Upload backend/hue_portal/core/tests/test_pure_semantic_search.py with huggingface_hub

backend/hue_portal/core/tests/test_pure_semantic_search.py

@@ -0,0 +1,156 @@
+"""
+Unit tests for Pure Semantic Search.
+"""
+import unittest
+from unittest.mock import Mock, patch, MagicMock
+from django.test import TestCase
+from django.db.models import QuerySet
+from hue_portal.core.pure_semantic_search import (
+    get_vector_scores,
+    parallel_vector_search,
+    pure_semantic_search,
+    calculate_exact_match_boost
+)
+
+
+class TestPureSemanticSearch(unittest.TestCase):
+    """Test Pure Semantic Search functions."""
+    
+    def setUp(self):
+        """Set up test fixtures."""
+        self.mock_queryset = Mock(spec=QuerySet)
+        self.mock_queryset.__iter__ = Mock(return_value=iter([]))
+        self.mock_queryset.__len__ = Mock(return_value=0)
+    
+    @patch('hue_portal.core.pure_semantic_search.get_embedding_model')
+    @patch('hue_portal.core.pure_semantic_search.generate_embedding')
+    @patch('hue_portal.core.pure_semantic_search.load_embedding')
+    @patch('hue_portal.core.pure_semantic_search.cosine_similarity')
+    def test_get_vector_scores(self, mock_cosine, mock_load, mock_gen, mock_model):
+        """Test get_vector_scores function."""
+        # Mock embedding model
+        mock_model.return_value = Mock()
+        mock_gen.return_value = [0.1] * 1024  # BGE-M3 dimension
+        mock_cosine.return_value = 0.8
+        
+        # Mock objects with embeddings
+        obj1 = Mock()
+        obj2 = Mock()
+        mock_load.side_effect = [[0.1] * 1024, [0.1] * 1024]
+        
+        self.mock_queryset.__iter__ = Mock(return_value=iter([obj1, obj2]))
+        self.mock_queryset.__len__ = Mock(return_value=2)
+        
+        results = get_vector_scores(self.mock_queryset, "test query", top_k=10)
+        
+        self.assertIsInstance(results, list)
+        # Should return results with scores
+        if results:
+            self.assertIsInstance(results[0], tuple)
+            self.assertEqual(len(results[0]), 2)
+    
+    def test_calculate_exact_match_boost(self):
+        """Test exact match boost calculation."""
+        obj = Mock()
+        obj.title = "Quy định điều 12"
+        obj.name = "Điều 12"
+        
+        # Test phrase match
+        boost = calculate_exact_match_boost(obj, "điều 12", ["title", "name"])
+        self.assertGreater(boost, 0.0)
+        self.assertLessEqual(boost, 1.0)
+        
+        # Test no match
+        boost2 = calculate_exact_match_boost(obj, "điều 99", ["title", "name"])
+        self.assertLess(boost2, boost)
+    
+    @patch('hue_portal.core.pure_semantic_search.get_vector_scores')
+    def test_parallel_vector_search_single_query(self, mock_get_scores):
+        """Test parallel_vector_search with single query."""
+        obj1 = Mock()
+        obj2 = Mock()
+        mock_get_scores.return_value = [(obj1, 0.9), (obj2, 0.8)]
+        
+        self.mock_queryset.__iter__ = Mock(return_value=iter([obj1, obj2]))
+        
+        results = parallel_vector_search(
+            ["test query"],
+            self.mock_queryset,
+            top_k_per_query=5,
+            final_top_k=2
+        )
+        
+        self.assertIsInstance(results, list)
+        # Should use single query search path
+    
+    @patch('hue_portal.core.pure_semantic_search.get_vector_scores')
+    def test_parallel_vector_search_multiple_queries(self, mock_get_scores):
+        """Test parallel_vector_search with multiple queries."""
+        obj1 = Mock()
+        obj2 = Mock()
+        obj3 = Mock()
+        
+        # Different results for different queries
+        mock_get_scores.side_effect = [
+            [(obj1, 0.9), (obj2, 0.8)],  # Query 1
+            [(obj2, 0.85), (obj3, 0.75)],  # Query 2
+        ]
+        
+        self.mock_queryset.__iter__ = Mock(return_value=iter([obj1, obj2, obj3]))
+        
+        results = parallel_vector_search(
+            ["query 1", "query 2"],
+            self.mock_queryset,
+            top_k_per_query=5,
+            final_top_k=3
+        )
+        
+        self.assertIsInstance(results, list)
+        # Should merge results from multiple queries
+        # obj2 should appear with max score (0.85)
+    
+    @patch('hue_portal.core.pure_semantic_search.parallel_vector_search')
+    def test_pure_semantic_search_single(self, mock_parallel):
+        """Test pure_semantic_search with single query."""
+        obj1 = Mock()
+        obj2 = Mock()
+        mock_parallel.return_value = [(obj1, 0.9), (obj2, 0.8)]
+        
+        results = pure_semantic_search(
+            ["test query"],
+            self.mock_queryset,
+            top_k=2
+        )
+        
+        self.assertIsInstance(results, list)
+        # Should return objects only (without scores)
+        self.assertEqual(len(results), 2)
+        self.assertEqual(results[0], obj1)
+        self.assertEqual(results[1], obj2)
+    
+    @patch('hue_portal.core.pure_semantic_search.parallel_vector_search')
+    def test_pure_semantic_search_multiple(self, mock_parallel):
+        """Test pure_semantic_search with multiple queries."""
+        obj1 = Mock()
+        obj2 = Mock()
+        mock_parallel.return_value = [(obj1, 0.9), (obj2, 0.8)]
+        
+        results = pure_semantic_search(
+            ["query 1", "query 2", "query 3"],
+            self.mock_queryset,
+            top_k=2
+        )
+        
+        self.assertIsInstance(results, list)
+        # Should use parallel_vector_search
+        mock_parallel.assert_called_once()
+    
+    def test_pure_semantic_search_empty_queries(self):
+        """Test pure_semantic_search with empty queries."""
+        results = pure_semantic_search([], self.mock_queryset, top_k=10)
+        self.assertEqual(results, [])
+
+
+if __name__ == "__main__":
+    unittest.main()
+