def lambda_handler(event, context):
"""
This is the main entry point for performing an action. The action evaluator puts
an action message on the actions queue and here's where they're popped
off and dealt with.
"""
records_table = dynamodb.Table(DYNAMODB_TABLE)
lambda_init_once()
for sqs_record in event["Records"]:
# TODO research max # sqs records / lambda_handler invocation
action_message = ActionMessage.from_aws_json(sqs_record["body"])
logger.info("Performing action: action_message = %s", action_message)
if action_performer := ActionPerformer.get(
action_message.action_label.value):
action_performer.perform_action(action_message)
ActionEvent(
content_id=action_message.content_key,
performed_at=datetime.datetime.now(),
action_label=action_message.action_label.value,
# v0 Hacks: the label rules model is super mutable we store basically the whole state
# for each action performed. (~gross but until we have a unique id and version
# it's what we've got).
# Right now this just make json blob for action_performer and a list of
# json blobs for action_rules that we can store and recover if needed.
action_performer=action_performer.to_aws_json(),
action_rules=[
rule.to_aws_json() for rule in action_message.action_rules
],
).write_to_table(records_table)
def action_history() -> ActionHistoryResponse:
"""
Return list of action event records for a given ID.
"""
if content_id := bottle.request.query.content_id or None:
return ActionHistoryResponse(
ActionEvent.get_from_content_id(dynamodb_table,
f"{content_id}"))
def test_query_action_event(self):
"""
Test ActionEvent write table with get_from_content_id query
"""
event = self.get_example_action_event()
event.write_to_table(self.get_table())
query_event = ActionEvent.get_from_content_id(
self.get_table(), TestContentModels.TEST_CONTENT_ID)[0]
assert event == query_event
def get_example_action_event():
enqueue_mini_castle_for_review_action_label = ActionLabel(
TestContentModels.TEST_ACTION_LABEL)
action_rules = [
ActionRule(
name="Enqueue Mini-Castle for Review",
action_label=enqueue_mini_castle_for_review_action_label,
must_have_labels=set([
ClassificationLabel("true_positive"),
]),
must_not_have_labels=set(),
),
]
banked_signal = BankedSignal(
banked_content_id="4169895076385542",
bank_id="303636684709969",
bank_source="te",
)
banked_signal.add_classification("true_positive")
action_performer = WebhookPostActionPerformer(
name="EnqueueForReview",
url="https://webhook.site/ff7ebc37-514a-439e-9a03-46f86989e195",
headers='{"Connection":"keep-alive"}',
# monitoring page:
# https://webhook.site/#!/ff7ebc37-514a-439e-9a03-46f86989e195
)
action_message = ActionMessage(
content_key=TestContentModels.TEST_CONTENT_ID,
content_hash=
"361da9e6cf1b72f5cea0344e5bb6e70939f4c70328ace762529cac704297354a",
matching_banked_signals=[banked_signal],
action_label=enqueue_mini_castle_for_review_action_label,
action_rules=action_rules,
)
return ActionEvent(
content_id=action_message.content_key,
performed_at=TestContentModels.TEST_TIME,
action_label=action_message.action_label.value,
action_performer=action_performer.to_aws_json(),
action_rules=[
rule.to_aws_json() for rule in action_message.action_rules
],
) # .write_to_table(table)
def pipeline_progress() -> ContentPipelineProgress:
"""
WARNING: UNOPTIMIZED. DO NOT CALL FROM AUTOMATED SYSTEMS.
Build a history of the stages that this piece of content has gone
through and what their results were. Do not call this from anything but
a UI. This is not optimized for performance.
"""
content_id = bottle.request.query.content_id or None
if not content_id:
return bottle.abort(400, "content_id must be provided.")
content_id = t.cast(str, content_id)
content_object = ContentObject.get_from_content_id(
dynamodb_table, content_id)
if not content_object:
return bottle.abort(400,
f"Content with id '{content_id}' not found.")
content_object = t.cast(ContentObject, content_object)
preview_url = get_preview_url(content_id, content_object)
# The result object will be gradually built up as records are retrieved.
result = ContentPipelineProgress(
content_id=content_id,
content_type=content_object.content_type,
content_preview_url=preview_url,
submitted_at=content_object.updated_at,
submission_additional_fields=list(
content_object.additional_fields),
)
hash_records = PipelineHashRecord.get_from_content_id(
dynamodb_table, content_id)
if len(hash_records) != 0:
result.hashed_at = max(hash_records,
key=lambda r: r.updated_at).updated_at
for hash_record in hash_records:
# Assume that each signal type has a single hash
if hash_record.signal_type.get_name() in result.hash_results:
return bottle.abort(
500,
f"Content with id '{content_id}' has multiple hash records for signal-type: '{hash_record.signal_type.get_name()}'.",
)
result.hash_results[hash_record.signal_type.get_name(
)] = hash_record.content_hash
match_records = MatchRecord.get_from_content_id(
dynamodb_table, content_id)
if len(match_records) != 0:
result.matched_at = max(match_records,
key=lambda r: r.updated_at).updated_at
# TODO #751 Until we resolve type agnostic storage of signal data,
# we can't populate match details.
# actually populate result.match_results.
# TODO: ActionEvaluation does not yet leave a trail. Either record
# action evaluation or remove the evaluation stage from the
# pipeline-progress indicator.
action_records = ActionEvent.get_from_content_id(
dynamodb_table, content_id)
if len(action_records) != 0:
result.action_performed_at = max(
action_records, key=lambda r: r.performed_at).performed_at
result.action_perform_results = [
r.action_label for r in action_records
]
return result