new update

app.py

@@ -3,6 +3,8 @@ import time
 import requests
 import json
 import os
+from urllib3.util import Retry
+from requests.adapters import HTTPAdapter
 
 API_URL = os.getenv("API_URL")
 API_KEY = os.getenv("API_KEY")
@@ -114,6 +116,26 @@ with gr.Blocks() as demo:
         history_messages = sys_msg + history_messages
         print(history_messages)
 
+        # Create a session object
+        session = requests.Session()
+
+        # Define the retry strategy
+        retries = Retry(
+            total=5,  # Total number of retries to allow
+            backoff_factor=1,  # A backoff factor to apply between attempts
+            status_forcelist=[
+                500,
+                502,
+                503,
+                504,
+            ],  # A set of HTTP status codes that we should force a retry on
+            method_whitelist=[
+                "HEAD",
+                "GET",
+                "OPTIONS",
+                "POST",
+            ],  # HTTP methods to retry on
+        )
         data = {
             "messages": history_messages,
             "stream": True,
@@ -129,34 +151,47 @@ with gr.Blocks() as demo:
             # },
         }
 
-        # Making the POST request and streaming the response
-        response = requests.post(
-            url, headers=headers, data=json.dumps(data), stream=True
-        )
-        for line in response.iter_lines():
-            # Filter out keep-alive new lines
-            if line:
-                data = line.decode("utf-8").lstrip("data: ")
-                # Check if the examples are valid
-                valid_check = is_valid_json(data)
-                if valid_check[0]:
-                    try:
-                        # Attempt to parse the JSON dataa
-                        # json_data = json.loads(data)
-                        json_data = valid_check[1]
-
-                        delta_content = (
-                            json_data.get("choices", [{}])[0]
-                            .get("delta", {})
-                            .get("content", "")
-                        )
-
-                        if delta_content:  # Ensure there's content to print
-                            history[-1][1] += delta_content
-                            time.sleep(0.05)
-                            yield history
-                    except json.JSONDecodeError as e:
-                        print(f"Error decoding JSON: {e} date: {data}")
+        # Mount it for http usage
+        session.mount("http://", HTTPAdapter(max_retries=retries))
+
+        # Making the POST request with increased timeout and retry logic
+        try:
+            response = session.post(
+                url,
+                headers=headers,
+                data=json.dumps(data),
+                stream=True,
+                timeout=(10, 30),
+            )
+
+            for line in response.iter_lines():
+                if line:
+                    for line in response.iter_lines():
+                        # Filter out keep-alive new lines
+                        if line:
+                            data = line.decode("utf-8").lstrip("data: ")
+                            # Check if the examples are valid
+                            valid_check = is_valid_json(data)
+                            if valid_check[0]:
+                                try:
+                                    # Attempt to parse the JSON dataa
+                                    # json_data = json.loads(data)
+                                    json_data = valid_check[1]
+
+                                    delta_content = (
+                                        json_data.get("choices", [{}])[0]
+                                        .get("delta", {})
+                                        .get("content", "")
+                                    )
+
+                                    if delta_content:  # Ensure there's content to print
+                                        history[-1][1] += delta_content
+                                        time.sleep(0.05)
+                                        yield history
+                                except json.JSONDecodeError as e:
+                                    print(f"Error decoding JSON: {e} date: {data}")
+        except requests.exceptions.RequestException as e:
+            print(f"An error occurred: {e}")
 
     msg.submit(
         user, [msg, chatbot], [msg, chatbot], queue=True, concurrency_limit=10