FROM python:3.11-slim AS base

ENV PYTHONDONTWRITEBYTECODE=1 \
    PYTHONUNBUFFERED=1 \
    PIP_NO_CACHE_DIR=1 \
    TOKENIZERS_PARALLELISM=false \
    MODEL_OUTPUT_DIR=/app/arabguard_model \
    CHECKPOINT_DIR=/tmp/arabguard_checkpoints \
    DASHBOARD_DATA_DIR=/app/dashboard_data

WORKDIR /app

COPY requirements.txt ./requirements.txt
RUN pip install --upgrade pip && pip install -r requirements.txt

FROM base AS trainer

# Keep downloads and checkpoints in this disposable build stage.
ENV HF_HOME=/tmp/huggingface

# Only source code is copied from Git. The dataset and base model are downloaded,
# trained, and saved into the image while Hugging Face builds the Space.
COPY train_model.py ./
RUN python train_model.py

FROM base AS runtime

# Copy only the trained artifacts, not the dataset cache or checkpoints.
COPY --from=trainer --chown=1000:1000 /app/arabguard_model ./arabguard_model
COPY --from=trainer --chown=1000:1000 /app/dashboard_data ./dashboard_data
COPY app.py ./

ENV HOME=/home/user \
    HF_HOME=/home/user/.cache/huggingface

RUN useradd --create-home --uid 1000 user \
    && chown -R user:user /app

USER user

EXPOSE 7860

CMD ["streamlit", "run", "app.py", "--server.address=0.0.0.0", "--server.port=7860", "--server.headless=true"]