# =============================================================================
# Multi-stage Dockerfile for Accessible Video Processing Platform
# =============================================================================
# Stage 1: Builder - Install dependencies
# Stage 2: Base - Common runtime for API and Worker
# Stage 3: API - FastAPI + Gunicorn (with ffmpeg for TTS audio conversion)
# Stage 4: Worker - Celery worker (with ffmpeg for video processing)
# =============================================================================

# -----------------------------------------------------------------------------
# Stage 1: Builder - Install Python dependencies using Poetry
# -----------------------------------------------------------------------------
FROM python:3.11-slim AS builder

# Install build dependencies
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential \
    curl \
    && rm -rf /var/lib/apt/lists/*

# Install Poetry
RUN pip install --no-cache-dir poetry==1.8.2

# Configure Poetry to not create virtual environment (we're in a container)
ENV POETRY_NO_INTERACTION=1 \
    POETRY_VIRTUALENVS_CREATE=false \
    POETRY_CACHE_DIR=/tmp/poetry_cache

WORKDIR /app

# Copy dependency files
COPY pyproject.toml poetry.lock ./

# Install dependencies using Poetry directly (simpler and more reliable)
RUN poetry config virtualenvs.create false \
    && poetry install --only main --no-interaction --no-ansi \
    && rm -rf $POETRY_CACHE_DIR

# -----------------------------------------------------------------------------
# Stage 2: Base - Common runtime environment
# -----------------------------------------------------------------------------
FROM python:3.11-slim AS base

# Install common runtime dependencies
RUN apt-get update && apt-get install -y --no-install-recommends \
    libmagic1 \
    curl \
    tini \
    && rm -rf /var/lib/apt/lists/* \
    && apt-get clean

# Create non-root user for security
RUN groupadd --gid 1000 app \
    && useradd --uid 1000 --gid app --shell /bin/bash --create-home app

# Copy Python packages from builder (installed globally, not in user dir)
COPY --from=builder /usr/local/lib/python3.11/site-packages /usr/local/lib/python3.11/site-packages
COPY --from=builder /usr/local/bin /usr/local/bin

# Set environment variables
ENV PYTHONPATH=/app \
    PYTHONUNBUFFERED=1 \
    PYTHONDONTWRITEBYTECODE=1

WORKDIR /app

# Copy application code
COPY --chown=app:app . .

# Switch to non-root user
USER app

# -----------------------------------------------------------------------------
# Stage 3: API - FastAPI + Gunicorn (Production API Server)
# -----------------------------------------------------------------------------
FROM base AS api

# Switch to root to install ffmpeg
USER root

# Install ffmpeg for TTS audio conversion
RUN apt-get update && apt-get install -y --no-install-recommends \
    ffmpeg \
    && rm -rf /var/lib/apt/lists/* \
    && apt-get clean

# Switch back to non-root user
USER app

# Set production environment variables
ENV APP_ENV=prod

# Health check for API
HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
    CMD curl -f http://localhost:8000/health || exit 1

# Expose API port
EXPOSE 8000

# Use tini as init system for proper signal handling
ENTRYPOINT ["tini", "--"]

# Start Gunicorn with Uvicorn workers
CMD ["gunicorn", "-c", "gunicorn_conf.py", "app.main:app"]

# -----------------------------------------------------------------------------
# Stage 4: Worker - Celery Worker (with ffmpeg for video processing)
# -----------------------------------------------------------------------------
FROM base AS worker

# Switch back to root to install ffmpeg
USER root

# Install ffmpeg for video processing
RUN apt-get update && apt-get install -y --no-install-recommends \
    ffmpeg \
    && rm -rf /var/lib/apt/lists/* \
    && apt-get clean

# Switch back to non-root user
USER app

# Set production environment variables
# WORKER_CONCURRENCY can be overridden at runtime (default: 8)
ENV APP_ENV=prod \
    C_FORCE_ROOT=0 \
    WORKER_CONCURRENCY=8

# Health check for worker (check if Celery is responding)
HEALTHCHECK --interval=60s --timeout=15s --start-period=10s --retries=3 \
    CMD python -c "from celery import Celery; app=Celery('accessible-video-tasks', broker='redis://redis:6379/0'); app.control.inspect().ping() or exit(1)" || exit 1

# Use tini as init system for proper signal handling
ENTRYPOINT ["tini", "--"]

# Start Celery worker listening to all queues EXCEPT whisper
# Whisper has a dedicated worker (see Stage 5) to prevent memory overload
# Concurrency is set via WORKER_CONCURRENCY env var (default: 8)
# When using Cloud Run for FFmpeg, can increase to handle more parallel HTTP calls
CMD celery -A celery_worker worker \
    -Q default,ingest,notify,render,ffmpeg \
    --loglevel=info \
    --concurrency=${WORKER_CONCURRENCY} \
    --max-tasks-per-child=100

# -----------------------------------------------------------------------------
# Stage 5: Whisper Worker - Dedicated worker for Whisper transcription
# -----------------------------------------------------------------------------
FROM base AS whisper-worker

# Switch back to root to install ffmpeg
USER root

# Install ffmpeg for audio extraction
RUN apt-get update && apt-get install -y --no-install-recommends \
    ffmpeg \
    && rm -rf /var/lib/apt/lists/* \
    && apt-get clean

# Switch back to non-root user
USER app

# Pre-download Whisper medium model during build to avoid cold start delays
# Model is cached in ~/.cache/huggingface/hub (~1.5GB)
RUN python -c "from faster_whisper import WhisperModel; WhisperModel('medium', device='cpu', compute_type='int8')"

# Set production environment variables
# WHISPER_WORKER_CONCURRENCY can be overridden at runtime
# Default: 1 for local mode (RAM constrained), set to 10 for Cloud Run mode
ENV APP_ENV=prod \
    C_FORCE_ROOT=0 \
    WHISPER_WORKER_CONCURRENCY=1

# Health check for worker (check if Celery is responding)
HEALTHCHECK --interval=60s --timeout=15s --start-period=10s --retries=3 \
    CMD python -c "from celery import Celery; app=Celery('accessible-video-tasks', broker='redis://redis:6379/0'); app.control.inspect().ping() or exit(1)" || exit 1

# Use tini as init system for proper signal handling
ENTRYPOINT ["tini", "--"]

# Start Celery worker listening ONLY to whisper queue
# Concurrency is set via WHISPER_WORKER_CONCURRENCY env var
# Local mode: 1 (RAM constrained - Whisper model uses ~4-6GB)
# Cloud Run mode: 10 (just HTTP calls, match Cloud Run max instances)
# --max-tasks-per-child=50 to periodically recycle workers and free memory
CMD celery -A celery_worker worker \
    -Q whisper \
    --loglevel=info \
    --concurrency=${WHISPER_WORKER_CONCURRENCY} \
    --max-tasks-per-child=50