Update app.py

app.py

@@ -42,6 +42,10 @@ def get_text_value(node, field_name):
         return node[field_name]
     return None
 
+def format_vector_for_pg(vector: List[float]) -> str:
+    """벡터를 PostgreSQL 포맷으로 변환합니다."""
+    return f"[{','.join(str(x) for x in vector)}]"
+
 def search_similar_chat(query: str, max_results: int = 100) -> List[Dict]:
     """
     다중 임베딩된 채팅 데이터에서 유사한 콘텐츠를 검색합니다.
@@ -71,21 +75,24 @@ def search_similar_chat(query: str, max_results: int = 100) -> List[Dict]:
         query_embedding = normalize(raw_embedding.reshape(1, -1), norm='l2')[0]
         print(f"임베딩 정규화 전/후 첫 5개 요소: {raw_embedding[:5]} -> {query_embedding[:5]}")
         
+        # Java 방식: 벡터를 문자열로 변환
+        query_vector = format_vector_for_pg(query_embedding)
+        
         # DB 연결
         conn = get_db_conn()
         register_vector(conn)
         
-        # 여러 필드를 가중치로 조합한 유사도 검색 SQL - 매개변수화된 쿼리 사용
-        sql = """
+        # Java 방식: 문자열 포맷팅 사용한 SQL 쿼리
+        sql = f"""
             WITH embeddings AS (
                 SELECT 
                     id, 
                     metadata, 
                     content,
-                    CASE WHEN full_embedding IS NOT NULL THEN 1 - (full_embedding <=> %s::vector) ELSE 0 END * %s as full_sim,
-                    CASE WHEN topic_embedding IS NOT NULL THEN 1 - (topic_embedding <=> %s::vector) ELSE 0 END * %s as topic_sim,
-                    CASE WHEN customer_embedding IS NOT NULL THEN 1 - (customer_embedding <=> %s::vector) ELSE 0 END * %s as customer_sim,
-                    CASE WHEN agent_embedding IS NOT NULL THEN 1 - (agent_embedding <=> %s::vector) ELSE 0 END * %s as agent_sim
+                    CASE WHEN full_embedding IS NOT NULL THEN 1 - (full_embedding <=> '{query_vector}'::vector) ELSE 0 END * {full_w} as full_sim,
+                    CASE WHEN topic_embedding IS NOT NULL THEN 1 - (topic_embedding <=> '{query_vector}'::vector) ELSE 0 END * {topic_w} as topic_sim,
+                    CASE WHEN customer_embedding IS NOT NULL THEN 1 - (customer_embedding <=> '{query_vector}'::vector) ELSE 0 END * {customer_w} as customer_sim,
+                    CASE WHEN agent_embedding IS NOT NULL THEN 1 - (agent_embedding <=> '{query_vector}'::vector) ELSE 0 END * {agent_w} as agent_sim
                 FROM vector_store_multi_embeddings
                 WHERE full_embedding IS NOT NULL
                 OR topic_embedding IS NOT NULL
@@ -99,20 +106,13 @@ def search_similar_chat(query: str, max_results: int = 100) -> List[Dict]:
                 (full_sim + topic_sim + customer_sim + agent_sim) as combined_similarity
             FROM embeddings
             ORDER BY combined_similarity DESC
-            LIMIT %s
+            LIMIT {limit}
         """
         
         with conn.cursor() as cur:
-            # 매개변수화된 쿼리 실행
-            params = (
-                query_embedding, full_w, 
-                query_embedding, topic_w, 
-                query_embedding, customer_w, 
-                query_embedding, agent_w, 
-                limit
-            )
-            print(f"쿼리 실행 - 파라미터: 가중치 설정={full_w}, {topic_w}, {customer_w}, {agent_w}, 결과 제한={limit}")
-            cur.execute(sql, params)
+            print(f"쿼리 실행 - Java 방식 포맷팅, 가중치 설정={full_w}, {topic_w}, {customer_w}, {agent_w}, 결과 제한={limit}")
+            # Java 방식: 매개변수 없이 직접 쿼리 실행
+            cur.execute(sql)
             rows = cur.fetchall()
             
             print(f"검색 결과: 총 {len(rows)}개 데이터 조회됨")
@@ -226,21 +226,24 @@ def search_similar_chat_by_date(
         query_embedding = normalize(raw_embedding.reshape(1, -1), norm='l2')[0]
         print(f"날짜 검색 - 임베딩 정규화 전/후 첫 5개 요소: {raw_embedding[:5]} -> {query_embedding[:5]}")
         
