Update app.py

app.py

@@ -7,7 +7,7 @@ client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
 def generate_dialogue(topic):
     # System message to instruct the model
-    system_message = "Generate a well-structured and engaging dialogue between two people on the given topic. Ensure proper spacing and formatting."
+    system_message = "Generate a short and engaging dialogue between two people on the given topic."
 
     # Conversation setup
     messages = [
@@ -16,48 +16,38 @@ def generate_dialogue(topic):
     ]
 
     response = ""
-    last_sentence = ""
 
-    # Generate the dialogue dynamically
+    # Generate the dialogue with streaming and delay
+    start_time = time.time()
+    
     for msg in client.chat_completion(
         messages,
-        max_tokens=512,  # Adjust for enough dialogue
+        max_tokens=250,  # Enough tokens for a dialogue
         stream=True,
         temperature=0.7,
         top_p=0.95,
     ):
-        token = msg.choices[0].delta.content.strip() if msg.choices[0].delta.content else ""
-
-        # Ensure correct spacing between words and punctuation
-        if response and not response.endswith((" ", "\n")) and not token.startswith((" ", "\n", ".", "?", "!", ",")):
-            response += " "
-
+        token = msg.choices[0].delta.content
         response += token
+        
+        # Yield response to stream the output progressively
+        yield response.strip()
 
-        # Track last complete sentence to avoid abrupt cuts
-        if token.endswith((".", "?", "!")):
-            last_sentence = response  
-
-        yield response  # Stream output
-
-        #time.sleep(0.1)  # Slow down slightly for smoother streaming
-
-    # Ensure the dialogue ends at a complete sentence
-    if response and response[-1] not in ".!?":
-        response = last_sentence  
-
-    yield response  # Final output
+        # # If the dialogue is complete, stop after 10 seconds
+        # if time.time() - start_time >= 10:
+        #     break  
+        # time.sleep(0.2)  # Slows token output for smoother streaming
 
 # Define the Gradio UI
 with gr.Blocks() as demo:
     gr.Markdown("## 🎭 AI Dialogue Generator")
-    gr.Markdown("Enter a topic, and the AI will generate a structured dialogue between two people.")
+    gr.Markdown("Enter a topic, and the AI will generate a short dialogue between two people.")
 
     with gr.Row():
         topic_input = gr.Textbox(label="Enter a topic", placeholder="e.g., Space Travel, AI in Healthcare")
         generate_btn = gr.Button("Generate Dialogue")
     
-    output_box = gr.TextArea(label="Generated Dialogue", interactive=False, lines=12, autoscroll=True)
+    output_box = gr.Textbox(label="Generated Dialogue", interactive=False, lines=10)
 
     generate_btn.click(generate_dialogue, inputs=topic_input, outputs=output_box)