added a simple docker based solution

[karlkurzer]

Added a simple Dockerfile to get it running :-)

[hydazz]

how big is the resulting docker image without cleaning up node...

[karlkurzer]

@hydazz the resulting image is about 2.9 GB

[hydazz]

npm prune --omit=dev --omit=optional && \ npm cache clean --force

[karlkurzer]

I guess that could be added, however then one can't really use it for development, right?

[hydazz]

Edit the source (local repo) and rebuild the image if one is to developing, either way nobody's going to do any "developing" from within the docker image, its best to recreate it.

[karlkurzer]

npm prune --omit=dev --omit=optional && npm cache clean --force

no change in image size

[hydazz]

yarn cache clean*

[karlkurzer]

yarn cache clean*

no change in image size

[hydazz]

🤔 https://github.com/imagegenius/docker-ai-chat-app/blob/main/Dockerfile#L28-L32

This image is only 400mb

perhaps

  rm -rf \
    /tmp/* \
    /root/.npm \
    /root/.cache

[jimzer]

Hey @karlkurzer :) Thanks for contributing! The Docker image will simplify the setup 🔥.

Do you want to try the rm trick from @hydazz before we merge?

[DennisTheD]

You could use a multi-stage Docker build to reduce the image size while keeping the Dockerfile simple and clean. Example: https://github.com/DennisTheD/ai-chat-app/blob/main/Dockerfile (861,76M image size). Not only do you reduce the image size, but also the total layer count. Using a smaller base image (like imagegenius) can help to reduce the size even further, but makes it harder to maintain.

[hydazz]

yarn >> npm >> npm prune --omit=dev --omit=optional

[DennisTheD]

yarn >> npm >> npm prune --omit=dev --omit=optional

Great idea! I updated my fork and gave it a try. Combined with the multi-stage build, I ended up with an image size of 481M, while still making use of the maintained node:lts-alpine image. If you like, I can provide you with another pull request. Alternatively, you might simply want to copy my Dockerfile? I also could implement a github action to automatically create the image after every update in this repo and push it to the Docker hub. This would simplify the setup even further. What do you think about this?

[karlkurzer]

@DennisTheD I think the proposal with the github action makes most sense :-), please feel free to create a new PR, and I'll close this one!

[hydazz]

@DennisTheD well I mean using npm rather than yarn for everything:

  npm install && \
  npm ci && \
  npm run build && \
  npm prune --omit=dev --omit=optional && \
  npm cache clean --force && \

this removes the need for yarn completely and makes the image tiny

btw prisma migrate deploy either by yarn or npx needs to be ran as a init script, as this initializes the database and if it is done in the dockerfile then you will get errors about tables not existing.

[DennisTheD]

@karlkurzer OK, I will prepare a new PR. @hydazz Since my Dockerfile uses the node:lts-alpine as a base image, yarn is included anyway. I agree that we could reduce the size even further by building a custom base image and removing this dependency (and maybe some other unused packages). But using the official node image brings in the benefits of a well maintained base image in regard to patches and things like this. To move the database migration into a separate init script might be a good idea when it comes to updating the image later on, I already thought about that.

[hydazz]

Just use alpine as the base and the only package that is needed is npm