Validate output format and edge cases

src/poprox_recommender/request_generator.py

@@ -0,0 +1,83 @@
+import random
+from datetime import datetime, timedelta
+from typing import List
+from uuid import uuid4
+
+from pydantic import ValidationError
+
+from poprox_concepts import AccountInterest, Click, InterestProfile
+from poprox_concepts.api.recommendations import RecommendationRequest
+from poprox_recommender.data.mind import MindData
+
+
+class RequestGenerator:
+    """
+    Class to generate recommendation request using click history, onboarding topics, and candidate articles from MIND
+    """
+
+    def __init__(self, mind_data: MindData):
+        self.mind_data = mind_data
+        self.candidate_articles = list()
+        self.past_articles = list()
+        self.added_topics = list()
+        self.clicks = list()
+
+    def set_num_recs(self, num_recs: int):
+        self.num_recs = num_recs
+
+    def add_clicks(self, num_clicks: int, num_days: int | None = None):
+        all_articles = list(self.mind_data.news_df.index)
+
+        if num_days:
+            start_date = datetime.now() - timedelta(days=num_days - 1)
+            timestamps = [start_date + timedelta(days=random.randint(0, num_days - 1)) for _ in range(num_clicks)]
+            random.shuffle(timestamps)
+        else:
+            timestamps = [datetime.now()] * num_clicks
+        # generate click history
+        self.clicks = [
+            Click(
+                article_id=self.mind_data.news_uuid_for_id(random.choice(all_articles)),
+                newsletter_id=uuid4(),
+                timestamp=timestamps[i],
+            )
+            for i in range(num_clicks)
+        ]
+
+        self.past_articles = [self.mind_data.lookup_article(uuid=click.article_id) for click in self.clicks]
+
+    def add_topics(self, topics: List[str]):
+        self.added_topics = [
+            AccountInterest(
+                account_id=uuid4(),
+                entity_id=uuid4(),
+                entity_name=topic,
+                preference=random.randint(1, 5),
+                frequency=None,
+            )
+            for topic in topics
+        ]
+
+    def add_candidates(self, num_candidates):
+        all_articles = list(self.mind_data.news_df.index)
+        selected_candidates = random.sample(all_articles, num_candidates)
+
+        self.candidate_articles = [self.mind_data.lookup_article(id=article_id) for article_id in selected_candidates]
+
+    def get_request(self) -> RecommendationRequest:
+        interest_profile = InterestProfile(
+            profile_id=uuid4(),
+            click_history=self.clicks,
+            onboarding_topics=self.added_topics,
+        )
+
+        try:
+            request = RecommendationRequest(
+                past_articles=self.past_articles,
+                todays_articles=self.candidate_articles,
+                interest_profile=interest_profile,
+                num_recs=self.num_recs,
+            )
+            return request
+        except ValidationError as e:
+            raise ValueError(f"Generated request is invalid: {e}")

src/poprox_recommender/testing.py

@@ -18,6 +18,7 @@
 from pytest import fixture
 
 from poprox_concepts.api.recommendations import RecommendationRequest, RecommendationResponse
+from poprox_recommender.data.mind import MindData
 
 logger = logging.getLogger(__name__)
 
@@ -128,3 +129,8 @@ def auto_service() -> Generator[TestService, None, None]:
         # use already-running docker
         port = os.environ.get("POPROX_TEST_PORT", "9000")
         yield DockerTestService(f"http://localhost:{port}/2015-03-31/functions/function/invocations")
+
+
+@fixture(scope="session")
+def mind_data():
+    yield MindData

