app.py

import json
import gradio as gr
from sentence_transformers import SentenceTransformer, util
import torch
import requests

# Load sentence transformer model
model = SentenceTransformer("all-MiniLM-L6-v2")

# Load Louvre data from Hugging Face
url = "https://huggingface.co/datasets/Agathe1489/Louvre-ARK/resolve/main/louvre_data.jsonl"
response = requests.get(url)
artworks = [json.loads(line) for line in response.text.strip().split("\n")]

# Precompute embeddings for title + description
texts = [
    f"{art.get('title', '')}. {art.get('description', '')}"
    for art in artworks
]
embeddings = model.encode(texts, convert_to_tensor=True)

# Search function
def search_artwork(query):
    query_embedding = model.encode(query, convert_to_tensor=True)
    scores = util.cos_sim(query_embedding, embeddings)[0]
    best_idx = torch.argmax(scores).item()
    result = artworks[best_idx]
    
    image_url = result.get("image", [{}])[0].get("urlImage", "")
    title = result.get("title", "Untitled")
    description = result.get("description", "No description available.")
    creator = (
        result.get("creator", [{}])[0].get("label", "Unknown artist")
        if result.get("creator") else "Unknown artist"
    )
    page_url = result.get("url", "")
    
    display_text = f"**{title}** by *{creator}*\n\n{description}\n\n[View on Louvre]({page_url})"
    
    return image_url, display_text

# Gradio UI
iface = gr.Interface(
    fn=search_artwork,
    inputs=gr.Textbox(placeholder="Describe a scene, object, theme, or keyword..."),
    outputs=[
        gr.Image(label="Artwork Image"),
        gr.Markdown(label="Details"),
    ],
    title="🎨 Louvre Artwork Finder",
    description="Enter a theme or description and discover related artwork from the Louvre Museum collection."
)

# Launch
if __name__ == "__main__":
    iface.launch()