app.py

   import gradio as gr
   import torch
   import os
   import sys
   import tempfile
   import shutil
   import subprocess
   # from huggingface_hub import HfApi, snapshot_download # For future model management if needed
   # import spaces # For @spaces.GPU decorator if you add it

   # --- Configuration ---
   # Path to the cloned UniRig repository directory within the Space
   UNIRIG_REPO_DIR = os.path.join(os.path.dirname(__file__), "UniRig")

   if not os.path.isdir(UNIRIG_REPO_DIR):
       print(f"ERROR: UniRig repository not found at {UNIRIG_REPO_DIR}. Please clone it there.")
       # Consider raising an error or displaying it in the UI if UniRig is critical for startup

   DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
   print(f"Using device: {DEVICE}")
   if DEVICE.type == 'cuda':
       print(f"CUDA Device Name: {torch.cuda.get_device_name(0)}")
       print(f"CUDA Version: {torch.version.cuda}")
   else:
       print("Warning: CUDA not available or not detected by PyTorch. UniRig performance will be severely impacted.")

   def run_unirig_command(command_args, step_name):
       """Helper function to run UniRig commands using subprocess."""
       python_exe = sys.executable
       # Ensure the command starts with the python executable and '-m' for module execution
       cmd = [python_exe, "-m"] + command_args
       
       print(f"Running {step_name}: {' '.join(cmd)}")
       
       process_env = os.environ.copy()
       # It's generally better to ensure UniRig's internal scripts handle PYTHONPATH if needed,
       # or that it's installed in a way that `python -m` works correctly from its root.
       # Setting cwd=UNIRIG_REPO_DIR is often the key for Hydra.

       try:
           # Execute the command from the UniRig directory for Hydra to find configs
           result = subprocess.run(cmd, cwd=UNIRIG_REPO_DIR, capture_output=True, text=True, check=True, env=process_env)
           print(f"{step_name} STDOUT:\n{result.stdout}")
           if result.stderr:
               print(f"{step_name} STDERR (non-fatal or warnings):\n{result.stderr}")
       except subprocess.CalledProcessError as e:
           print(f"ERROR during {step_name}:")
           print(f"Command: {' '.join(e.cmd)}")
           print(f"Return code: {e.returncode}")
           print(f"Stdout: {e.stdout}")
           print(f"Stderr: {e.stderr}")
           # Provide a more user-friendly error, potentially masking long tracebacks
           error_summary = e.stderr.splitlines()[-5:] # Last 5 lines of stderr
           raise gr.Error(f"Error in UniRig {step_name}. Details: {' '.join(error_summary)}")
       except FileNotFoundError:
           print(f"ERROR: Could not find executable or script for {step_name}. Is UniRig cloned correctly in {UNIRIG_REPO_DIR} and Python environment setup?")
           raise gr.Error(f"Setup error for UniRig {step_name}. Check server logs and UniRig directory structure.")
       except Exception as e_general:
           print(f"An unexpected Python exception occurred in run_unirig_command for {step_name}: {e_general}")
           raise gr.Error(f"Unexpected Python error during {step_name}: {str(e_general)[:500]}")


   # If you are using @spaces.GPU, you would import it:
   # import spaces
   # @spaces.GPU # You can specify type like @spaces.GPU(type="t4") or count
   def rig_glb_mesh_multistep(input_glb_file_obj):
       """
       Takes an input GLB file object (from gr.File with type="filepath"),
       rigs it using the new UniRig multi-step process,
       and returns the path to the final rigged GLB file.
       """
       if not os.path.isdir(UNIRIG_REPO_DIR):
            raise gr.Error(f"UniRig repository not found at {UNIRIG_REPO_DIR}. Cannot proceed. Please check Space setup.")

       if input_glb_file_obj is None:
           # This case should ideally be handled by Gradio's input validation if `allow_none=False` (default)
           raise gr.Error("No input file provided. Please upload a .glb mesh.")

       # When type="filepath", input_glb_file_obj is the path string directly
       input_glb_path = input_glb_file_obj
       print(f"Input GLB path received: {input_glb_path}")

       # Create a dedicated temporary directory for all intermediate and final files
       processing_temp_dir = tempfile.mkdtemp(prefix="unirig_processing_")
       print(f"Using temporary processing directory: {processing_temp_dir}")

       try:
           base_name = os.path.splitext(os.path.basename(input_glb_path))[0]
           
           # Step 1: Skeleton Prediction
           temp_skeleton_path = os.path.join(processing_temp_dir, f"{base_name}_skeleton.fbx")
           print("Step 1: Predicting Skeleton...")
           run_unirig_command([
               "unirig.predict_skeleton",
               f"input.path={os.path.abspath(input_glb_path)}", # Use absolute path for robustness
               f"output.path={os.path.abspath(temp_skeleton_path)}",
               # f"device={str(DEVICE)}" # If UniRig's script accepts this override and handles it
           ], "Skeleton Prediction")
           if not os.path.exists(temp_skeleton_path):
               raise gr.Error("Skeleton prediction failed to produce an output file. Check logs for UniRig errors.")

