from typing import Any, Optional from smolagents.tools import Tool from PIL import Image # Add this import import os # Add this import import datetime # Add this import class FinalAnswerTool(Tool): name = "final_answer" description = "Provides a final answer to the given problem. Can include a textual response and/or a path to a generated image." # IMPORTANT: Update the inputs to explicitly allow 'text' and 'image_path' inputs = { 'text': {'type': 'str', 'description': 'The final textual answer to the problem'}, 'image_path': {'type': 'str', 'description': 'Optional path to a generated image file (e.g., .webp, .png, .jpg) to be displayed.'} } output_type = "str" # The output will be a Markdown string def forward(self, text: Optional[str] = None, image_path: Optional[str] = None) -> str: """ Processes the final answer, potentially embedding an image from a given path. """ output_content = [] if image_path and os.path.exists(image_path): # Gradio should typically handle .webp directly, but converting to PNG # is a robust fallback for broader compatibility or specific display quirks. image_display_path = image_path if image_path.lower().endswith(".webp"): try: img = Image.open(image_path) # Create a temporary filename for the PNG # Use a more robust temporary path strategy temp_dir = os.path.dirname(image_path) if os.path.dirname(image_path) else "/tmp" os.makedirs(temp_dir, exist_ok=True) # Ensure temp_dir exists base_name = os.path.basename(image_path) name_without_ext = os.path.splitext(base_name)[0] # Add microsecond precision to avoid collisions if multiple images are generated rapidly temp_png_path = os.path.join( temp_dir, f"{name_without_ext}_{datetime.datetime.now().strftime('%Y%m%d%H%M%S%f')}.png" ) img.save(temp_png_path, "PNG") image_display_path = temp_png_path print(f"[DEBUG] Converted {image_path} to {temp_png_path} for display.") except Exception as e: print(f"[ERROR] Failed to convert image {image_path} to PNG: {e}. Attempting to use original path.") # Fallback to original path if conversion fails # Embed the image using Markdown syntax, which Gradio will render output_content.append(f"![Generated Image]({image_display_path})") if text: output_content.append(text) if not output_content: return "No final answer provided." return "\n\n".join(output_content) def __init__(self, *args, **kwargs): # Important: Call the superclass __init__ to properly initialize the Tool super().__init__(*args, **kwargs) self.is_initialized = False # You can keep this if it's used elsewhere