Update app.py

app.py

@@ -16,7 +16,7 @@ client = OpenAI(api_key=OPENAI_API_KEY)
 
 def gpt_call(history, user_message,
              model="gpt-4o",
-             max_tokens=1500,  # Increased from 512 to 1500 to prevent truncation
+             max_tokens=3000,  # Increased to 3000 to prevent truncation
              temperature=0.7,
              top_p=0.95):
     """
@@ -37,19 +37,28 @@ def gpt_call(history, user_message,
     # 3) Add the user's new message
     messages.append({"role": "user", "content": user_message})
 
-    # 4) Call OpenAI API
-    completion = client.chat.completions.create(
-        model=model,
-        messages=messages,
-        max_tokens=max_tokens,  # Increased to allow longer responses
-        temperature=temperature,
-        top_p=top_p
-    )
-
-    # 5) Ensure full response is returned without being cut off
-    full_response = "".join(choice.message.content for choice in completion.choices).strip()
-
-    return full_response
+    # 4) Call OpenAI API (with continuation handling)
+    full_response = ""
+    while True:
+        completion = client.chat.completions.create(
+            model=model,
+            messages=messages,
+            max_tokens=max_tokens,  # Increased to allow longer responses
+            temperature=temperature,
+            top_p=top_p
+        )
+        
+        response_part = completion.choices[0].message.content.strip()
+        full_response += " " + response_part
+
+        # If the response looks incomplete, force the AI to continue
+        if len(response_part) < max_tokens - 50:  # Ensures near full completion
+            break  # Stop loop if response is complete
+
+        # Add last response back into conversation history to continue it
+        messages.append({"role": "assistant", "content": response_part})
+
+    return full_response.strip()
 
 
 def respond(user_message, history):