Dockerfile

# Hugging Face Spaces (Docker SDK) image for the VLPS Gradio demo.
# Models are pre-downloaded at build time (warmup.py) so first use is fast.
FROM python:3.12-slim

# System libs needed by OpenCV / EasyOCR.
RUN apt-get update && apt-get install -y --no-install-recommends \
        libgl1 libglib2.0-0 && \
    rm -rf /var/lib/apt/lists/*

# HF Spaces runs as a non-root user (uid 1000); keep caches in a writable home.
RUN useradd -m -u 1000 user
USER user
ENV HOME=/home/user \
    PATH=/home/user/.local/bin:$PATH \
    HF_HOME=/home/user/.cache/huggingface \
    KMP_DUPLICATE_LIB_OK=TRUE \
    VLPS_IMAGE_URLS=1 \
    GRADIO_ANALYTICS_ENABLED=False

WORKDIR /home/user/app

# Install Python deps first (better layer caching).
# CPU-only torch/torchvision keeps the image small (the CUDA build is several GB and
# overflows the free-tier build); the LVLM runs remotely, so no GPU is needed here.
COPY --chown=user requirements-space.txt .
RUN pip install --no-cache-dir --upgrade pip && \
    pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/cpu && \
    pip install --no-cache-dir -r requirements-space.txt

# Bake model weights into the image.
COPY --chown=user warmup.py .
RUN python warmup.py

# App code.
COPY --chown=user src ./src
COPY --chown=user web ./web
COPY --chown=user core.py server.py ./

EXPOSE 7860
CMD ["uvicorn", "server:app", "--host", "0.0.0.0", "--port", "7860"]