dockerfile

# Use an official Nvidia CUDA runtime as a parent image
FROM nvidia/cuda:12.4.1-devel-ubuntu22.04

# Avoid interactive prompts - auto-select defaults for any prompts 
ARG DEBIAN_FRONTEND=noninteractive

# Set timezone for tzdata package as it's a dependency for some packages
ENV TZ=America/Los_Angeles

# Set the working directory in the container
WORKDIR /app

# Install build dependencies: build-essential includes GCC, G++, and make; libffi-dev for Foreign Function Interface (FFI); libssl-dev for SSL support
RUN apt-get update && apt-get install -y \
	software-properties-common \
	build-essential \
	libffi-dev \
	libssl-dev \
	poppler-utils \
	libreoffice \
	git \
	net-tools \
	iproute2 \
	cuda-toolkit-12-4

# Test LibreOffice installation - any errors will interrupt the docker build process
RUN libreoffice --version

# Copy the current directory contents into the container at /app
COPY . /app

# Clone specific llama.cpp release
RUN git clone -b b3447 https://github.com/ggerganov/llama.cpp.git

# Build llama.cpp
WORKDIR /app/llama.cpp
RUN make GGML_CUDA=1 -j $(nproc)

# Add the build path to the system PATH
ENV PATH "/app/llama.cpp:${PATH}"

# Change back to the main app directory
WORKDIR /app

# Add deadsnakes PPA for Python 3.11 - since the Ubuntu rep may not contain our desired version of Python, we install a popular Personal Package Archive (PPA), deadsnakes, to get the desired version which will pull the latest 3.11.X (other specific subversions are not hosted)
RUN add-apt-repository ppa:deadsnakes/ppa -y

# Install Python 3.11 and upgrade pip, while ensuring pip is installed
RUN apt-get update && apt-get install -y python3.11 python3.11-venv python3.11-dev && python3.11 -m ensurepip

# Create a virtual environment to ensure Python Packages are isolated from the system
RUN python3.11 -m venv /venv

# Activate the venv (modifying PATH is typical for a container env rather than 'activate')
ENV PATH="/venv/bin:$PATH"

# Now upgrade pip from within the venv
RUN pip install --upgrade pip

# Install any needed packages specified in requirements.txt
RUN pip install -r requirements_linux.txt

# make port 5000 available to the world outside this container
EXPOSE 5000
EXPOSE 8080

# Run application
CMD ["python3", "web_app/app.py"]