Dockerfile

# QModel 7 - Islamic RAG API
# =============================
# Dockerfile for QModel API
# Supports both Ollama and HuggingFace backends via .env configuration
#
# Build: docker build -t qmodel .
# Run: docker run -p 8000:8000 --env-file .env qmodel

# ── Builder stage: compile deps, keep toolchain out of the final image ──
FROM python:3.11-slim AS builder

ENV PYTHONDONTWRITEBYTECODE=1 \
    PYTHONUNBUFFERED=1 \
    PIP_NO_CACHE_DIR=1

# - build-essential/cmake: for compiling llama-cpp-python and other native deps
# - libopenblas-dev: for numerical operations (FAISS, numpy)
# - libomp-dev: for OpenMP (FAISS parallelization)
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential \
    cmake \
    libopenblas-dev \
    libomp-dev \
    && rm -rf /var/lib/apt/lists/*

RUN python -m venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH"

COPY requirements.txt .

# faiss-cpu means this deployment doesn't use GPU acceleration — install the
# CPU-only torch wheel first so requirements.txt doesn't pull the default
# CUDA-enabled build (several GB of unused NVIDIA libraries).
RUN pip install --default-timeout=1000 --retries 10 --no-cache-dir \
        --extra-index-url https://download.pytorch.org/whl/cpu torch \
    && pip install --default-timeout=1000 --retries 10 --no-cache-dir -r requirements.txt

# ── Final stage: slim runtime image ─────────────────────────────────────
FROM python:3.11-slim

LABEL author="Abdullah Elgendy"
LABEL maintainer="Abdullah Elgendy"
LABEL description="QModel v7 - Quran & Hadith RAG API"
LABEL version="7.0.0"

ENV PYTHONDONTWRITEBYTECODE=1 \
    PYTHONUNBUFFERED=1 \
    PATH="/opt/venv/bin:$PATH"

WORKDIR /app

# Runtime-only system deps (no compiler toolchain here)
RUN apt-get update && apt-get install -y --no-install-recommends \
    libopenblas0-pthread \
    libomp5 \
    curl \
    && rm -rf /var/lib/apt/lists/*

COPY --from=builder /opt/venv /opt/venv

# Copy application code
COPY . .

# Run as non-root
RUN useradd --create-home --uid 1000 appuser \
    && chown -R appuser:appuser /app
USER appuser

# Expose port for API
EXPOSE 8000

# Health check — /health returns 503 until the RAG pipeline finishes loading
HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \
    CMD curl -f http://localhost:8000/health || exit 1

# Start application
# Configure via .env: LLM_BACKEND=ollama or LLM_BACKEND=hf
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]