import gradio as gr
from huggingface_hub import InferenceClient
import requests
from bs4 import BeautifulSoup
import re

# Initialize Hugging Face InferenceClient
client = InferenceClient(model="HuggingFaceH4/zephyr-7b-beta")

def is_url(input_string):
    url_pattern = re.compile(r'^(https?://)?([a-zA-Z0-9.-]+(?:\.[a-zA-Z]{2,4}))(?:/[^\s]*)?/?$')
    return url_pattern.match(input_string) is not None

def scrape_website(url):
    try:
        response = requests.get(url)
        soup = BeautifulSoup(response.text, 'html.parser')
        return ' '.join([p.text for p in soup.find_all('p')])
    except Exception as e:
        return f"Failed to scrape the website. Error: {str(e)}"

def generate_response(input_text):
    if is_url(input_text):
        scraped_content = scrape_website(input_text)
        return f"I've found the following content from the website:\n\n{scraped_content}"
    else:
        response = client.text_generation(f"User: {input_text}\nAssistant:", max_new_tokens=100, temperature=0.6)
        return response['generated_text']

def process_input(input_text):
    return generate_response(input_text)

# Define Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Conversational Chatbot with Web Scraping Ability")
    
    with gr.Row():
        with gr.Column():
            text_input = gr.Textbox(label="Enter your message or URL here")
            submit_button = gr.Button("Submit")
        
        with gr.Column():
            response_output = gr.Textbox(label="Chatbot Response", interactive=False)
    
    submit_button.click(fn=process_input, inputs=[text_input], outputs=[response_output])
    
demo.launch()