Dockerfile.rag-api

FROM python:3.11-slim

# Set working directory
WORKDIR /app

# Install system dependencies
RUN apt-get update && apt-get install -y \
    curl \
    sqlite3 \
    build-essential \
    && rm -rf /var/lib/apt/lists/*

# Copy requirements and install Python dependencies (using Docker-specific requirements)
COPY requirements-docker.txt ./requirements.txt
RUN pip install --no-cache-dir -r requirements.txt

# Copy RAG system code and backend dependencies
COPY rag_system/ ./rag_system/
COPY backend/ ./backend/

# Create necessary directories
RUN mkdir -p lancedb index_store shared_uploads logs

# Expose port
EXPOSE 8001

# Health check. The RAG API is deliberately single-threaded (its shared
# in-memory pipeline/agent state is only safe because one request runs at a
# time), so one long /chat or /index blocks /health. The generous start_period
# covers model loading, and 10 retries ride out routine long requests instead
# of flapping the container to unhealthy.
HEALTHCHECK --interval=30s --timeout=10s --start-period=120s --retries=10 \
    CMD curl -f http://localhost:8001/health || exit 1

# Run the RAG API server
CMD ["python", "-m", "rag_system.api_server"]