Update app.py

app.py

@@ -2,10 +2,12 @@ import gradio as gr
 from transformers import pipeline
 import torch
 
-# 1. SETUP
+# 1. SETUP: Qwen 0.5B - Optimized for Free CPU
 model_id = "Qwen/Qwen2.5-0.5B-Instruct"
-print("Loading Mochi to CPU...")
 
+print("Loading Mochi to CPU... Stay chill.")
+
+# Initialize the pipeline
 pipe = pipeline(
     "text-generation", 
     model=model_id, 
@@ -14,47 +16,60 @@ pipe = pipeline(
 )
 
 def chat_fn(message, history):
-    # THE SOUL
+    # THE SOUL: Define Mochi's personality
     character_card = (
         "You are Mochi, a chill, slightly sarcastic, but deeply loyal best friend. "
         "Talk like a real person in their 20s: use lowercase, occasional slang, and short sentences. "
         "Be supportive but roast the user occasionally. Keep answers short and snappy."
     )
 
+    # Start with the personality
     messages = [{"role": "system", "content": character_card}]
     
-    # THE MEMORY (Auto-detecting format to prevent 'unpack' errors)
+    # 2. THE CLEANER: This part is the most important.
+    # It converts any version of Gradio history into a clean format for the AI.
     for entry in history:
-        if isinstance(entry, dict):
-            # If history is [{'role': 'user', 'content': '...'}, ...]
-            messages.append({"role": entry["role"], "content": entry["content"]})
-        elif isinstance(entry, (list, tuple)):
-            # If history is [[user_msg, bot_msg], ...]
-            messages.append({"role": "user", "content": str(entry[0])})
-            messages.append({"role": "assistant", "content": str(entry[1])})
-    
-    # Add the current message
+        # Check if entry is the new Gradio 6.0 dict format
+        if isinstance(entry, dict) and "role" in entry and "content" in entry:
+            content = entry["content"]
+            # If content is a list (Gradio 6.0 style), extract the text
+            if isinstance(content, list):
+                text = next((item["text"] for item in content if item.get("type") == "text"), "")
+            else:
+                text = str(content)
+            messages.append({"role": entry["role"], "content": text})
+            
+        # Check if entry is the old Gradio 5.0 tuple/list format
+        elif isinstance(entry, (list, tuple)) and len(entry) == 2:
+            if entry[0]: messages.append({"role": "user", "content": str(entry[0])})
+            if entry[1]: messages.append({"role": "assistant", "content": str(entry[1])})
+
+    # Add the current message from the user
     messages.append({"role": "user", "content": message})
     
-    # THE GENERATION
+    # 3. GENERATION
     try:
+        # We limit max_new_tokens to 128 to keep it fast on CPU
         generation = pipe(
             messages, 
             max_new_tokens=128, 
             do_sample=True, 
-            temperature=0.85, 
+            temperature=0.8, 
             top_p=0.9,
             truncation=True 
         )
+        # Extract response
         return generation[0]['generated_text'][-1]['content']
     except Exception as e:
-        print(f"Error: {e}")
+        print(f"Error during generation: {e}")
         return "my bad, my brain just glitched. lol try again?"
 
-# 4. THE UI (Minimalist to avoid version conflicts)
+# 4. THE UI
 with gr.Blocks() as demo:
     gr.Markdown("# 🐾 Mochi AI")
-    gr.Markdown("*Unlimited chill vibes.*")
+    gr.Markdown("*Your chill, unlimited best friend.*")
+    
+    # We remove 'type' and let the 'Cleaner' above handle whatever Gradio sends
     gr.ChatInterface(fn=chat_fn)
 
 # 5. THE LAUNCH