Dockerfile

# Agri-Intelligence platform — container image.
# Works for: local `docker compose up`, CI, and Hugging Face Spaces (Docker SDK).
# HF Spaces reads `app_port: 8000` from README.md and runs this image as uid 1000.
FROM python:3.11-slim

# System libs needed by lightgbm / shap / numpy / torch wheels.
RUN apt-get update && apt-get install -y --no-install-recommends \
    libgomp1 curl \
    && rm -rf /var/lib/apt/lists/*

# Non-root user (required/recommended by Hugging Face Spaces) + writable caches.
RUN useradd -m -u 1000 user
ENV HOME=/home/user \
    PATH=/home/user/.local/bin:$PATH \
    MPLCONFIGDIR=/home/user/.cache/matplotlib \
    TORCH_HOME=/home/user/.cache/torch \
    PYTHONUNBUFFERED=1 \
    # Single-threaded OpenMP: torch + lightgbm + sklearn each bundle an OpenMP
    # runtime; this avoids a native init-order crash when loaded in one process,
    # and is correct for single-request inference workers.
    OMP_NUM_THREADS=1 MKL_NUM_THREADS=1 KMP_DUPLICATE_LIB_OK=TRUE

WORKDIR /app

# CPU-only PyTorch FIRST — avoids the ~2 GB CUDA download and keeps the image
# small enough for free hosts. Pinned to the versions the models were saved with.
RUN pip install --no-cache-dir \
    torch==2.2.1 torchvision==0.17.1 \
    --index-url https://download.pytorch.org/whl/cpu

# Remaining deps (torch/torchvision already satisfied above, so not re-downloaded).
COPY api/requirements.txt ./api/requirements.txt
RUN pip install --no-cache-dir -r api/requirements.txt

# Application code + model/data artifacts the pipelines load at runtime.
# (.dockerignore excludes only training images + the raw price xlsx.)
COPY api ./api
COPY models ./models
COPY data ./data
COPY vision ./vision
COPY yield_pred ./yield_pred
COPY weather ./weather
COPY irrigation ./irrigation
COPY fertilizer ./fertilizer
COPY profit ./profit
COPY carbon ./carbon
COPY market ./market
COPY schemes ./schemes
COPY assistant ./assistant
COPY dashboards ./dashboards
COPY twin ./twin
COPY scripts ./scripts
COPY conftest.py pytest.ini ./

# Hand ownership to the non-root user (covers app code + cache dirs).
RUN mkdir -p /home/user/.cache && chown -R user:user /app /home/user
USER user

# The assistant needs an LLM key at runtime — pass it in, never bake it in:
#   docker run -e NVIDIA_API_KEY=... -p 8000:8000 agri-intelligence
#   (on HF Spaces, set NVIDIA_API_KEY as a Space secret)

EXPOSE 8000

HEALTHCHECK --interval=30s --timeout=5s --start-period=40s \
    CMD curl -fsS http://localhost:8000/health || exit 1

CMD ["uvicorn", "api.main:app", "--host", "0.0.0.0", "--port", "8000"]