import gradio as gr
import pandas as pd
import requests
import io
import base64
from PIL import Image, ImageOps, ImageEnhance

DEEPSIGHT_API_URL = "https://api.deepseek.com/v2/chat/completions"
DEEPSIGHT_API_KEY = "YOUR_API_KEY"

SYSTEM_PROMPT = open("system_prompt.txt").read()

def call_deepsight(product_data):
    """Send product row to DeepSight v2 for structured generation."""
    payload = {
        "model": "deepseek-chat",
        "messages": [
            {"role": "system", "content": SYSTEM_PROMPT},
            {"role": "user", "content": product_data}
        ],
        "temperature": 0.2
    }

    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {DEEPSIGHT_API_KEY}"
    }

    response = requests.post(DEEPSIGHT_API_URL, json=payload, headers=headers)
    return response.json()["choices"][0]["message"]["content"]


def download_image(url):
    try:
        img_bytes = requests.get(url, timeout=10).content
        return Image.open(io.BytesIO(img_bytes)).convert("RGBA")
    except:
        return None


def remove_bg(img):
    """Simple white background fallback if rembg is not installed."""
    # Optional: integrate rembg here if available
    bg = Image.new("RGBA", img.size, "WHITE")
    bg.paste(img, mask=img.split()[3])
    return bg


def apply_watermark(img, watermark_path="watermark.png"):
    try:
        wm = Image.open(watermark_path).convert("RGBA")
        wm = wm.resize((int(img.size[0] * 0.3), int(img.size[1] * 0.3)))
        img.paste(wm, (img.size[0]-wm.size[0]-10, img.size[1]-wm.size[1]-10), wm)
    except:
        pass
    return img


def enhance_image(img):
    img = ImageEnhance.Sharpness(img).enhance(1.4)
    img = ImageEnhance.Brightness(img).enhance(1.05)
    return img


def process_csv(file):
    df = pd.read_csv(file)
    output_rows = []
    output_images = []

    for idx, row in df.iterrows():
        title = str(row.get("post_title", "")).strip()
        short = str(row.get("post_excerpt", "")).strip()
        long = str(row.get("post_content", "")).strip()
        image_link = row.get("image_link", "")

        # Create a combined row prompt
        row_prompt = f"""
        Product Title: {title}
        Short Description: {short}
        Long Description: {long}
        """

        # Call DeepSight for content generation
        ai_output = call_deepsight(row_prompt)
        ai_dict = eval(ai_output)  # Expecting strict JSON from DeepSight

        # Process image
        final_image_path = None
        if isinstance(image_link, str) and image_link.startswith("http"):
            img = download_image(image_link)
            if img:
                img = ImageOps.contain(img, (800, 800))
                img = remove_bg(img)
                img = enhance_image(img)
                img = apply_watermark(img)

                save_path = f"processed_{idx}.png"
                img.save(save_path, "PNG")
                final_image_path = save_path

        output_rows.append({
            "seo_title": ai_dict["seo_title"],
            "short_description": ai_dict["short_description"],
            "long_description": ai_dict["long_description"],
            "processed_image": final_image_path
        })

    result_df = pd.DataFrame(output_rows)
    result_path = "output.csv"
    result_df.to_csv(result_path, index=False)

    return result_path


# Gradio Interface
with gr.Blocks() as demo:
    gr.Markdown("# 🚀 WooCommerce Product Optimizer (DeepSight v2)")
    gr.Markdown("Upload your CSV and let DeepSight clean titles, descriptions, and images.")

    csv_input = gr.File(label="Upload WooCommerce CSV")
    output_csv = gr.File(label="Download Optimized CSV")

    run_btn = gr.Button("Process CSV")

    run_btn.click(process_csv, inputs=csv_input, outputs=output_csv)

demo.launch()