Improve unit test coverage for NLU pipeline components

app/bot/nlu/intent_classifiers/tf_intent_classifer.py

@@ -7,9 +7,8 @@
 import spacy
 import tensorflow as tf
 from sklearn.preprocessing import LabelBinarizer
-from tensorflow.python.keras import Sequential
-from tensorflow.python.layers.core import Dense
-from tensorflow.python.layers.core import Dropout
+from tensorflow.keras import Sequential
+from tensorflow.keras.layers import Dense, Dropout
 from app.bot.nlu.pipeline import NLUComponent
 
 np.random.seed(1)

tests/__init__.py

@@ -0,0 +1,3 @@
+
+
+

tests/test_nlu_pipeline.py

@@ -0,0 +1,82 @@
+import pytest
+from unittest.mock import MagicMock, patch
+from app.bot.nlu.featurizers.spacy_featurizer import SpacyFeaturizer
+from app.bot.nlu.intent_classifiers.sklearn_intent_classifer import SklearnIntentClassifier
+from app.bot.nlu.intent_classifiers.tf_intent_classifer import TfIntentClassifier
+from app.bot.nlu.entity_extractors.crf_entity_extractor import CRFEntityExtractor
+from app.bot.nlu.entity_extractors.synonym_replacer import SynonymReplacer
+from app.bot.nlu.llm.zero_shot_nlu_openai import ZeroShotNLUOpenAI
+
+@pytest.fixture
+def sample_text():
+    return "Order a large pepperoni pizza"
+
+@pytest.fixture
+def spacy_model():
+    import spacy
+    return spacy.blank("en")
+
+@pytest.fixture
+def spacy_featurizer(spacy_model):
+    return SpacyFeaturizer(model_name="en")
+
+
+def test_spacy_featurizer_process(spacy_featurizer, sample_text):
+    message = {"text": sample_text}
+    result = spacy_featurizer.process(message)
+    assert "spacy_doc" in result
+    assert result["spacy_doc"].text == sample_text
+
+
+def test_sklearn_intent_classifier_process():
+    clf = SklearnIntentClassifier()
+    clf.model = MagicMock()
+    clf.model.classes_ = ["order_pizza", "greet"]
+    clf.model.predict_proba.return_value = [[0.8, 0.2]]
+    message = {"text": "Order pizza", "spacy_doc": MagicMock(vector=[1.0, 2.0])}
+    result = clf.process(message)
+    assert "intent" in result
+    assert "intent_ranking" in result
+
+
+def test_tf_intent_classifier_process():
+    clf = TfIntentClassifier()
+    clf.model = MagicMock()
+    clf.label_encoder = MagicMock()
+    clf.label_encoder.classes_ = ["order_pizza", "greet"]
+    clf.graph = MagicMock()
+    clf.model.predict.return_value = [[0.7, 0.3]]
+    message = {"text": "Order pizza"}
+    with patch.object(clf, "nlp", MagicMock()):
+        result = clf.process(message)
+    assert "intent" in result
+    assert "intent_ranking" in result
+
+
+def test_crf_entity_extractor_process():
+    extractor = CRFEntityExtractor()
+    extractor.tagger = MagicMock()
+    extractor.tagger.tag.return_value = ["B-pizza_size", "O", "B-pizza_topping"]
+    message = {"text": "large pepperoni", "spacy_doc": MagicMock()}
+    with patch.object(extractor, "pos_tagger", return_value=[("large", "JJ"), ("pepperoni", "NN")]):
+        with patch.object(extractor, "sent_to_features", return_value=[["feature1"], ["feature2"]]):
+            result = extractor.process(message)
+    assert "entities" in result
+
+
+def test_synonym_replacer_process():
+    synonyms = {"big": "large"}
+    replacer = SynonymReplacer(synonyms=synonyms)
+    message = {"entities": {"pizza_size": "big"}}
+    result = replacer.process(message)
+    assert result["entities"]["pizza_size"] == "large"
+
+
+def test_zero_shot_nlu_openai_process():
+    zsnlu = ZeroShotNLUOpenAI(intents=["order_pizza"], entities=["pizza_size"])
+    zsnlu.chain = MagicMock()
+    zsnlu.chain.invoke.return_value = {"intent": "order_pizza", "entities": {"pizza_size": "large"}}
+    message = {"text": "I want a large pizza"}
+    result = zsnlu.process(message)
+    assert result["intent"]["intent"] == "order_pizza"
+    assert result["entities"]["pizza_size"] == "large"