Dockerfile

FROM python:3.10-slim

WORKDIR /app

# Install the engine (Same engine Ollama uses)
RUN apt-get update && apt-get install -y build-essential cmake git
RUN pip install --no-cache-dir gradio huggingface_hub llama-cpp-python

# Copy the app
COPY . .

# Start the server
CMD ["python", "app.py"]