diff --git a/src/json_utils.py b/src/json_utils.py
index cbbab8d..84c0817 100644
--- a/src/json_utils.py
+++ b/src/json_utils.py
@@ -98,5 +98,8 @@ def extract_json_fragment(text: str) -> Any:
             last_error = err
             continue
     
+    print("CLEANED:", cleaned)
+    print("FRAGMENT:", fragment)
+    print("FRAGMENT repr:", repr(fragment))
     print("Failed JSON text:", text) # 디버깅용 출력
     raise ValueError(f"JSON 파싱에 실패했습니다: {last_error}")
\ No newline at end of file
diff --git a/src/lambda_function.py b/src/lambda_function.py
index 42de60e..ae3c482 100644
--- a/src/lambda_function.py
+++ b/src/lambda_function.py
@@ -1,4 +1,5 @@
 import json
+import hashlib
 from trafilatura import extract
 import boto3
 
@@ -50,22 +51,23 @@ def lambda_handler(event, context):
 
     # url에서 쿼리 파라미터(?), 해시(#) 제거
     url = url.split('?')[0].split('#')[0]
+    url_hashed = hashlib.sha256(url.encode('utf-8')).hexdigest()
 
     s3 = boto3.client("s3")
     bucket = 'inha-capstone-20-tos-content'
     dynamodb = boto3.resource('dynamodb')
     table = dynamodb.Table('inha-capstone-20-tos-analysis')
-    key = url
+    key = url_hashed
 
     # try: S3에서 해당 url을 key로 갖는 객체 가져오기
     try:
-        response = s3.get_object(Bucket=bucket, Key=url)
+        response = s3.get_object(Bucket=bucket, Key=key)
         saved_tos_content = response['Body'].read().decode('utf-8')
 
         # 기존 tos_content와 비교
         if saved_tos_content == tos_content:
             # 동일하면 DynamoDB에서 이전 분석 결과를 가져와 return
-            db_response = table.get_item(Key={'url': url})
+            db_response = table.get_item(Key={'url': key})
             if 'Item' in db_response:
                 evaluation_result = db_response['Item']
                 print("캐시 존재, 이전 분석 결과 반환")
@@ -140,7 +142,7 @@ def lambda_handler(event, context):
 
     # DynamoDB에 분석 결과 저장
     table.put_item(Item={
-        'url': url,
+        'url': key,
         'overall_evaluation': evaluation_result.get("overall_evaluation"),
         'evaluation_for_each_clause': evaluation_result.get("evaluation_for_each_clause")
     })
diff --git a/src/llm_client.py b/src/llm_client.py
index 6bb7012..e3896c5 100644
--- a/src/llm_client.py
+++ b/src/llm_client.py
@@ -82,4 +82,4 @@ def _generate_response_bedrock_claude(self, system_instruction: str, message: st
             messages=[{"role": "user", "content": [{"text": message}]}]
         )
 
-        return response['output']['message']['content'][0]['text'] if not selected_model.startswith("openai") else response['output']['message']['content'][1]['text']
\ No newline at end of file
+        return response['output']['message']['content'][0]['text'] if not selected_model.startswith("openai") else response['output']['message']['content'][-1]['text']
\ No newline at end of file