import torch
import gradio as gr
from transformers import pipeline

# Use a Hebrew question-answering model
model_name = "avichr/heBERT"

question_answer = pipeline("question-answering", model=model_name, tokenizer=model_name)

def read_file_content(file_obj):
    """
    Reads the content of a file object and returns it.
    Parameters:
    file_obj (file object): The file object to read from.
    Returns:
    str: The content of the file.
    """
    try:
        with open(file_obj.name, 'r', encoding='utf-8') as file:
            context = file.read()
            return context
    except Exception as e:
        return f"An error occurred: {e}"

def get_answer(file, question):
    context = read_file_content(file)
    answer = question_answer(question=question, context=context)
    return answer["answer"]

demo = gr.Interface(
    fn=get_answer,
    inputs=[gr.File(label="Upload your file"), gr.Textbox(label="Input your question", lines=1)],
    outputs=[gr.Textbox(label="Answer text", lines=1)],
    title="Document Q & A - Hebrew",
    description="This application will be used to answer questions based on the context provided."
)

demo.launch()