FROM python:3.12-slim

ARG TORCH_VERSION=2.14.1

# Install the CPU wheel explicitly before Laya so pip does not select the much
# larger CUDA build. Model weights remain runtime-only in the mounted cache.
RUN python -m pip install --no-cache-dir \
        "torch==${TORCH_VERSION}" \
        --index-url https://download.pytorch.org/whl/cpu \
    && python -m pip install --no-cache-dir "laya[serve]>=0.3.22,<0.4" \
    && python -m pip check \
    && python -c "import torch; assert torch.version.cuda is None, torch.version.cuda"

ENV HF_HOME=/models \
    LAYA_HOST=0.0.0.0 \
    LAYA_PORT=8000 \
    LAYA_PRELOAD=0 \
    LAYA_DEVICE=cpu

VOLUME ["/models"]
EXPOSE 8000

CMD ["laya-serve"]