import streamlit as st
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load model and tokenizer from Hugging Face Hub
@st.cache_resource
def load_model():
    model_name = "sshleifer/tiny-gpt2"
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForCausalLM.from_pretrained(model_name)
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.to(device)
    model.eval()
    return tokenizer, model, device

tokenizer, model, device = load_model()

# Initialize chat history in session
if "messages" not in st.session_state:
    st.session_state.messages = []

# Chat UI
st.title("🤖 Tiny GPT-2 Chatbot")
st.markdown("Ask me anything! This bot runs locally with no API key.")

# Display chat history
for msg in st.session_state.messages:
    role = "🧑‍💻" if msg["role"] == "user" else "🤖"
    st.markdown(f"**{role}:** {msg['content']}")

# Input box
user_input = st.text_input("Type your message...", key="user_input")

if user_input:
    # Append user message
    st.session_state.messages.append({"role": "user", "content": user_input})

    # Build prompt
    prompt = "\n".join([m["content"] for m in st.session_state.messages])

    # Tokenize and generate
    inputs = tokenizer(prompt, return_tensors="pt").to(device)
    outputs = model.generate(
        **inputs,
        max_new_tokens=100,
        temperature=0.7,
        top_p=0.95,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id
    )

    output_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    reply = output_text[len(prompt):].strip().split("\n")[0]

    # Append bot reply
    st.session_state.messages.append({"role": "assistant", "content": reply})

    # Refresh the page to show new message
    st.experimental_rerun()