Add dockerfile

[yarray]

Add dockerfile and modify documentation along with testing script accordingly. Note when ENV=docker, test-installation.sh will not check the existence of the model file to avoid error.

[tmansson]

Great work, but a suggestion for a slightly more efficient Dockerfile.

And i suggest using port 8000 instead of 443, since 443 is restricted on linux systems.

FROM ghcr.io/ggerganov/llama.cpp:light

RUN mkdir -p /llama.cpp/models
RUN mv /main /llama.cpp/

RUN apt update && apt install -y curl xz-utils
RUN curl -O https://nodejs.org/dist/v18.16.0/node-v18.16.0-linux-x64.tar.xz && tar xf node-v18.16.0-linux-x64.tar.xz
RUN mv node-v18.16.0-linux-x64/bin/* /usr/bin/ && mv node-v18.16.0-linux-x64/lib/* /usr/lib/ && mv node-v18.16.0-linux-x64/include/* /usr/include/

COPY ./ /gpt-llama.cpp

WORKDIR /gpt-llama.cpp

RUN npm install
EXPOSE 8000

ENTRYPOINT ["/bin/bash", "-c"]
CMD ["npm start"]

and a docker-compose file to go along with it.

version: "3.9"

services:
  gpt-llama:
    container_name: gpt-llama
    build: .
    ports:
      - 8000:8000
    environment:
      - PORT=8000
    volumes:
      - ../llama.cpp/models:/llama.cpp/models:ro

[star5o]

Great work! I will test it locally soon.