[NOTICKET] fix-revert string change

src/rag/retrievers/schema.py

@@ -52,11 +52,11 @@ class SchemaRetriever(BaseRetriever):
         emb_str = "[" + ",".join(str(x) for x in embedding) + "]"
 
         if operator == "<#>":
-            score_sql = "(lpe.embedding <#> :emb::vector) * -1"
+            score_sql = f"(lpe.embedding <#> '{emb_str}'::vector) * -1"
         elif operator == "<->":
-            score_sql = "1.0 / (1.0 + (lpe.embedding <-> :emb::vector))"
+            score_sql = f"1.0 / (1.0 + (lpe.embedding <-> '{emb_str}'::vector))"
         else:
-            score_sql = "1.0 - (lpe.embedding <=> :emb::vector)"
+            score_sql = f"1.0 - (lpe.embedding <=> '{emb_str}'::vector)"
 
         sql = text(f"""
             SELECT lpe.document, lpe.cmetadata, {score_sql} AS score
@@ -65,12 +65,12 @@ class SchemaRetriever(BaseRetriever):
             WHERE lpc.name = 'document_embeddings'
               AND lpe.cmetadata->>'user_id' = :user_id
               AND lpe.cmetadata->>'source_type' = 'database'
-            ORDER BY lpe.embedding {operator} :emb::vector ASC
+            ORDER BY lpe.embedding {operator} '{emb_str}'::vector ASC
             LIMIT :k
         """)
 
         async with _pgvector_engine.connect() as conn:
-            result = await conn.execute(sql, {"user_id": user_id, "k": k * 4, "emb": emb_str})
+            result = await conn.execute(sql, {"user_id": user_id, "k": k * 4})
             rows = result.fetchall()
 
         return [
@@ -90,11 +90,11 @@ class SchemaRetriever(BaseRetriever):
         emb_str = "[" + ",".join(str(x) for x in embedding) + "]"
 
         if operator == "<#>":
-            score_sql = "(lpe.embedding <#> :emb::vector) * -1"
+            score_sql = f"(lpe.embedding <#> '{emb_str}'::vector) * -1"
         elif operator == "<->":
-            score_sql = "1.0 / (1.0 + (lpe.embedding <-> :emb::vector))"
+            score_sql = f"1.0 / (1.0 + (lpe.embedding <-> '{emb_str}'::vector))"
         else:
-            score_sql = "1.0 - (lpe.embedding <=> :emb::vector)"
+            score_sql = f"1.0 - (lpe.embedding <=> '{emb_str}'::vector)"
 
         sql = text(f"""
             SELECT lpe.document, lpe.cmetadata, {score_sql} AS score
@@ -105,12 +105,12 @@ class SchemaRetriever(BaseRetriever):
               AND lpe.cmetadata->>'source_type' = 'document'
               AND (lpe.cmetadata->'data'->>'file_type' = 'csv'
                 OR lpe.cmetadata->'data'->>'file_type' = 'xlsx')
-            ORDER BY lpe.embedding {operator} :emb::vector ASC
+            ORDER BY lpe.embedding {operator} '{emb_str}'::vector ASC
             LIMIT :k
         """)
 
         async with _pgvector_engine.connect() as conn:
-            result = await conn.execute(sql, {"user_id": user_id, "k": k * 4, "emb": emb_str})
+            result = await conn.execute(sql, {"user_id": user_id, "k": k * 4})
             rows = result.fetchall()
 
         results = []