app.py

import gradio as gr
import os
import openai
import requests
import re
from newspaper import Article
import transformers
from transformers import GPT2Tokenizer

# Silence transformers backend warnings
transformers.logging.set_verbosity_error()

def text_prompt(request: str, page_url: str, api_key: str, temp: float):
    # Fetch & parse
    try:
        headers = {'User-Agent': 'Chrome/83.0.4103.106'}
        resp = requests.get(page_url, headers=headers, timeout=10)
        resp.raise_for_status()
        page = Article('')
        page.set_html(resp.text)
        page.parse()
    except Exception as e:
        return "", f"Error fetching URL: {e}", ""

    # Tokenize & truncate to ~1800 GPT-2 tokens
    tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
    sentences = page.text.split('.')
    tokens_accum, truncated_text = [], ""
    for sent in sentences:
        toks = tokenizer.tokenize(sent + ".")
        if len(tokens_accum) + len(toks) > 1800:
            break
        tokens_accum += toks
        truncated_text += sent + ". "
    num_input_tokens = len(tokens_accum)

    if num_input_tokens < 10:
        return page.text, f"Not enough text ({num_input_tokens} tokens)", num_input_tokens

    # Call GPT-4o mini via ChatCompletion
    openai.api_key = api_key
    try:
        chat_resp = openai.ChatCompletion.create(
            model="gpt-4o-mini",
            messages=[
                {"role": "system", "content": "You are a helpful assistant."},
                {"role": "user",   "content": request + "\n\n>>\n" + truncated_text + "\n<<"}
            ],
            temperature=temp,
            max_tokens=2048,
            top_p=0.9,
        )
        ai_text      = re.sub(r'\s+', ' ', chat_resp.choices[0].message.content).strip()
        total_tokens = chat_resp.usage.total_tokens
        return page.text, ai_text, total_tokens

    except Exception as e:
        return page.text, f"OpenAI API error: {e}", num_input_tokens

if __name__ == "__main__":
    iface = gr.Interface(
        fn=text_prompt,
        inputs=[
            gr.Textbox(label="Prompt:"),
            gr.Textbox(label="URL to parse:"),
            gr.Textbox(label="API-Key:", type="password"),
            gr.Slider(0.0,1.0,value=0.3, label="Temperature:")
        ],
        outputs=[
            gr.Textbox(label="Input Text:"),
            gr.Textbox(label="AI Output:"),
            gr.Textbox(label="Total Tokens:")
        ],
        title="GPT-4o-mini URL Summarizer",
        description="Uses GPT-4o-mini via ChatCompletion to summarize webpage text."
    )

    try:
        iface.queue()
        iface.launch()
    except Exception as e:
        print("Failed to launch:", e)