That's awesome! Are u handling the image recognition separately or is GPT-5 dialed into that too? Sounds like a complex pipeline. Would love to see some behind the scene action, maybe a blog post or a GitHub repo?