🪄 ImageSorcery MCP

from typing import Annotated from fastmcp import FastMCP from pydantic import Field # Import the central logger from imagesorcery_mcp.logging_config import logger def register_prompt(mcp: FastMCP): @mcp.prompt(name="remove-background") def remove_background( image_path: Annotated[ str, Field(description="Full path to the input image (must be a full path)") ], target_objects: Annotated[ str, Field( description="Description of the objects to keep in the foreground (e.g., 'person', 'car and person', 'main subject')" ), ] = "", output_path: Annotated[ str, Field( description="Full path for the output image with background removed (optional, will auto-generate if not provided)" ), ] = "", ) -> str: """ Guides the AI through a comprehensive background removal workflow. This prompt provides a step-by-step approach to remove backgrounds from images using object detection and masking tools. It's designed to work with the ImageSorcery MCP server's detect and fill tools. The workflow includes: 1. Object detection to identify the target object 2. Mask generation for precise selection 3. Background removal using fill operations 4. Optional refinement steps Args: image_path: Full path to the input image target_objects: Description of what to keep (default: empty for auto-detection) output_path: Where to save the result (auto-generated if empty) Returns: A detailed prompt guiding the AI through the background removal process """ logger.info(f"Remove background prompt requested for image: {image_path}") logger.debug(f"Target objects: {target_objects}, Output path: {output_path}") # Generate output path if not provided if not output_path: if image_path.lower().endswith(('.png', '.jpg', '.jpeg')): base_path = image_path.rsplit('.', 1)[0] output_path = f"{base_path}_no_background.png" else: output_path = f"{image_path}_no_background.png" # Build the prompt based on whether target_objects is specified if target_objects: prompt = f"""I need to remove the background from an image while preserving the {target_objects}. Please follow this step-by-step workflow: **Step 1: Find Target Objects** Use the `find` tool to locate the specific objects: - Call `find` on '{image_path}' with: - description: "{target_objects}" - confidence: 0.3 (lower threshold for better recall) - return_geometry: true - geometry_format: "mask" - This will use text-based object identification to locate the {target_objects} **Step 2: Remove Background** Use the `fill` tool to remove the background: - Call `fill` on '{image_path}' with: - areas: Use mask files from find - color: null - output_path: '{output_path}' **Step 3: Clean Up** - Remove the temporary mask files created during the process **Important Notes:** - Save the final result as a PNG file to preserve transparency Please execute this workflow step by step.""" else: prompt = f"""I need to remove the background from an image. Please follow this step-by-step workflow: **Step 1: Detect Objects** Use the `detect` tool to identify objects in the image: - Call `detect` on '{image_path}' with: - confidence: 0.5 (to catch more objects) - return_geometry: true - geometry_format: "mask" - Review the detected objects and identify the main subjects to preserve **Step 3: Remove Background** Use the `fill` tool to remove the background: - Call `fill` on '{image_path}' with: - areas: Use mask files from detect - color: null - output_path: '{output_path}' **Step 4: Clean Up** - Remove the temporary mask files created during the process **Important Notes:** - Save the final result as a PNG file to preserve transparency Please execute this workflow step by step.""" logger.info(f"Generated remove background prompt for targets: {target_objects or 'auto-detect'}") return prompt