Vibe Coding

app.py

@@ -6,6 +6,7 @@ import json
 import torch
 from datetime import datetime, timedelta
 from threading import Thread
+from pathlib import Path
 
 # Gradio and HuggingFace imports
 import gradio as gr
@@ -40,7 +41,6 @@ conversations = []
 # Uncomment this line to login with your token
 # login(token=os.environ.get("HF_TOKEN"))
 
-
 def save_to_dataset():
     """Save the current conversations to a HuggingFace dataset"""
     if not conversations:
@@ -77,17 +77,17 @@ def save_to_dataset():
     
     return dataset, status_msg
 
-
 @spaces.GPU(duration=120)
-def predict(message, chat_history, temperature, top_p, conversation_id=None):
-    """Generate a response using the model and save the conversation"""
-    # Create/retrieve conversation ID for tracking
-    if conversation_id is None or conversation_id == "":
+def chat_model(message, history, temperature=0.7, top_p=0.9):
+    """Chat function for use with ChatInterface"""
+    conversation_id = getattr(chat_model, "conversation_id", None)
+    if conversation_id is None:
         conversation_id = str(uuid.uuid4())
+        chat_model.conversation_id = conversation_id
     
     # Format chat history for the model
     formatted_history = []
-    for human_msg, ai_msg in chat_history:
+    for human_msg, ai_msg in history:
         formatted_history.append({"role": "user", "content": human_msg})
         if ai_msg:  # Skip None values that might occur during streaming
             formatted_history.append({"role": "assistant", "content": ai_msg})
@@ -126,14 +126,14 @@ def predict(message, chat_history, temperature, top_p, conversation_id=None):
     # Yield partial text as it's generated
     for new_text in streamer:
         partial_text += new_text
-        yield chat_history + [[message, partial_text]], conversation_id
-    
-    # Store conversation data
-    existing_conv = next((c for c in conversations if c["conversation_id"] == conversation_id), None)
+        yield partial_text
     
-    # Update history with final response
+    # Store conversation data in the global conversations list
     formatted_history.append({"role": "assistant", "content": partial_text})
     
+    # Find existing conversation or create new one
+    existing_conv = next((c for c in conversations if c["conversation_id"] == conversation_id), None)
+    
     # Update or create conversation record
     current_time = datetime.now().isoformat()
     if existing_conv:
@@ -162,16 +162,12 @@ def predict(message, chat_history, temperature, top_p, conversation_id=None):
     if current_time_dt - last_save_time > timedelta(minutes=SAVE_INTERVAL_MINUTES):
         save_to_dataset()
         last_save_time = current_time_dt
-    
-    return chat_history + [[message, partial_text]], conversation_id
-
 
 def save_dataset_manually():
-    """Manually trigger dataset save"""
+    """Manually trigger dataset save and return status"""
     _, status = save_to_dataset()
     return status
 
-
 def get_stats():
     """Get current stats about conversations and saving"""
     mins_until_save = SAVE_INTERVAL_MINUTES - (datetime.now() - last_save_time).seconds // 60
@@ -185,8 +181,7 @@ def get_stats():
         "dataset_name": DATASET_NAME
     }
 
-
-# Create a Stanford theme using the simpler approach from Gradio examples
+# Create a Stanford theme
 theme = gr.themes.Default(
     primary_hue=gr.themes.utils.colors.red,
     secondary_hue=gr.themes.utils.colors.gray,
@@ -207,7 +202,7 @@ theme = gr.themes.Default(
     block_label_background_fill="#f9f9f9"
 )
 
-# Custom CSS for additional Stanford styling
+# Custom CSS
 css = """
 .gradio-container {
     font-family: 'Source Sans Pro', sans-serif !important;
@@ -218,41 +213,47 @@ css = """
 }
 """
 
-# Set up the Gradio app
+# Set up the Gradio app with Blocks for more control
 with gr.Blocks(theme=theme, title="Stanford Soft Raccoon Chat", css=css) as demo:
-    conversation_id = gr.State("")
-    
     with gr.Row():
         with gr.Column(scale=3):
-            chatbot = gr.Chatbot(
-                label="Soft Raccoon Chat",
-                avatar_images=(None, "🌲"),  # Stanford tree emoji
-                height=600
+            # Use ChatInterface for the main chat functionality
+            chatbot = gr.ChatInterface(
+                fn=chat_model,
+                chatbot=gr.Chatbot(
+                    label="Soft Raccoon Chat",
+                    avatar_images=(None, "🌲"),  # Stanford tree emoji
+                    height=600,
+                    placeholder="<strong>Soft Raccoon AI Assistant</strong><br>Ask me anything!"
+                ),
+                additional_inputs=[
+                    gr.Slider(
+                        minimum=0.1,
+                        maximum=2.0,
+                        value=0.7,
+                        step=0.1,
+                        label="Temperature"
+                    ),
+                    gr.Slider(
+                        minimum=0.1,
+                        maximum=1.0,
+                        value=0.9,
+                        step=0.05,
+                        label="Top-P"
+                    )
+                ],
+                title="Stanford Soft Raccoon Chat",
+                description="AI assistant powered by the Soft Raccoon language model",
+                examples=[
+                    "Tell me about Stanford University",
+                    "How can I learn about artificial intelligence?",
+                    "What's your favorite book?"
+                ],
+                cache_examples=True,
+                retry_btn="Regenerate",
+                undo_btn="Undo",
+                clear_btn="Clear",
             )
-            
-            with gr.Row():
-                msg = gr.Textbox(
-                    placeholder="Send a message...",
-                    show_label=False,
-                    container=False
-                )
-                submit_btn = gr.Button("Send", variant="primary")
-            
-            with gr.Accordion("Generation Parameters", open=False):
-                temperature = gr.Slider(
-                    minimum=0.1,
-                    maximum=2.0,
-                    value=0.7,
-                    step=0.1,
-                    label="Temperature"
-                )
-                top_p = gr.Slider(
-                    minimum=0.1,
-                    maximum=1.0,
-                    value=0.9,
-                    step=0.05,
-                    label="Top-P"
-                )
         
         with gr.Column(scale=1):
             with gr.Group():
@@ -270,20 +271,6 @@ with gr.Blocks(theme=theme, title="Stanford Soft Raccoon Chat", css=css) as demo
                 refresh_btn = gr.Button("Refresh Stats")
     
     # Set up event handlers
-    submit_btn.click(
-        predict,
-        [msg, chatbot, temperature, top_p, conversation_id],
-        [chatbot, conversation_id],
-        api_name="chat"
-    )
-    
-    msg.submit(
-        predict,
-        [msg, chatbot, temperature, top_p, conversation_id],
-        [chatbot, conversation_id],
-        api_name=False
-    )
-    
     save_button.click(
         save_dataset_manually,
         [],
@@ -309,10 +296,11 @@ with gr.Blocks(theme=theme, title="Stanford Soft Raccoon Chat", css=css) as demo
     demo.load(
         update_stats, 
         [], 
-        [convo_count, next_save, last_save_time_display, dataset_name_display]
+        [convo_count, next_save, last_save_time_display, dataset_name_display],
+        every=30  # Refresh every 30 seconds
     )
     
-    # Ensure we save on shutdown using atexit
+    # Ensure we save on shutdown
     import atexit
     atexit.register(save_to_dataset)