init inference

app.py

@@ -18,26 +18,26 @@ def chat_with_model(user_input, chat_history):
     """
     Process user input and generate a response from the model.
     :param user_input: User's input string
-    :param chat_history: Conversation history
+    :param chat_history: List of [user_message, ai_response] pairs
     :return: Updated chat history
     """
-    # Format chat history for the Llama model
+    # Combine chat history into a single prompt
     prompt = ""
-    for turn in chat_history:
-        prompt += f"User: {turn['user']}\nAI: {turn['ai']}\n"
+    for user, ai in chat_history:
+        prompt += f"User: {user}\nAI: {ai}\n"
     prompt += f"User: {user_input}\nAI:"
 
     # Generate response from the model
     response = llm(prompt)["choices"][0]["text"].strip()
 
-    # Update chat history
-    chat_history.append({"user": user_input, "ai": response})
+    # Update chat history as a list of tuples
+    chat_history.append((user_input, response))
     return chat_history, chat_history
 
 # Gradio UI
 with gr.Blocks() as demo:
     gr.Markdown("# 🦙 LLaMA GGUF Chatbot")
-    chat_box = gr.Chatbot(label="Chat with the GGUF Model")
+    chatbot = gr.Chatbot(label="Chat with the GGUF Model")
 
     with gr.Row():
         with gr.Column(scale=4):
@@ -51,7 +51,7 @@ with gr.Blocks() as demo:
     submit_btn.click(
         chat_with_model,
         inputs=[user_input, chat_history],
-        outputs=[chat_box, chat_history],
+        outputs=[chatbot, chat_history],
         show_progress=True,
     )