FROM ollama/ollama:latest

# HF Spaces (free CPU) expects port 7860
ENV OLLAMA_HOST=0.0.0.0:7860
ENV OLLAMA_MAX_LOADED_MODELS=1

EXPOSE 7860

# Override default entrypoint so we can run normal commands
ENTRYPOINT []

# Start Ollama server + pull glm4:9b (quantized for CPU) + keep alive
CMD ollama serve & sleep 15 && ollama pull glm4:9b && echo "✅ deepseek-r1:8b ready!" && tail -f /dev/null