           # Step 2: Skinning Weight Prediction
           temp_skin_path = os.path.join(processing_temp_dir, f"{base_name}_skin.fbx")
           print("Step 2: Predicting Skinning Weights...")
           run_unirig_command([
               "unirig.predict_skin",
               f"input.skeleton_path={os.path.abspath(temp_skeleton_path)}",
               f"input.source_mesh_path={os.path.abspath(input_glb_path)}",
               f"output.path={os.path.abspath(temp_skin_path)}",
           ], "Skinning Prediction")
           if not os.path.exists(temp_skin_path):
               raise gr.Error("Skinning prediction failed to produce an output file. Check logs for UniRig errors.")

           # Step 3: Merge Skeleton/Skin with Original Mesh
           final_rigged_glb_path = os.path.join(processing_temp_dir, f"{base_name}_rigged_final.glb")
           print("Step 3: Merging Results...")
           run_unirig_command([
               "unirig.merge_skeleton_skin",
               f"input.source_rig_path={os.path.abspath(temp_skin_path)}",
               f"input.target_mesh_path={os.path.abspath(input_glb_path)}",
               f"output.path={os.path.abspath(final_rigged_glb_path)}",
           ], "Merging")
           if not os.path.exists(final_rigged_glb_path):
               raise gr.Error("Merging process failed to produce the final rigged GLB file. Check logs for UniRig errors.")

           # final_rigged_glb_path is in processing_temp_dir.
           # Gradio's gr.Model3D output component will handle serving this file.
           return final_rigged_glb_path

       except gr.Error: # Re-raise Gradio errors directly
           if os.path.exists(processing_temp_dir): # Clean up on known Gradio error
               shutil.rmtree(processing_temp_dir)
               print(f"Cleaned up temporary directory: {processing_temp_dir}")
           raise
       except Exception as e:
           print(f"An unexpected error occurred in rig_glb_mesh_multistep: {e}")
           if os.path.exists(processing_temp_dir): # Clean up on unexpected error
               shutil.rmtree(processing_temp_dir)
               print(f"Cleaned up temporary directory: {processing_temp_dir}")
           raise gr.Error(f"An unexpected error occurred during processing: {str(e)[:500]}")


   # --- Gradio Interface ---
   theme = gr.themes.Soft(
       primary_hue=gr.themes.colors.sky,
       secondary_hue=gr.themes.colors.blue,
       neutral_hue=gr.themes.colors.slate,
       font=[gr.themes.GoogleFont("Inter"), "ui-sans-serif", "system-ui", "sans-serif"],
   )

   # Ensure UNIRIG_REPO_DIR check happens before interface is built if it's critical
   if not os.path.isdir(UNIRIG_REPO_DIR) and __name__ == "__main__": # Check only if running as main script
       print(f"CRITICAL STARTUP ERROR: UniRig repository not found at {UNIRIG_REPO_DIR}. The application will not work.")

   # Define the interface
   # Note: The @spaces.GPU decorator would go above the function `rig_glb_mesh_multistep`
   iface = gr.Interface(
       fn=rig_glb_mesh_multistep, 
       inputs=gr.File(
           label="Upload .glb Mesh File",
           type="filepath"  # Corrected type for Gradio 5.x.x
       ),
       outputs=gr.Model3D(
           label="Rigged 3D Model (.glb)",
           clear_color=[0.8, 0.8, 0.8, 1.0],
       ),
       title="UniRig Auto-Rigger (Python 3.11 / PyTorch 2.3+)",
       description=(
           "Upload a 3D mesh in `.glb` format. This application uses the latest UniRig to automatically rig the mesh.\n"
           "The process involves: 1. Skeleton Prediction, 2. Skinning Weight Prediction, 3. Merging.\n"
           "This may take several minutes. Ensure your GLB has clean geometry.\n"
           f"Running on: {str(DEVICE).upper()}. UniRig repo expected at: '{os.path.basename(UNIRIG_REPO_DIR)}'.\n"
           f"UniRig Source: https://github.com/VAST-AI-Research/UniRig"
       ),
       cache_examples=False,
       theme=theme,
       allow_flagging="never"
   )

   if __name__ == "__main__":
       if not os.path.isdir(UNIRIG_REPO_DIR):
           print(f"CRITICAL: UniRig repository not found at {UNIRIG_REPO_DIR}. Ensure it's cloned in the Space's root.")
       
       iface.launch()