import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load the mT5-small model and tokenizer
model_name = "google/mt5-small"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Define the chatbot function for summarization and answering questions
def chatbot(user_input):
    # Tokenize the user input
    inputs = tokenizer(user_input, return_tensors="pt", max_length=512, truncation=True)
    
    # Generate a response (you can customize max_length and num_beams for different outputs)
    outputs = model.generate(inputs["input_ids"], max_length=150, num_beams=2, early_stopping=True)
    
    # Decode and return the generated text
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response

# Set up the Gradio interface
demo = gr.Interface(fn=chatbot, inputs="text", outputs="text", title="mT5-Small Chatbot")

# Launch the app
demo.launch()