tools/vision_tools.py

import base64

import requests
from config import IP_WINDOWS
import ollama
from smolagents import tool


@tool
def analyze_image(image_path: str, question: str) -> str:
    """ Analyze an image using a local vision model and answer a question about it.
    Use this tool when you need to extract information from a jpg/png file.

    Args:
        image_path: The local path to the image file (e.g. 'images/grafico.png').
        question: The specific question about what to look for in the image (e.g. 'What value is the red bar?').

    Returns:
        str: The answer to the question, based on the image.

    Example:
        >>> result = analise_image("images/grafico.png", "What value is the red bar?")
    """
    # url = f"http://{IP_WINDOWS}:11434/api/generate"
    
    try:
        # Codificamos la imagen a base64 para enviarla por red
        with open(image_path, "rb") as image_file:
            img_str = base64.b64encode(image_file.read()).decode('utf-8')

        # payload = {
        #    "model": "llava", # Asegúrate de tener este modelo en Windows
        #    "prompt": question,
        #    "images": [img_str],
        #    "stream": False
        #}
        
        # response = requests.post(url, json=payload)
        response = ollama.chat(model='llava',
                               messages=[{'role': 'user',
                                          'content': question,
                                          'images': image_file}])
        return response['messages']['content']

    except Exception as e:
        return f"Error conectando con Windows: {str(e)}"