diff --git a/bird/llm/run/run_gpt.sh b/bird/llm/run/run_gpt.sh
index 47d408b8..69041c8c 100644
--- a/bird/llm/run/run_gpt.sh
+++ b/bird/llm/run/run_gpt.sh
@@ -8,6 +8,7 @@ cot='True'
 no_cot='Fales'
 
 YOUR_API_KEY=''
+BASE_URL='https://api.openai.com/v1'
 
 engine1='code-davinci-002'
 engine2='text-davinci-003'
@@ -21,11 +22,11 @@ data_kg_output_path='./exp_result/turbo_output_kg/'
 
 
 echo 'generate GPT3.5 batch with knowledge'
-python3 -u ./src/gpt_request.py --db_root_path ${db_root_path} --api_key ${YOUR_API_KEY} --mode ${mode} \
+python3 -u ./src/gpt_request.py --db_root_path ${db_root_path} --api_key ${YOUR_API_KEY} --base_url ${BASE_URL} --mode ${mode} \
 --engine ${engine3} --eval_path ${eval_path} --data_output_path ${data_kg_output_path} --use_knowledge ${use_knowledge} \
 --chain_of_thought ${no_cot}
 
 echo 'generate GPT3.5 batch without knowledge'
-python3 -u ./src/gpt_request.py --db_root_path ${db_root_path} --api_key ${YOUR_API_KEY} --mode ${mode} \
+python3 -u ./src/gpt_request.py --db_root_path ${db_root_path} --api_key ${YOUR_API_KEY} --base_url ${BASE_URL} --mode ${mode} \
 --engine ${engine3} --eval_path ${eval_path} --data_output_path ${data_output_path} --use_knowledge ${not_use_knowledge} \
 --chain_of_thought ${no_cot}
diff --git a/bird/llm/src/gpt_request.py b/bird/llm/src/gpt_request.py
index da85e336..f7d20fea 100644
--- a/bird/llm/src/gpt_request.py
+++ b/bird/llm/src/gpt_request.py
@@ -10,14 +10,13 @@
 from typing import Dict, List, Tuple
 
 import backoff
-import openai
+from openai import OpenAI
 import pandas as pd
 import sqlparse
 from tqdm import tqdm
-'''openai configure'''
-
-openai.debug=True
 
+# Initialize OpenAI client instead of using the global openai module
+client = None  # We'll initialize this with the API key later
 
 def new_directory(path):  
     if not os.path.exists(path):  
@@ -147,51 +146,85 @@ def generate_combined_prompts_one(db_path, question, knowledge=None):
     return combined_prompts
 
 def quota_giveup(e):
-    return isinstance(e, openai.error.RateLimitError) and "quota" in str(e)
+    return "quota" in str(e)
 
+# Updated backoff decorator to use more generic exception handling
 @backoff.on_exception(
     backoff.constant,
-    openai.error.OpenAIError,
+    Exception,  # Using generic Exception instead of OpenAI specific error
     giveup=quota_giveup,
     raise_on_giveup=True,
     interval=20
 )
 def connect_gpt(engine, prompt, max_tokens, temperature, stop):
-    # print(prompt)
+    global client
     try:
-        result = openai.Completion.create(engine=engine, prompt=prompt, max_tokens=max_tokens, temperature=temperature, stop=stop)
+        # Use the new client-based API
+        if engine.startswith("code-"):
+            # For codex models, use completions endpoint
+            response = client.completions.create(
+                model=engine,
+                prompt=prompt,
+                max_tokens=max_tokens,
+                temperature=temperature,
+                stop=stop
+            )
+            return response
+        else:
+            # For newer models, use chat completions endpoint
+            response = client.chat.completions.create(
+                model=engine,
+                messages=[{"role": "user", "content": prompt}],
+                max_tokens=max_tokens,
+                temperature=temperature,
+                stop=stop
+            )
+            return response
     except Exception as e:
-        result = 'error:{}'.format(e)
-    return result
-def collect_response_from_gpt(db_path_list, question_list, api_key, engine, knowledge_list=None):
+        return f'error:{e}'
+
+def collect_response_from_gpt(db_path_list, question_list, api_key, base_url, engine, knowledge_list=None):
     '''
     :param db_path: str
     :param question_list: []
     :return: dict of responses collected from openai
     '''
+    global client
+    client = OpenAI(api_key=api_key, base_url=base_url)
+
     responses_dict = {}
     response_list = []
-    openai.api_key = api_key
+
     for i, question in tqdm(enumerate(question_list)):
         print('--------------------- processing {}th question ---------------------'.format(i))
         print('the question is: {}'.format(question))
-        
+
         if knowledge_list:
             cur_prompt = generate_combined_prompts_one(db_path=db_path_list[i], question=question, knowledge=knowledge_list[i])
         else:
             cur_prompt = generate_combined_prompts_one(db_path=db_path_list[i], question=question)
         