+        # Java 방식: 벡터를 문자열로 변환
+        query_vector = format_vector_for_pg(query_embedding)
+        
         # DB 연결
         conn = get_db_conn()
         register_vector(conn)
         
-        # 여러 필드를 가중치로 조합한 유사도 검색 SQL - 매개변수화
-        sql = """
+        # Java 방식: 문자열 포맷팅 사용한 SQL 쿼리 시작
+        sql = f"""
             WITH embeddings AS (
                 SELECT 
                     id, 
                     metadata, 
                     content,
-                    CASE WHEN full_embedding IS NOT NULL THEN 1 - (full_embedding <=> %s::vector) ELSE 0 END * %s as full_sim,
-                    CASE WHEN topic_embedding IS NOT NULL THEN 1 - (topic_embedding <=> %s::vector) ELSE 0 END * %s as topic_sim,
-                    CASE WHEN customer_embedding IS NOT NULL THEN 1 - (customer_embedding <=> %s::vector) ELSE 0 END * %s as customer_sim,
-                    CASE WHEN agent_embedding IS NOT NULL THEN 1 - (agent_embedding <=> %s::vector) ELSE 0 END * %s as agent_sim
+                    CASE WHEN full_embedding IS NOT NULL THEN 1 - (full_embedding <=> '{query_vector}'::vector) ELSE 0 END * {full_w} as full_sim,
+                    CASE WHEN topic_embedding IS NOT NULL THEN 1 - (topic_embedding <=> '{query_vector}'::vector) ELSE 0 END * {topic_w} as topic_sim,
+                    CASE WHEN customer_embedding IS NOT NULL THEN 1 - (customer_embedding <=> '{query_vector}'::vector) ELSE 0 END * {customer_w} as customer_sim,
+                    CASE WHEN agent_embedding IS NOT NULL THEN 1 - (agent_embedding <=> '{query_vector}'::vector) ELSE 0 END * {agent_w} as agent_sim
                 FROM vector_store_multi_embeddings
                 WHERE full_embedding IS NOT NULL
                 OR topic_embedding IS NOT NULL
@@ -248,21 +251,12 @@ def search_similar_chat_by_date(
                 OR agent_embedding IS NOT NULL
         """
         
-        params = [
-            query_embedding, full_w, 
-            query_embedding, topic_w, 
-            query_embedding, customer_w, 
-            query_embedding, agent_w
-        ]
-        
         # 날짜 필터 추가
         if start_timestamp is not None:
-            sql += " AND (metadata->>'startTime')::bigint >= %s"
-            params.append(start_timestamp)
+            sql += f" AND (metadata->>'startTime')::bigint >= {start_timestamp}"
         
         if end_timestamp is not None:
-            sql += " AND (metadata->>'startTime')::bigint <= %s"
-            params.append(end_timestamp)
+            sql += f" AND (metadata->>'startTime')::bigint <= {end_timestamp}"
         
         sql += """
             )
@@ -276,11 +270,10 @@ def search_similar_chat_by_date(
             LIMIT %s
         """
         
-        params.append(limit)
-        
         with conn.cursor() as cur:
             print(f"날짜 검색 쿼리 실행: 시작일={start_date}({start_timestamp}), 종료일={end_date}({end_timestamp})")
-            cur.execute(sql, tuple(params))
+            # 여기서는 limit만 매개변수로 전달
+            cur.execute(sql, (limit,))
             rows = cur.fetchall()
             
             print(f"날짜 필터링 검색 결과: 총 {len(rows)}개 데이터 조회됨")