Dockerfile

FROM ollama/ollama:latest

# Install nginx and bash (for scripting) inside the container
# ollama/ollama is likely Debian/Ubuntu-based (check with 'cat /etc/os-release')
# RUN apt-get update && \
#     apt-get install -y nginx bash && \
#     apt-get clean && \
#     rm -rf /var/lib/apt/lists/*

# f16 - High precision, high memory usage (default)
# q8_0 - 8-bit quantization, ~50% memory reduction with minimal quality loss
# q4_0 - 4-bit quantization, ~75% memory reduction with noticeable quality impact

# Set environment variables to configure Ollama
ENV OLLAMA_HOST=0.0.0.0:7860 \
    OLLAMA_NOHISTORY=true \
    OLLAMA_MULTIUSER_CACHE=false \
    OLLAMA_NOPRUNE=true \
    OLLAMA_MODELS=/dev/shm/ollama/models \
    OLLAMA_TEMP=/dev/shm/ollama/tmp \
    OLLAMA_FLASH_ATTENTION=0 \
    OLLAMA_KV_CACHE_TYPE=q4_0 \
    OLLAMA_NUM_THREADS=2 \
    OLLAMA_MAX_MEMORY=16GB \
    OLLAMA_KEEP_ALIVE=1m \
    OLLAMA_LOAD_TIMEOUT=1m \
    OLLAMA_MAX_LOADED_MODELS=2
    
# OLLAMA_MODEL=tinyllama

# Expose both ports: 8080 for Ollama internally, 7860 for nginx externally
EXPOSE 7860 

# Copy your custom nginx config to container
COPY nginx.conf /etc/nginx/nginx.conf

# Copy entrypoint script
COPY entrypoint.sh /entrypoint.sh
RUN chmod +x /entrypoint.sh


# Create RAM directories with permissions
RUN mkdir -p /dev/shm/ollama/models /dev/shm/ollama/tmp && chmod -R 777 /dev/shm/ollama && \
    rm -rf /.ollama && mkdir -p /.ollama && chmod -R 777 /.ollama


ENV PYTHONUNBUFFERED=1 \
	GRADIO_ALLOW_FLAGGING=never \
	GRADIO_NUM_PORTS=1 \
	GRADIO_SERVER_NAME=0.0.0.0 \
	GRADIO_THEME=huggingface \
	SYSTEM=spaces \
	SHELL=/bin/bash

CMD ["serve"]
# ENTRYPOINT ["/entrypoint.sh"]