Update app.py

app.py

@@ -157,8 +157,19 @@ def search():
         return jsonify({"categorized_articles": {}, "has_articles": False, "loading": False})
     
     try:
-        results = vector_db.similarity_search_with_relevance_scores(query, k=50, score_threshold=0.5)
-        enriched_articles = format_articles_from_db_results(results)
+        # 1. Use similarity_search_with_score to get the raw distance score.
+        # This returns a list of (Document, float) tuples.
+        results_with_scores = vector_db.similarity_search_with_score(query, k=50)
+
+        # 2. Filter based on the raw L2 distance score. Lower is better.
+        # A threshold of 1.0 is a good starting point. You can make it smaller (e.g., 0.8)
+        # for stricter matches, or larger for looser matches.
+        score_threshold = 1.0
+        filtered_results = [(doc, score) for doc, score in results_with_scores if score < score_threshold]
+        
+        # 3. Pass the correctly filtered list to the formatting function.
+        # This function is already set up to handle this data structure.
+        enriched_articles = format_articles_from_db_results(filtered_results)
 
         categorized_articles = {}
         for article in enriched_articles:
@@ -174,6 +185,7 @@ def search():
         logger.error(f"Semantic search error: {e}", exc_info=True)
         return jsonify({"categorized_articles": {}, "has_articles": False, "loading": False}), 500
 
+
 @app.route('/get_all_articles/<category>')
 def get_all_articles(category):
     try: