requirements.txt

fastapi>=0.111
uvicorn[standard]>=0.30
gradio>=4.38
requests>=2.31

# Match vLLM ↔ Torch versions
vllm==0.5.5

# Use CUDA 12.1 wheels for torch 2.4.0
--extra-index-url https://download.pytorch.org/whl/cu121
torch==2.4.0

transformers>=4.44
accelerate>=0.30