app.py

import os
import openai
import gradio as gr
from bs4 import BeautifulSoup
import requests

openai.api_key = os.getenv("OPENAI_API_KEY")

def extract_text_from_url(url):
    try:
        resp = requests.get(url, timeout=30, headers={
            "User-Agent": "Mozilla/5.0 (compatible; Bot/1.0)"
        })
        soup = BeautifulSoup(resp.content, "html.parser")
        candidates = soup.find_all(['h1','h2','h3','h4','p','span','li'])
        text = ' '.join([c.get_text(strip=True) for c in candidates])
        text = text[:4000]
        if len(text) < 100:
            raise ValueError("Could not extract enough content (site may require JavaScript). Please enter keywords manually.")
        return text
    except Exception as e:
        raise ValueError(f"URL extraction error: {e}")

def extract_keywords(text):
    prompt = f"""
    Extract up to 10 concise, relevant SEO keywords suitable for an automotive advertisement from the following content:
    {text}
    Keywords:
    """
    response = openai.ChatCompletion.create(
        model="gpt-4",
        messages=[{"role": "user", "content": prompt}],
        temperature=0.6,
        max_tokens=100
    )
    output = response.choices[0].message.content.strip()
    if ',' in output:
        keywords = output.split(',')
    else:
        keywords = output.split('\n')
    return [kw.strip() for kw in keywords if kw.strip()]

def generate_ad_copy(platform, keywords):
    prompt = f"""
    Create a compelling, SEO-optimized {platform} ad using these keywords: {', '.join(keywords)}. 
    Include a clear and enticing call-to-action.
    """
    response = openai.ChatCompletion.create(
        model="gpt-4",
        messages=[{"role": "user", "content": prompt}],
        temperature=0.7,
        max_tokens=300
    )
    return response.choices[0].message.content.strip()

def generate_ad_image(keywords):
    kw_str = ", ".join(keywords)
    # Enhanced prompt for better visuals
    image_prompt = (
        f"High-quality, photorealistic automotive ad photo of a luxury car. "
        f"Clean background, professional lighting, stylish dealership setting. "
        f"Keywords: {kw_str}. Room for text overlay, wide format, visually appealing."
    )
    response = openai.Image.create(
        prompt=image_prompt,
        n=1,
        size="512x512"
    )
    image_url = response["data"][0]["url"]
    img_data = requests.get(image_url).content
    img_file = "generated_ad_image.png"
    with open(img_file, "wb") as f:
        f.write(img_data)
    return img_file

def platform_html(platform, ad_text):
    # Platform-specific color and icons
    if platform == "Facebook":
        color = "#1877F2"
        icon = "🌐"
    elif platform == "Instagram":
        # Instagram gradient
        color = "linear-gradient(90deg, #f58529 0%, #dd2a7b 50%, #8134af 100%)"
        icon = "📸"
    elif platform == "X (Twitter)":
        color = "#14171A"
        icon = "🐦"
    else: # Google Search
        color = "#4285F4"
        icon = "🔍"

    if platform == "Instagram":
        # Gradient needs to be on a child div (not background-color)
        content = f"""
        <div style="background: {color}; padding: 2px; border-radius: 12px; margin-bottom:16px;">
            <div style="background: white; color: #333; padding: 18px 20px; border-radius: 10px;">
                <span style="font-size: 1.5em;">{icon} <b>{platform}</b></span>
                <div style="margin-top: 12px; font-size: 1.1em; line-height:1.6;">{ad_text}</div>
            </div>
        </div>
        """
    else:
        content = f"""
        <div style="background: {color}; color: white; padding: 18px 20px; border-radius: 12px; margin-bottom:16px; min-height: 120px;">
            <span style="font-size: 1.5em;">{icon} <b>{platform}</b></span>
            <div style="margin-top: 12px; font-size: 1.1em; line-height:1.6;">{ad_text}</div>
        </div>
        """
    return content

def main_workflow(input_mode, url_or_keywords):
    error = None
    keywords = []
    ad_copies = {}
    image_path = None

    if input_mode == "URL":
        try:
            text = extract_text_from_url(url_or_keywords)
            keywords = extract_keywords(text)
        except Exception as e:
            return None, None, None, f"{e}"
    else:
        keywords = [kw.strip() for kw in url_or_keywords.split(",") if kw.strip()]
        if not keywords:
            return None, None, None, "Please provide at least one keyword."
    # Generate ad copies
    platforms = ["Facebook", "Instagram", "X (Twitter)", "Google Search"]
    for platform in platforms:
        ad_copies[platform] = generate_ad_copy(platform, keywords)
    # Generate image
    try:
        image_path = generate_ad_image(keywords)
    except Exception as e:
        error = f"Image generation error: {e}"

    # Save ads to txt
    output_txt = "generated_ads.txt"
    with open(output_txt, "w", encoding="utf-8") as f:
        for platform, content in ad_copies.items():
            f.write(f"--- {platform} Ad Copy ---\n{content}\n\n")
    return keywords, ad_copies, image_path, error

def run_space(input_mode, url, keywords):
    url_or_keywords = url if input_mode == "URL" else keywords
    keywords, ad_copies, image_path, error = main_workflow(input_mode, url_or_keywords)
    ad_previews = ""
    if ad_copies:
        for platform, ad in ad_copies.items():
            ad_previews += platform_html(platform, ad)
    return (
        keywords,
        ad_previews,
        image_path,
        "generated_ads.txt" if ad_copies else None,
        error
    )

with gr.Blocks() as demo:
    gr.Markdown("# 🚗 Auto Ad Generator\nPaste a car listing URL **or** enter your own keywords, then preview AI-generated ads for each social media platform, plus an auto-generated image!")
    input_mode = gr.Radio(["URL", "Keywords"], value="URL", label="Input Type")
    url_input = gr.Textbox(label="Listing URL", placeholder="https://www.cars.com/listing/...", visible=True)
    kw_input = gr.Textbox(label="Manual Keywords (comma separated)", placeholder="e.g. BMW, used car, sunroof", visible=False)
    submit_btn = gr.Button("Generate Ads")

    gr.Markdown("## Keywords")
    kw_out = gr.JSON(label="Extracted/Provided Keywords")

    gr.Markdown("## Ad Copy Previews")
    ad_out = gr.HTML(label="Ad Copy Preview")   # Now HTML, not Markdown

    gr.Markdown("## Generated Ad Image")
    img_out = gr.Image(label="Generated Ad Image", type="filepath")

    gr.Markdown("## Download Ad Copies")
    file_out = gr.File(label="Download TXT")

    err_out = gr.Textbox(label="Errors", interactive=False)

    def show_hide_fields(choice):
        return (
            gr.update(visible=choice == "URL"),
            gr.update(visible=choice == "Keywords"),
        )

    input_mode.change(show_hide_fields, input_mode, [url_input, kw_input])

    submit_btn.click(
        run_space,
        inputs=[input_mode, url_input, kw_input],
        outputs=[kw_out, ad_out, img_out, file_out, err_out]
    )

demo.launch()