Dockerfile.train

# Training Dockerfile - matches deployment environment exactly
# Uses Python 3.11 to ensure model compatibility with deployment

FROM python:3.11-slim

# Set environment variables
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1
ENV NUMBA_CACHE_DIR=/tmp/numba_cache

# Install system dependencies (same as deployment)
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential \
    python3-dev \
    gfortran \
    libopenblas-dev \
    && rm -rf /var/lib/apt/lists/*

WORKDIR /app

# Copy requirements (same versions as deployment)
# Note: requirements.txt is in services/coordinate-api/
COPY services/coordinate-api/requirements.txt ./requirements.txt
RUN pip install --no-cache-dir -r requirements.txt

# Install additional dependencies for training (not needed for API)
RUN pip install --no-cache-dir python-dotenv supabase

# Copy ONLY the training scripts we need (not the entire scripts directory)
COPY scripts/pritzker/train_umap_model.py /app/train_umap_model.py
COPY scripts/pritzker/train_hdbscan_model.py /app/train_hdbscan_model.py
COPY scripts/pritzker/sync_market_data.py /app/sync_market_data.py

# Create models directory (will be mounted as volume, but create for safety)
RUN mkdir -p /app/models

# Set working directory
WORKDIR /app

# Default entrypoint (will be overridden by docker run command)
CMD ["python3", "--version"]