Update app.py

app.py

@@ -17,29 +17,28 @@ history_openai_format = [
 
 # Define the function that takes a name and text input to generate the speech
 def generate_speech(name, input_text):
+    global history_openai_format  # Use the global history variable to maintain state
+    
     # Include the user's name in the dialogue if provided
     personalized_greeting = f"Welcome to the Tonight Show, {name}. " if name else ""
     
     # Append user message to history with personalized greeting
-    history_openai_format.append({"role": "user", "content": personalized_greeting + input_text})
+    current_exchange = [{"role": "user", "content": personalized_greeting + input_text}]
 
-    # Build completion with OpenAI using the accumulated history
+    # Build completion with OpenAI using the current exchange and the accumulated history
     completion = client.chat.completions.create(
         model="gpt-3.5-turbo-1106",
-        messages=history_openai_format
+        messages=history_openai_format + current_exchange
     )
 
     # Extract generated text (response by the assistant) from completion
     message_content = completion.choices[0].message.content.strip()
 
-    # Append assistant's message to history
-    # NOTE: This is crucial for keeping memory of the conversation.
-    history_openai_format.append({"role": "assistant", "content": message_content})
-
-    # Prepare the text for TTS conversion of just the latest reply, NOT the entire conversation
-    text_to_speech = message_content  # Only the latest message from the assistant is used
+    # Update the conversation history with the current exchange
+    # This ensures the full history is preserved for the next call but only the latest messages are sent to TTS
+    history_openai_format.extend(current_exchange + [{"role": "assistant", "content": message_content}])
 
-    # ElevenLabs TTS API settings and request for the latest response
+    # Use ElevenLabs TTS API settings and request for the latest assistant response
     url = "https://api.elevenlabs.io/v1/text-to-speech/eIH76iW9yjOWnOKIkd6y/stream"
     headers = {
         "Accept": "audio/mpeg",
@@ -61,10 +60,12 @@ def generate_speech(name, input_text):
     	}
     }
 
+    # Send the request to ElevenLabs API
     response = requests.post(url, json=data, headers=headers)
 
-    # Return the response content directly if successful, otherwise print error details
+    # Return the response content if successful, otherwise, print error details
     if response.status_code == 200:
+        # Return only the audio of the latest assistant message
         return response.content
     else:
         print("Error with ElevenLabs API:", response.status_code, response.text)