# syntax=docker/dockerfile:1
# ---------------------------------------------------------------------------
# Hugging Face Spaces image for the Document-Agent backend (free CPU tier).
#
# HF Spaces (Docker SDK) builds the Dockerfile at the repo ROOT with the repo
# root as the build context, so source paths are prefixed with backend/. HF
# runs the container as UID 1000 and routes traffic to the port declared as
# `app_port:` in README.md (8000 here).
#
# Free Spaces have NO persistent disk, so DATA_DIR is an ephemeral, user-owned
# directory: uploads / SQLite / the vector store reset on restart or rebuild.
# That is fine for a demo. The ML models are baked in at build time (same as
# the production image) so cold start is fast and there are no runtime
# downloads.
# ---------------------------------------------------------------------------
FROM python:3.12-slim

ENV PYTHONUNBUFFERED=1 \
    PYTHONDONTWRITEBYTECODE=1 \
    PIP_NO_CACHE_DIR=1 \
    PIP_DISABLE_PIP_VERSION_CHECK=1 \
    HF_HUB_DISABLE_SYMLINKS_WARNING=1 \
    # Fixed, shared cache locations so the non-root runtime user finds the
    # models baked in during build.
    HF_HOME=/opt/models/hf \
    EASYOCR_MODULE_PATH=/opt/models/easyocr \
    TORCH_HOME=/opt/models/torch \
    DATA_DIR=/data

# Native libs needed by docling / opencv / easyocr.
RUN apt-get update && apt-get install -y --no-install-recommends \
        libgl1 \
        libglib2.0-0 \
        libgomp1 \
    && rm -rf /var/lib/apt/lists/*

WORKDIR /app

# Install CPU-only torch/torchvision FIRST so docling's torch dependency
# resolves to the lightweight CPU wheels instead of multi-GB CUDA builds.
RUN pip install --index-url https://download.pytorch.org/whl/cpu \
        torch torchvision

COPY backend/requirements.txt .
RUN pip install -r requirements.txt

# Bake the models into the image.
COPY backend/scripts/prefetch_models.py scripts/prefetch_models.py
RUN mkdir -p /opt/models && python scripts/prefetch_models.py

# Backend app source (repo root is the build context on HF Spaces).
COPY backend/ .

# HF Spaces runs the container as UID 1000. Make the data dir and the baked
# model/app trees writable/readable by that user.
RUN useradd -m -u 1000 user \
    && mkdir -p /data \
    && chown -R user:user /data /app /opt/models
USER user
ENV HOME=/home/user

EXPOSE 8000

# Honour $PORT if a host injects it; default 8000 matches app_port in README.
HEALTHCHECK --interval=30s --timeout=5s --start-period=60s --retries=3 \
    CMD python -c "import os,urllib.request,sys; sys.exit(0 if urllib.request.urlopen('http://localhost:%s/api/health' % os.environ.get('PORT','8000')).status==200 else 1)"

CMD ["sh", "-c", "uvicorn app.main:app --host 0.0.0.0 --port ${PORT:-8000}"]