image_agent.py

import os
import PIL
import requests
from io import BytesIO
from PIL import Image
from smolagents import CodeAgent, OpenAIServerModel, InferenceClientModel
from huggingface_hub import login

image_urls = [
    "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcTcB9-ieIzm2uhtTkMkV5F50xbL4itvh3j8bg&s",  # Joker image
    "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcR0SWQC74KuQWrJA33pJqUmOuDx62K7p1uEcQ&s"  # Joker image
]

stored_images = []

for url in image_urls:
    headers = {
        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36"
    }
    response = requests.get(url, headers)
    try:
        image = Image.open(BytesIO(response.content)).convert("RGB")
        stored_images.append(image)
    except PIL.UnidentifiedImageError:
        print("⚠️ The downloaded file is not a valid image.")
        # Optional: save the raw bytes for inspection
        with open("debug_image.bin", "wb") as f:
            f.write(response.content)
    # image = Image.open(BytesIO(response.content)).convert("RGB")


def init_stored_image_agent() -> CodeAgent:
    # Now based on the images indicate if it's wonder women or joker.
    model = OpenAIServerModel(model_id="gpt-4o")
    stored_image_agent = CodeAgent(
        tools=[],
        model=model,
        max_steps=5,
        verbosity_level=2
    )
    return stored_image_agent