Update helpers.py

helpers.py

@@ -131,20 +131,42 @@ key = os.getenv("GOOGLE_API_KEY")
 #     return DoclingLoader(file_path=file_path, chunker=chunker  # This will break your doc into manageable pieces.
 #                         ).load()
 
-def extract_metadata(input_string):
-    # Use regex to find the content inside curly braces
-    match = re.search(r'\{.*?\}', input_string)
-    if match:
-        metadata_str = match.group()  # This returns the substring with the braces
+# def extract_metadata(input_string):
+#     # Use regex to find the content inside curly braces
+#     match = re.search(r'\{.*?\}', input_string)
+#     if match:
+#         metadata_str = match.group()  # This returns the substring with the braces
+#         try:
+#             # Safely evaluate the string to a dictionary
+#             new_metadata = ast.literal_eval(metadata_str)
+#         except Exception as e:
+#             print(f"Error evaluating metadata: {e}")
+#             new_metadata = {}
+#     else:
+#         new_metadata = None
+#     return new_metadata
+
+def extract_metadata(response_string):
+    # Tìm tất cả các dictionary trong chuỗi đầu vào
+    matches = re.findall(r'\{.*?\}', response_string, re.DOTALL)
+    if not matches:
+        return None
+    
+    smallest_dict = None
+    min_length = float("inf")
+    
+    for match in matches:
         try:
-            # Safely evaluate the string to a dictionary
-            new_metadata = ast.literal_eval(metadata_str)
-        except Exception as e:
-            print(f"Error evaluating metadata: {e}")
-            new_metadata = {}
-    else:
-        new_metadata = None
-    return new_metadata
+            parsed_dict = ast.literal_eval(match)  # Chuyển đổi string thành dictionary
+            if isinstance(parsed_dict, dict):
+                dict_length = len(str(parsed_dict))  # Độ dài chuỗi của dict
+                if dict_length < min_length:
+                    smallest_dict = parsed_dict
+                    min_length = dict_length
+        except Exception:
+            continue  # Bỏ qua nếu không phải dictionary hợp lệ
+    
+    return smallest_dict
 
 # # Example usage:
 # input_str = "Some random text before and then {'a': 'abc', 'b': 'bcd'} and some random text after."