Dockerfile

FROM nvidia/cuda:12.0.0-cudnn8-devel-ubuntu22.04

# 1) Configure HF cache locations up front
ENV HF_HOME="/home/user/.cache/huggingface" \
    HF_HUB_CACHE="/home/user/.cache/huggingface/hub" \
    TRANSFORMERS_CACHE="/home/user/.cache/huggingface/transformers"

# 2) Create non-root user
RUN useradd -m -u 1000 user

# 3) Install Python & system libs
RUN apt-get update && \
    apt-get install -y --no-install-recommends \
       python3 python3-pip python3-dev \
       build-essential git libpoppler-cpp-dev poppler-utils libmagic-dev && \
    rm -rf /var/lib/apt/lists/*

WORKDIR /home/user/app
COPY --chown=user requirements.txt .

# 4) Install Python deps
RUN pip install --no-cache-dir torch torchvision --extra-index-url https://download.pytorch.org/whl/cu118
RUN pip install --no-cache-dir huggingface_hub
RUN pip install --no-cache-dir -r requirements.txt

# 5) Prep model cache dir and HF cache, set ownership
RUN mkdir -p /home/user/app/model_cache $HF_HUB_CACHE && \
    chown -R user:user /home/user/app /home/user/.cache/huggingface

# 6) **Switch to non-root user** and pre-download your model
USER user
RUN python3 - <<EOF
from huggingface_hub import snapshot_download
import os
snapshot_download(
    repo_id="numind/NuExtract-1.5",         # <-- your HF model ID
    local_dir="/home/user/app/model_cache", # <-- where your FastAPI will load from
    cache_dir=os.getenv("HF_HUB_CACHE"),    # <-- HF’s own cache
    resume_download=True
)
EOF

# 7) Copy your code and launch
COPY --chown=user . .
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]