def test_feedback_inserted(sample_feedback):
"""Test inserting events into the fact_store."""
fact_store = FactStore(True)
fact_store.session.query(FeedbackModel).delete()
for feedback in sample_feedback:
fact_store.write_feedback(**feedback)
items = fact_store.readall_feedback()
assert len(items) is 3
def test_feedback_inserted():
"""Test inserting events into the fact_store."""
CUSTOMER_ID = "#123456"
fact_store = FactStore(True)
fact_store.session.query(FeedbackModel).delete()
fact_store.write_feedback(
predict_id="a2b35c5b-016d-4e2c-8ec5-87d1b962b2f8", message="222JSJSJJS",
anomaly_status=True, customer_id=CUSTOMER_ID
)
fact_store.write_feedback(predict_id="18bd090d-ae27-4b19-a0db-ed5f589b4e2e",
customer_id=CUSTOMER_ID, message="SSJJSJS", anomaly_status=True)
fact_store.write_feedback(predict_id="74a6b1bd-efea-4e7b-87a9-8f7330885160", message="AJJSJS",
customer_id=CUSTOMER_ID, anomaly_status=False)
items = fact_store.readall_feedback()
assert len(items) is 3
def feedback():
"""Feedback Service to provide user input on which false predictions this model provided."""
try:
content = request.json
# When deploying fact_store per customer you should set env var
customer_id = os.getenv("CUSTOMER_ID")
logging.info("id: {} ".format(content["lad_id"]))
logging.info("anomaly: {} ".format(content["is_anomaly"]))
logging.info("message: {} ".format(content["message"]))
logging.info("customer_id: {} ".format(customer_id))
if not content["lad_id"] or not content["is_anomaly"]:
raise Exception("This service requires that you provide the id" +
", anomaly=True|False and notes are optional ")
fs = FactStore()
# Record feedback
HUMAN_FEEDBACK_COUNT.labels(
customer_id=customer_id,
anomaly_status=content["is_anomaly"]).inc()
# Note id is the prediction id that is found in the email.
if (fs.write_feedback(predict_id=content["lad_id"],
message=content["message"],
anomaly_status=bool(content["is_anomaly"]),
customer_id=customer_id) is False):
raise Exception("Predict ID must be unique. This anomaly" +
" feedback has been reported before")
except Exception as e:
logging.info(e)
HUMAN_FEEDBACK_ERROR_COUNT.labels(
err_msg="failure to write to db in factstore").inc()
result = {"feedback_service": "failure", "error_msg": "{0}".format(e)}
return make_response(jsonify(result), 403)
else:
result = {"feedback_service": "success"}
return make_response(jsonify(result), 200)
return ""
def false_positive():
"""Service to provide list of false anomalies to be relabeled during ml training run."""
fs = FactStore()
df = fs.readall_false_positive()
return jsonify({"feedback": df})