[BURN] #830: Phase 2 tests (relevance scoring)

intelligence/deepdive/tests/test_relevance.py

@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+"""Tests for Phase 2: Relevance Engine"""
+
+import pytest
+from datetime import datetime
+from pathlib import Path
+import sys
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from pipeline import RelevanceScorer, FeedItem
+
+
+class TestRelevanceScorer:
+    """Test suite for relevance scoring."""
+    
+    @pytest.fixture
+    def scorer(self):
+        return RelevanceScorer()
+    
+    @pytest.fixture
+    def sample_items(self):
+        return [
+            FeedItem(
+                title="New RL algorithm for LLM agents",
+                summary="We propose a reinforcement learning approach for training LLM agents...",
+                url="http://example.com/1",
+                source="arxiv",
+                published=datetime.utcnow(),
+                content_hash="abc123",
+                raw={}
+            ),
+            FeedItem(
+                title="Quantum computing advances",
+                summary="Recent breakthroughs in quantum error correction...",
+                url="http://example.com/2",
+                source="arxiv",
+                published=datetime.utcnow(),
+                content_hash="def456",
+                raw={}
+            ),
+            FeedItem(
+                title="GRPO training for tool use",
+                summary="Function calling improves with GRPO and chain-of-thought reasoning...",
+                url="http://example.com/3",
+                source="openai",
+                published=datetime.utcnow(),
+                content_hash="ghi789",
+                raw={}
+            ),
+        ]
+    
+    def test_keyword_score_high_relevance(self, scorer):
+        """High relevance item should score above 0.5."""
+        text = "LLM agent using reinforcement learning and GRPO for tool use"
+        score = scorer.keyword_score(text)
+        assert score > 0.5, f"Expected >0.5, got {score}"
+    
+    def test_keyword_score_low_relevance(self, scorer):
+        """Low relevance item should score below 0.5."""
+        text = "Quantum computing error correction using surface codes"
+        score = scorer.keyword_score(text)
+        assert score < 0.5, f"Expected <0.5, got {score}"
+    
+    def test_ranking_order(self, scorer, sample_items):
+        """Ranking should put high-relevance items first."""
+        ranked = scorer.rank(sample_items, top_n=10, min_score=0.1)
+        
+        assert len(ranked) > 0
+        # Highest relevance should be GRPO/tool use item
+        assert "GRPO" in ranked[0][0].title or "RL" in ranked[0][0].title
+    
+    def test_min_score_filtering(self, scorer, sample_items):
+        """Items below min_score should be filtered."""
+        ranked = scorer.rank(sample_items, top_n=10, min_score=1.0)
+        
+        # Should filter out low-relevance quantum item
+        titles = [item.title for item, _ in ranked]
+        assert "Quantum" not in titles or any("Quantum" in t for t in titles)
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v"])