ml_server.py

from bottle import run, post, request
from threading import Thread
from queue import Queue
import datetime
import requests
import shutil
import tempfile
from os.path import join
import time

HOST = '0.0.0.0'
PORT = 22222
ENDPOINT = 'https://127.0.0.1/api/v0/finish'


def log(*arg):
    t = '{0:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now())
    print(t, *arg)


## job queue
jobQueue = Queue()


## worker
def worker():
    print('Running worker')
    while True:
        job = jobQueue.get(block=True)
        log('processing job:', job)
        try:
            process(job)
        except:
            pass
        log('finished job:', job)

## modify this
def predict(imgPath):
    time.sleep(1)
    return [
        {'class': 'a', 'score': 0.90},
        {'class': 'b', 'score': 0.48},
        {'class': 'c', 'score': 0.38},
        {'class': 'e', 'score': 0.32}
    ]


def process(job):
    r = requests.get(job['image_url'], verify=False, timeout=10, stream=True)
    print(r.status_code)
    if r.status_code == 200:
        tempdir = tempfile.gettempdir()
        path = join(tempdir, 'image')
        print(path)
        with open(path, 'wb') as f:
            r.rawdecode_content = True
            shutil.copyfileobj(r.raw, f)
        result = predict(path)
        requests.post(ENDPOINT, json={
            'job_id': job['job_id'],
            'result': result
        })
    else:
        ## post without result = failed
        request.post(ENDPOINT, json={
            'job_id': job['job_id']
        })


## background thread for processing
Thread(target=worker).start()


@post('/')
def index():
    json = request.json
    log('received job:', json)
    jobQueue.put(json)


print(f'Listening on port {PORT}')
run(host=HOST, port=PORT, reloader=False)