Dockerfile

# Use the official Ollama Docker image
FROM ollama/ollama:latest

# Set the working directory inside the container
WORKDIR /app

# Define the model you want to download (replace with your chosen sub-500MB model)
# Examples: smollm:135m, smollm:360m, tinyllama:1.1b-chat-v1.0-q2_K
ENV OLLAMA_MODEL_NAME="smollm:135m"

# Pull the model during the Docker build process
# This ensures the model is present when the container starts,
# reducing startup time and avoiding dynamic downloads.
# The `ollama pull` command can be slow, so pulling it once here is efficient.
RUN ollama pull ${OLLAMA_MODEL_NAME}

# Set OLLAMA_HOST to listen on all network interfaces inside the container
# This is crucial for Hugging Face Spaces to be able to connect to Ollama.
ENV OLLAMA_HOST="0.0.0.0:11434"

# Allow all origins for the API (CORS configuration)
ENV OLLAMA_ORIGINS="*"

# Expose the port Ollama listens on
EXPOSE 11434

# Command to start the Ollama server.
# The base image typically has ENTRYPOINT ["ollama"], so CMD just provides the argument.
CMD ["serve"]

# Optional: Add metadata for Hugging Face Spaces (improves discoverability)
LABEL org.opencontainers.image.title="Ollama with a tiny model"
LABEL org.opencontainers.image.description="Runs Ollama with a pre-loaded model less than 500MB"
LABEL org.opencontainers.image.licenses="MIT"
LABEL org.opencontainers.image.authors="Your Name/Org"