Dockerfile

FROM python:3.9

# Setup
RUN apt-get update && apt-get install -y curl wget netcat-traditional && \
   curl -fsSL https://ollama.com/install.sh | sh && \
   useradd -m -u 1000 user

WORKDIR /app

# Install dependencies
COPY --chown=user requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Copy files
COPY --chown=user . .

# Create files
COPY --chown=user Modelfile /app/Modelfile

# Start script
RUN echo '#!/bin/bash
ulimit -v unlimited
ollama serve --verbose &
timeout=60
until nc -z localhost 11434 || [ $timeout -le 0 ]; do
   sleep 1
   ((timeout--))
done
cd /app
ollama create llama3.2:1b-papalia -f Modelfile
exec uvicorn app:app --host 0.0.0.0 --port 7860 --workers 1 --limit-concurrency 1' > /app/start.sh

RUN chmod +x /app/start.sh

USER user
ENV PATH="/home/user/.local/bin:$PATH" \
   MALLOC_ARENA_MAX=2

HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \
   CMD curl -f http://localhost:7860/health || exit 1

CMD ["/app/start.sh"]