Dockerfile

FROM python:3.10-slim

# HF Spaces requires a non-root user with uid 1000
RUN useradd -m -u 1000 user

WORKDIR /app

# Install CPU-only PyTorch before requirements.txt so sentence-transformers
# doesn't pull the full CUDA build (~2 GB) as a transitive dependency.
RUN pip install --no-cache-dir \
    torch \
    --index-url https://download.pytorch.org/whl/cpu

# Copy and install remaining dependencies as a separate layer so source-only
# changes don't invalidate the (slow) pip install cache.
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Pre-download the embedding model into the image so first-request latency
# on the Space is not penalised by a ~90 MB download at runtime.
RUN python -c "from sentence_transformers import SentenceTransformer; SentenceTransformer('all-MiniLM-L6-v2')"

# Copy source last — this layer changes most often
COPY --chown=user:user . .

USER user

ENV PYTHONUNBUFFERED=1 \
    # Bind to all interfaces so HF Spaces proxy can reach the server
    GRADIO_SERVER_NAME=0.0.0.0 \
    GRADIO_SERVER_PORT=7860 \
    # Keep HuggingFace model cache inside the container's home
    HF_HOME=/home/user/.cache/huggingface

EXPOSE 7860

CMD ["python", "app_gradio.py"]