tests/web_service/test_basic_request.py

@@ -0,0 +1,59 @@
+"""
+Test basic request by initializing test data.
+"""
+
+import logging
+import warnings
+
+from pydantic import ValidationError
+from pytest import mark, skip
+
+from poprox_concepts.api.recommendations import RecommendationRequest, RecommendationResponse
+from poprox_recommender.config import allow_data_test_failures
+from poprox_recommender.recommenders import recommendation_pipelines
+from poprox_recommender.request_generator import RequestGenerator
+from poprox_recommender.testing import auto_service as service
+from poprox_recommender.testing import mind_data
+
+logger = logging.getLogger(__name__)
+try:
+    PIPELINES = recommendation_pipelines().keys()
+except Exception as e:
+    warnings.warn("failed to load models, did you run `dvc pull`?")
+    if allow_data_test_failures():
+        skip("recommendation pipelines unavailable", allow_module_level=True)
+    else:
+        raise e
+
+
+@mark.docker
+@mark.parametrize("pipeline", PIPELINES)
+def test_basic_request(service, mind_data, pipeline):  # noqa: F811
+    """
+    Initialize request data
+    """
+    request_generator = RequestGenerator(mind_data())
+    request_generator.add_candidates(100)
+    request_generator.add_clicks(num_clicks=37, num_days=7)
+    request_generator.add_topics(
+        [
+            "Science",
+            "Technology",
+            "Sports",
+            "Lifestyle",
+            "Oddities",
+        ]
+    )
+    request_generator.set_num_recs(10)
+    req_body = request_generator.get_request()
+
+    logger.info("sending request")
+    response = service.request(req_body, pipeline)
+    logger.info("response: %s", response.model_dump_json(indent=2))
+    assert response.recommendations
+    assert response.recommendations.values()
+    recs = next(iter(response.recommendations.values()))
+    assert len(recs) > 0
+    assert len(recs) == request_generator.num_recs
+    article_ids = [article.article_id for article in recs]
+    assert len(article_ids) == len(set(article_ids))

tests/web_service/test_heavy_interaction_history.py

@@ -0,0 +1,68 @@
+"""
+Validate edge case: a user with long history of clicks and onboarding topics
+"""
+
+import logging
+import warnings
+
+from pydantic import ValidationError
+from pytest import mark, skip
+
+from poprox_concepts.api.recommendations import RecommendationRequest, RecommendationResponse
+from poprox_recommender.config import allow_data_test_failures
+from poprox_recommender.recommenders import recommendation_pipelines
+from poprox_recommender.request_generator import RequestGenerator
+from poprox_recommender.testing import auto_service as service
+from poprox_recommender.testing import mind_data
+
+logger = logging.getLogger(__name__)
+try:
+    PIPELINES = recommendation_pipelines().keys()
+except Exception as e:
+    warnings.warn("failed to load models, did you run `dvc pull`?")
+    if allow_data_test_failures():
+        skip("recommendation pipelines unavailable", allow_module_level=True)
+    else:
+        raise e
+
+
+@mark.docker
+@mark.parametrize("pipeline", PIPELINES)
+def test_heavy_interaction_history(service, mind_data, pipeline):  # noqa: F811
+    """
+    Initialize request data
+    """
+    request_generator = RequestGenerator(mind_data())
+    request_generator.add_candidates(100)
+    request_generator.add_clicks(num_clicks=100, num_days=10)
+    request_generator.add_topics(
+        [
+            "U.S. news",
+            "World news",
+            "Politics",
+            "Business",
+            "Entertainment",
+            "Sports",
+            "Health",
+            "Science",
+            "Technology",
+            "Lifestyle",
+            "Religion",
+            "Climate and environment",
+            "Education",
+            "Oddities",
+        ]
+    )
+    request_generator.set_num_recs(10)
+    req_body = request_generator.get_request()
+
+    logger.info("sending request")
+    response = service.request(req_body, pipeline)
+    logger.info("response: %s", response.model_dump_json(indent=2))
+    assert response.recommendations
+    assert response.recommendations.values()
+    recs = next(iter(response.recommendations.values()))
+    assert len(recs) > 0
+    assert len(recs) == request_generator.num_recs
+    article_ids = [article.article_id for article in recs]
+    assert len(article_ids) == len(set(article_ids))

tests/web_service/test_no_clicks.py

@@ -0,0 +1,59 @@
+"""
+Validate edge case: user has no click history
+"""
+
+import logging
+import warnings
+
+from pydantic import ValidationError
+from pytest import mark, skip
+
+from poprox_concepts.api.recommendations import RecommendationRequest, RecommendationResponse
+from poprox_recommender.config import allow_data_test_failures
+from poprox_recommender.recommenders import recommendation_pipelines
+from poprox_recommender.request_generator import RequestGenerator
+from poprox_recommender.testing import auto_service as service
+from poprox_recommender.testing import mind_data
+
+logger = logging.getLogger(__name__)
+try:
+    PIPELINES = recommendation_pipelines().keys()
+except Exception as e:
+    warnings.warn("failed to load models, did you run `dvc pull`?")
+    if allow_data_test_failures():
+        skip("recommendation pipelines unavailable", allow_module_level=True)
+    else:
+        raise e
+
+
+@mark.docker
+@mark.parametrize("pipeline", PIPELINES)
+def test_no_clicks(service, mind_data, pipeline):  # noqa: F811
+    """
+    Initialize request data
+    """
+    request_generator = RequestGenerator(mind_data())
+    request_generator.add_candidates(100)
+    request_generator.add_clicks(num_clicks=0, num_days=10)
+    request_generator.add_topics(
+        [
+            "Science",
+            "Technology",
+            "Sports",
+            "Lifestyle",
+            "Oddities",
+        ]
+    )
+    request_generator.set_num_recs(10)
+    req_body = request_generator.get_request()
+
+    logger.info("sending request")
+    response = service.request(req_body, pipeline)
+    logger.info("response: %s", response.model_dump_json(indent=2))
+    assert response.recommendations
+    assert response.recommendations.values()
+    recs = next(iter(response.recommendations.values()))
+    assert len(recs) > 0
+    assert len(recs) == request_generator.num_recs
+    article_ids = [article.article_id for article in recs]
+    assert len(article_ids) == len(set(article_ids))

