predict-api-server.py

# run: gunicorn --bind=0.0.0.0:9696 --chdir=server --log-level=debug  predict:app
# predict: curl -X POST -H 'Content-Type: application/json' localhost:9696/predict -d '{"customer_age":100,"gender":"F","dependent_count":2,"education_level":2,"marital_status":"married","income_category":2,"card_category":"blue","months_on_book":6,"total_relationship_count":3,"credit_limit":4000,"total_revolving_bal":2500}'

import os

from model_service import ModelService
from multiprocessing.pool import RUN
import logging
from datetime import datetime

from flask import Flask, request, jsonify
from pymongo import MongoClient
import requests
from model_loader import ModelLoader


EVIDENTLY_SERVICE_ADDRESS = os.getenv('EVIDENTLY_SERVICE', 'http://127.0.0.1:8085')
MONGODB_ADDRESS = os.getenv("MONGODB_ADDRESS", "mongodb://127.0.0.1:27018")
MONGODB_DB = os.getenv("MONGODB_DB", "prediction_service")
MONGODB_COLLECTION = os.getenv("MONGODB_COLLECTION", "data")
RUN_ID = os.getenv('RUN_ID', 'No RUN_ID provided')

log = logging.getLogger('gunicorn.error')

def get_mongo_collection(database_name, collection_name):
    mongo_client = MongoClient(MONGODB_ADDRESS)
    db = mongo_client.get_database(database_name)
    return db.get_collection(collection_name)

mongo_collection = get_mongo_collection(MONGODB_DB, MONGODB_COLLECTION)
model, dv = ModelLoader().load_model_from_mlflow(RUN_ID)
model_service = ModelService(model, dv)

app = Flask('duration-prediction')
app.logger.handlers = log.handlers


# GET payload example:
    # {
    #   "customer_age": 100,
    #   "gender": "F"",
    #   "dependent_count": 2,
    #   "education_level": 2,
    #   "marital_status": "married",
    #   "income_category": 2,
    #   "card_category": "blue",
    #   "months_on_book": 6,
    #   "total_relationship_count": 3,
    #   "credit_limit": 4000,
    #   "total_revolving_bal": 2500
    # }
@app.route('/predict', methods=['POST'])
def predict_endpoint():
    log.info(f'Request payload: {request.get_data()}')

    input = request.get_json()
    # log.info(f'Request json: {input}')

    features = model_service.prepare_features(input)
    pred = model_service.predict(features)

    result = {
        'churn chance': float(str(pred)),
        'model_run_id': RUN_ID
    }

    save_to_db(input, float(pred))
    send_to_evidently_service(input, float(pred))
    
    log.info(f'Response payload: {result}')
    return jsonify(result)


def save_to_db(record, prediction):
    rec = record.copy()
    rec['prediction'] = prediction
    rec['created_at'] = datetime.now()
    rec['model_run_id'] = RUN_ID
    mongo_collection.insert_one(rec)


def send_to_evidently_service(record, prediction):
    rec = record.copy()
    rec['prediction'] = prediction
    requests.post(f"{EVIDENTLY_SERVICE_ADDRESS}/iterate/capstone", json=[rec])



if __name__ == "__main__":
    log.info('App starting ...')
    app.run(debug=True, host='0.0.0.0', port=9696)