import gradio as gr
import re
import spacy
from transformers import pipeline

# Load spaCy's English model
nlp = spacy.load("en_core_web_sm")

def preprocess_text(text):
    doc = nlp(text.lower())  # Tokenize and lowercase the text
    tokens = [token.text for token in doc if not token.is_punct]  # Remove punctuation
    return tokens

# Load the multilingual model for question answering
qa_model = pipeline("question-answering", model="deepset/xlm-roberta-large-squad2")

# Function to generate the answer based on question and uploaded context
def answer_question(question, context):
    try:
        preprocessed_context = preprocess_text(context)
        result = qa_model(question=question, context=" ".join(preprocessed_context))
        return result['answer']
    except Exception as e:
        return f"Error: {str(e)}"

# Gradio interface
def qa_app(text_file, question):
    try:
        with open(text_file.name, 'r') as file:
            context = file.read()
        return answer_question(question, context)
    except Exception as e:
        return f"Error reading file: {str(e)}"

# Create Gradio interface with updated syntax
iface = gr.Interface(
    fn=qa_app,  # The function that processes input
    inputs=[gr.File(label="Upload your text file"), gr.Textbox(label="Enter your question")],
    outputs="text",
    title="Multilingual Question Answering",
    description="Upload a text file and ask a question based on its content."
)

# Launch the Gradio app
iface.launch()