Dockerfile

# Use an official Python runtime as a parent image
FROM python:3.10-slim

# Set the working directory in the container
WORKDIR /code

# Set environment variables - ALL EPHEMERAL CACHING (no persistent storage)
# All caches in fast /tmp/ storage (cleared on restart - prevents future bloat)
ENV TRANSFORMERS_CACHE="/tmp/transformers_cache"
ENV HF_HOME="/tmp/hf_cache"
ENV MPLCONFIGDIR="/tmp/matplotlib_cache"
ENV TORCH_HOME="/tmp/torch_cache"
ENV PYTORCH_KERNEL_CACHE_PATH="/tmp/torch_kernels"

# PyTorch CUDA memory allocation optimization
ENV PYTORCH_CUDA_ALLOC_CONF="expandable_segments:True"

# Copy and install dependencies FIRST (for better caching)
COPY requirements.txt .
RUN pip install --no-cache-dir --upgrade pip && \
    pip install --no-cache-dir --trusted-host pypi.python.org -r requirements.txt

# Copy application code LAST
COPY . .

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