tests/web_service/test_no_clicks_no_onboarding.py

@@ -0,0 +1,51 @@
+"""
+Validate edge case: user has no click history & no onboarding preferences
+"""
+
+import logging
+import warnings
+
+from pydantic import ValidationError
+from pytest import mark, skip
+
+from poprox_concepts.api.recommendations import RecommendationRequest, RecommendationResponse
+from poprox_recommender.config import allow_data_test_failures
+from poprox_recommender.recommenders import recommendation_pipelines
+from poprox_recommender.request_generator import RequestGenerator
+from poprox_recommender.testing import auto_service as service
+from poprox_recommender.testing import mind_data
+
+logger = logging.getLogger(__name__)
+try:
+    PIPELINES = recommendation_pipelines().keys()
+except Exception as e:
+    warnings.warn("failed to load models, did you run `dvc pull`?")
+    if allow_data_test_failures():
+        skip("recommendation pipelines unavailable", allow_module_level=True)
+    else:
+        raise e
+
+
+@mark.docker
+@mark.parametrize("pipeline", PIPELINES)
+def test_no_clicks_no_onboarding(service, mind_data, pipeline):  # noqa: F811
+    """
+    Initialize request data
+    """
+    request_generator = RequestGenerator(mind_data())
+    request_generator.add_candidates(100)
+    request_generator.add_clicks(num_clicks=0, num_days=0)
+    request_generator.add_topics([])
+    request_generator.set_num_recs(10)
+    req_body = request_generator.get_request()
+
+    logger.info("sending request")
+    response = service.request(req_body, pipeline)
+    logger.info("response: %s", response.model_dump_json(indent=2))
+    assert response.recommendations
+    assert response.recommendations.values()
+    recs = next(iter(response.recommendations.values()))
+    assert len(recs) > 0
+    assert len(recs) == request_generator.num_recs
+    article_ids = [article.article_id for article in recs]
+    assert len(article_ids) == len(set(article_ids))

tests/web_service/test_no_onboarding.py

@@ -0,0 +1,51 @@
+"""
+Validate edge case: user has no on boarding topic
+"""
+
+import logging
+import warnings
+
+from pydantic import ValidationError
+from pytest import mark, skip
+
+from poprox_concepts.api.recommendations import RecommendationRequest, RecommendationResponse
+from poprox_recommender.config import allow_data_test_failures
+from poprox_recommender.recommenders import recommendation_pipelines
+from poprox_recommender.request_generator import RequestGenerator
+from poprox_recommender.testing import auto_service as service
+from poprox_recommender.testing import mind_data
+
+logger = logging.getLogger(__name__)
+try:
+    PIPELINES = recommendation_pipelines().keys()
+except Exception as e:
+    warnings.warn("failed to load models, did you run `dvc pull`?")
+    if allow_data_test_failures():
+        skip("recommendation pipelines unavailable", allow_module_level=True)
+    else:
+        raise e
+
+
+@mark.docker
+@mark.parametrize("pipeline", PIPELINES)
+def test_no_onboarding(service, mind_data, pipeline):  # noqa: F811
+    """
+    Initialize request data
+    """
+    request_generator = RequestGenerator(mind_data())
+    request_generator.add_candidates(100)
+    request_generator.add_clicks(num_clicks=37, num_days=7)
+    request_generator.add_topics([])
+    request_generator.set_num_recs(10)
+    req_body = request_generator.get_request()
+
+    logger.info("sending request")
+    response = service.request(req_body, pipeline)
+    logger.info("response: %s", response.model_dump_json(indent=2))
+    assert response.recommendations
+    assert response.recommendations.values()
+    recs = next(iter(response.recommendations.values()))
+    assert len(recs) > 0
+    assert len(recs) == request_generator.num_recs
+    article_ids = [article.article_id for article in recs]
+    assert len(article_ids) == len(set(article_ids))