have working client example

.dockerignore

@@ -0,0 +1,6 @@
+Dockerfile
+README.md
+*.pyc
+*.pyo
+*.pyd
+__pycache__

Dockerfile

@@ -6,21 +6,22 @@ FROM python:3.8.1-slim
 ENV APP_HOME /app
 WORKDIR $APP_HOME
 
-#workaround: https://github.com/moby/moby/issues/28617
-ADD .env /app
-RUN cat .env >> /etc/environment
+# https://github.com/moby/moby/issues/28617
+#*COPY_ENV_VARS_HERE*
+ENV GCS_BUCKET_NAME shakti0
+ENV GOOGLE_APPLICATION_CREDENTIALS gcs_creds.json
+ENV PROJECT_ID shakti123
 
-##install required files
 
-#must run deploy command from folder containing server code
+# must run deploy command from folder containing server code
 COPY . ./
 
 # Install production dependencies.
-RUN pip install Flask gunicorn
+RUN pip install Flask gunicorn joblib python-dotenv numpy scikit-learn
 
 # Run the web service on container startup. Here we use the gunicorn
 # webserver, with one worker process and 2 threads.
 # For environments with multiple CPU cores, increase the number of workers
 # to be equal to the cores available.
-#to make app:app true, flask server variable name (left side of :) must be app
+# to make app:app true, flask server variable name (left side of :) must be app
 CMD exec gunicorn --bind :$PORT --workers 1 --threads 2 app:app

app.py

@@ -0,0 +1,39 @@
+import os
+
+from flask import Flask, jsonify, request
+from joblib import load
+from dotenv import load_dotenv
+import sklearn
+import numpy as np
+import glob
+
+app = Flask(__name__)
+model = None
+
+
+@app.before_request
+def load_resources():
+    '''Flask template is currently only for scikit-learn models'''
+    load_dotenv()
+    # in future don't make user have to write flask server themselves
+    # that way don't need to use global b/c not thread-safe & using gunicorn
+    global model
+    if not model:
+        model = load(glob.glob("*joblib")[0])
+
+
+def transform_data(input_data):
+    # will convert from 1D to required 2D
+    return np.array(input_data.tolist()[:784]).reshape(1, -1)
+
+
+@app.route("/", methods=["GET", "POST"])
+def predict():
+    img_nparray = np.fromstring(request.files["image"].read(), np.uint8)
+    transformed_data = transform_data(img_nparray)
+    prediction = model.predict(transformed_data).tolist()[0]
+    return jsonify({"prediction": prediction})
+
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=int(os.environ.get("PORT", 8080)))