requirements.txt

# ALICE Dependencies
# NOTE: Python package versions match SAM's bundled Python environment
# to ensure generate_image_diffusers.py works identically in both projects
#
# IMPORTANT: torch and torchvision are NOT listed here because they require
# platform-specific installation (ROCm on AMD, CUDA on NVIDIA, CPU-only, etc.)
# Install scripts handle PyTorch installation separately.

# FastAPI and server
fastapi==0.104.1
uvicorn[standard]==0.24.0
pydantic==2.12.5
pydantic-settings==2.1.0

# Diffusion models - MATCHED TO SAM
# Note: SAM uses diffusers 0.36.0.dev0, we use latest stable
# Updated to git version for Qwen-Image pipeline support
git+https://github.com/huggingface/diffusers
transformers==4.57.3
accelerate==1.12.0
safetensors==0.7.0

# Prompt weighting and long prompts - MATCHED TO SAM
compel==2.3.1

# Image processing - MATCHED TO SAM
pillow==12.0.0

# Configuration - MATCHED TO SAM
pyyaml==6.0.3

# Async file handling
aiofiles==23.2.1

# Async HTTP client (for download manager)
aiohttp==3.9.1

# Multipart form data (for file uploads)
python-multipart==0.0.6

# System monitoring - MATCHED TO SAM
psutil==7.1.3