app.py

import re
import gradio as gr
from transformers import pipeline, BartTokenizer, BartForConditionalGeneration

# Load the BART model and tokenizer for text generation (answer suggestions)
model_name = "facebook/bart-large-cnn"
tokenizer = BartTokenizer.from_pretrained(model_name)
model = BartForConditionalGeneration.from_pretrained(model_name)

# Question detection function
def detect_questions(email_text):
    # Use regex to find questions in the email
    questions = re.findall(r'([^\.\!\?]*\?)', email_text)
    return questions

# Generate answers using the BART model
def generate_answers(question):
    # Use the BART model to generate a response
    inputs = tokenizer(question, return_tensors="pt", max_length=1024, truncation=True)
    summary_ids = model.generate(inputs["input_ids"], num_beams=4, max_length=50, early_stopping=True)
    answer = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
    return answer

# Main function to handle the email input
def process_email(email_text):
    questions = detect_questions(email_text)
    responses = {}
    
    for question in questions:
        response = generate_answers(question)
        responses[question] = response
    
    return responses

# Create a Gradio interface
iface = gr.Interface(
    fn=process_email,
    inputs="textbox",
    outputs="json",
    title="Email Question Detector and Responder",
    description="Input an email, and the AI will detect questions and provide suggested responses.",
)

# Launch the interface
iface.launch()