         plain_result = connect_gpt(engine=engine, prompt=cur_prompt, max_tokens=256, temperature=0, stop=['--', '\n\n', ';', '#'])
-        # pdb.set_trace()
-        # plain_result = connect_gpt(engine=engine, prompt=cur_prompt, max_tokens=256, temperature=0, stop=['</s>'])
-        # determine wheter the sql is wrong
-        
-        if type(plain_result) == str:
+
+        # Parse the response based on its type
+        if isinstance(plain_result, str):
             sql = plain_result
         else:
-            sql = 'SELECT' + plain_result['choices'][0]['text']
-        
-        # responses_dict[i] = sql
+            # Handle different response formats based on the API version
+            if hasattr(plain_result, 'choices') and hasattr(plain_result.choices[0], 'text'):
+                # Old completions API format
+                sql = 'SELECT' + plain_result.choices[0].text
+            elif hasattr(plain_result, 'choices') and hasattr(plain_result.choices[0], 'message'):
+                # New chat completions API format
+                message_content = plain_result.choices[0].message.content
+                if message_content.startswith('SELECT'):
+                    sql = message_content
+                else:
+                    sql = 'SELECT' + message_content
+            else:
+                # Fallback
+                sql = f"Error: Unexpected response format: {plain_result}"
+
         db_id = db_path_list[i].split('/')[-1].split('.sqlite')[0]
         sql = sql + '\t----- bird -----\t' + db_id # to avoid unpredicted \t appearing in codex results
         response_list.append(sql)
@@ -245,6 +278,7 @@ def generate_sql_file(sql_lst, output_path=None):
     args_parser.add_argument('--db_root_path', type=str, default='')
     # args_parser.add_argument('--db_name', type=str, required=True)
     args_parser.add_argument('--api_key', type=str, required=True)
+    args_parser.add_argument('--base_url', type=str, default='https://api.openai.com/v1')
     args_parser.add_argument('--engine', type=str, required=True, default='code-davinci-002')
     args_parser.add_argument('--data_output_path', type=str)
     args_parser.add_argument('--chain_of_thought', type=str)
@@ -259,9 +293,9 @@ def generate_sql_file(sql_lst, output_path=None):
     assert len(question_list) == len(db_path_list) == len(knowledge_list)
     
     if args.use_knowledge == 'True':
-        responses = collect_response_from_gpt(db_path_list=db_path_list, question_list=question_list, api_key=args.api_key, engine=args.engine, knowledge_list=knowledge_list)
+        responses = collect_response_from_gpt(db_path_list=db_path_list, question_list=question_list, api_key=args.api_key, base_url=args.base_url, engine=args.engine, knowledge_list=knowledge_list)
     else:
-        responses = collect_response_from_gpt(db_path_list=db_path_list, question_list=question_list, api_key=args.api_key, engine=args.engine, knowledge_list=None)
+        responses = collect_response_from_gpt(db_path_list=db_path_list, question_list=question_list, api_key=args.api_key, base_url=args.base_url, engine=args.engine, knowledge_list=None)
     
     if args.chain_of_thought == 'True':
         output_name = args.data_output_path + 'predict_' + args.mode + '_cot.json'