', method="POST")
def predict():
    response.content_type = 'application/json'
    try:
        body = request.json
        body = body if isinstance(body, list) else [body]
        X = pandas.DataFrame.from_dict(body)
        
        if (hasattr(model, 'get_booster')):
            # HACK https://github.com/dmlc/xgboost/issues/1238
            X = X[model.get_booster().feature_names]

        result = model.predict(X)
        return json.dumps(result.tolist())
    except Exception as error:
        response.status = 500
        return json.dumps({'error': str(error)})


@route('/predictproba', method="POST")
def predictproba():
    response.content_type = 'application/json'
    try:
        body = request.json
        body = body if isinstance(body, list) else [body]
        X = pandas.DataFrame.from_dict(body)

        if (hasattr(model, 'get_booster')):
            # HACK https://github.com/dmlc/xgboost/issues/1238
            X = X[model.get_booster().feature_names]

        result = pandas.DataFrame(model.predict_proba(X),
                                  columns=model.classes_)
        return result.to_json(orient="records")
    except Exception as error:
        response.status = 500
        return json.dumps({'error': str(error)})


@route('/featurenames')
def featurenames():
    response.content_type = 'application/json'
    try:
        result = model.get_booster().feature_names
        return json.dumps(result)
    except Exception as error:
        response.status = 500
        return json.dumps({'error': str(error)})


#run(host='localhost', port=8080, debug=True, reloader=True)
run(host='0.0.0.0', port=8080)

# curl -H "Content-Type: application/json" -X GET -d '[{"sepal length (cm)":4.4}]' http://localhost:8080/predictproba