Deploying a Machine Learning model as REST API using Flask. Error: RuntimeError: Working outside of request context

Question

I've developed a classifier for test classification. and trying to access it using REST API. Here is the code:

clf_model = joblib.load('MNB_Clf.pkl','r')

app = Flask(__name__)

@app.route('/spend_api',methods=['POST'])
def make_predict():
    data = request.get_json(force=True)

    test_data = pd.read_csv(data)

    pred_proba_class = clf_model.predict_proba(test_data['ColumnName1'])


    final_pred_file = pd.DataFrame(pred_proba_class)

    sub_file = 'output_'+str(datetime.datetime.now().strftime("%Y-%m-%d-%H-%M")) + '.csv'
    return jsonify(results=final_pred_file.to_csv(sub_file))
if __name__ == '__main__':
  app.run(port = 9000,debug=True)

I am trying to send the CSV file to the api using the below code:

url = 'http://localhost:9000/spend_api'
files = {'file': ('Test_data_final.csv')}
r = request.post(url,files=files)

I am getting a Runtime Error. Could you please help how to resolve the issue.

Here is the error:

  RuntimeError                              Traceback (most recent call 
  last)
  <ipython-input-15-4b8522aa1eb0> in <module>()
         3 url = 'http://localhost:9000/spend_api'
         4 files = {'file': ('Test_data_final.csv')}
   ----> 5 r = request.post(url,files=files)
         6 
         7 

      C:\Users\pavansubhash_t\AppData\Local\Continuum\Anaconda2\lib\site -
    packages\werkzeug\local.pyc in __getattr__(self, name)
         345         if name == '__members__':
    346             return dir(self._get_current_object())
--> 347         return getattr(self._get_current_object(), name)
    348 
    349     def __setitem__(self, key, value):

C:\Users\pavansubhash_t\AppData\Local\Continuum\Anaconda2\lib\site-packages\werkzeug\local.pyc in _get_current_object(self)
    304         """
    305         if not hasattr(self.__local, '__release_local__'):
--> 306             return self.__local()
    307         try:
    308             return getattr(self.__local, self.__name__)

C:\Users\pavansubhash_t\AppData\Local\Continuum\Anaconda2\lib\site-packages\flask\globals.pyc in _lookup_req_object(name)
     35     top = _request_ctx_stack.top
     36     if top is None:
---> 37         raise RuntimeError(_request_ctx_err_msg)
     38     return getattr(top, name)
     39 

RuntimeError: Working outside of request context.

This typically means that you attempted to use functionality that needed an active HTTP request. Consult the documentation on testing for information about how to avoid this problem.

Answer 1

If I understand well your requirement, you have a pre-trained classifier and you want to serve classification over an API. Your API receives a CSV file as input and sends back another CSV file. To refactor your code to do that, you need to add the follow changes.

Fix how you are sending the file in a request:

Fix the way you are loading the csv file into the files dictionary as follows:

url = 'http://localhost:9000/spend_api'
files = {'file': open('Test_data_final.csv','rb')}
r = request.post(url,files=files)

you can find more details in this SO thread.

Fix how to send the json response

clf_model = joblib.load('MNB_Clf.pkl','r')

app = Flask(__name__)

@app.route('/spend_api',methods=['POST'])
def make_predict():
    data = request.get_json(force=True)

    test_data = pd.read_csv(data)
    pred_proba_class = clf_model.predict_proba(test_data['ColumnName1'])
    final_pred_file = pd.DataFrame(pred_proba_class)
    return jsonify(results=final_pred_file.to_dict(orient="records"))
if __name__ == '__main__':
  app.run(port = 9000,debug=True)

If you want to send a CSV file instead

here is a standalone working example:

from flask import Flask 
from flask import send_file 
from StringIO import StringIO 
import pandas as pd 

app = Flask("csv")

@app.route("/get_csv")
def hello():

    st = """col1|col2
    1|2
    3|4 
    """
    df = pd.read_csv(StringIO(st), sep="|")
    df.to_csv("/tmp/my_test_csv.csv", index=False, sep="|")
    return send_file("/tmp/my_test_csv.csv")


if __name__ == '__main__':
    app.run(port=5000)

As a side note, I suggest that you rethink your design to use JSON instead.