import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-small")
model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-small")

# Chat history (global per session)
chat_history_ids = None

def vibebot_response(user_input, history=[]):
    global chat_history_ids
    # Encode user input
    new_input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors='pt')

    # Append to chat history if exists
    bot_input_ids = torch.cat([chat_history_ids, new_input_ids], dim=-1) if chat_history_ids is not None else new_input_ids

    # Generate response
    chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)
    response = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)

    # Append for gradio history
    history.append((user_input, response))
    return history, history

# Gradio interface
demo = gr.ChatInterface(fn=vibebot_response, title="VibeBot: Gen-Z Therapist",
                        chatbot=gr.Chatbot(height=400), theme="default")

demo.launch()