Update app.py

app.py

@@ -1,20 +1,20 @@
 import os
-import gradio as gradio
+import gradio as gr
 from huggingface_hub import hf_hub_download
 from llama_cpp import Llama
 
 # 1. Download the highly optimized Llama 3.2 3B model from Hugging Face repository
-print("Downloading model... This happens only on the first run.")
+print("Downloading model...")
 model_path = hf_hub_download(
     repo_id="bartowski/Llama-3.2-3B-Instruct-GGUF",
-    filename="Llama-3.2-3B-Instruct-Q4_K_M.gguf" # Compressed to fit perfectly in 16GB RAM
+    filename="Llama-3.2-3B-Instruct-Q4_K_M.gguf"
 )
 
 # 2. Initialize the model on the CPU
 print("Initializing model...")
 llm = Llama(
     model_path=model_path,
-    n_ctx=2048,       # Context length (how much text it remembers)
+    n_ctx=2048,       # Context length
     n_threads=2       # Utilizes both free CPU cores fully
 )
 
@@ -25,7 +25,11 @@ def respond(message, chat_history):
     formatted_prompt += "You are a helpful, direct, and honest AI assistant.<|eot_id|>"
     
     # Inject chat history so the bot remembers the conversation context
-    for user_msg, bot_msg in chat_history:
+    for turn in chat_history:
+        # Check if history is structured as objects or dicts (Gradio 6 style)
+        user_msg = turn.get("text") if isinstance(turn, dict) else turn[0]
+        bot_msg = turn.get("text") if isinstance(turn, dict) else turn[1]
+        
         if user_msg:
             formatted_prompt += f"<|start_header_id|>user<|end_header_id|>\n{user_msg}<|eot_id|>"
         if bot_msg:
@@ -44,12 +48,12 @@ def respond(message, chat_history):
     
     token_accumulator = ""
     for token in output:
-        token_text = token["choices"][0]["text"]
+        token_text = token["choices"]["text"]
         token_accumulator += token_text
         yield token_accumulator
 
-# 4. Create the web dashboard layout using Gradio
-demo = gradio.ChatInterface(
+# 4. Create the web dashboard layout using Gradio ChatInterface
+demo = gr.ChatInterface(
     fn=respond,
     title="🤖 Free Llama 3.2 CPU Chatbot",
     description="Running 24/7/365 for free on Hugging Face Spaces using CPU inference.",