Update app.py

app.py

@@ -4,7 +4,8 @@ import sys
 import json 
 import requests
 import random
-
+import timeout_decorator
+from tenacity import retry, wait_fixed, stop_after_attempt
 
 MODEL = "o1-preview"
 API_URL = os.getenv("API_URL")
@@ -16,6 +17,13 @@ NUM_THREADS = int(os.getenv("NUM_THREADS"))
 
 print (NUM_THREADS)
 
+@retry(stop=stop_after_attempt(5), wait=wait_fixed(2))
+@timeout_decorator.timeout(120)
+def call_openai_api(payload, headers):
+    response = requests.post(API_URL, headers=headers, json=payload, stream=True)
+    response.raise_for_status()
+    return response
+
 def exception_handler(exception_type, exception, traceback):
     print("%s: %s" % (exception_type.__name__, exception))
 sys.excepthook = exception_handler
@@ -78,7 +86,7 @@ def predict(inputs, top_p, temperature, chat_counter, chatbot, history, request:
 
     try:
         # make a POST request to the API endpoint using the requests.post method, passing in stream=True
-        response = requests.post(API_URL, headers=headers, json=payload, stream=True)
+        response = call_openai_api(payload, headers) #requests.post(API_URL, headers=headers, json=payload, stream=True)
         response_code = f"{response}"
         #if response_code.strip() != "<Response [200]>":
         #    #print(f"response code - {response}")