[SOLVED] Adding technologies to appwrite

Im trying to get tensorflow working with appwrite with the python sdk but the requirements.txt wont allow any installation i looked up online and tensorflow can be installed as docker too but im not familiar with docker and docker functions anything that can help me out id appreciate.

I think the problem with tensor flow is there are additional dependencies required on Alpine (our base image).

Related issues (make sure to 👍)

• https://github.com/appwrite/appwrite/issues/1037
• https://github.com/open-runtimes/open-runtimes/issues/86

Cc @Eldad

as far as i can think of when executing functions a container gets created for that function and put outside of the appwrite docker instance which is 22 total containers, makes sense why you guys can not show the function error logs yet.

yeah after reading these its impossible for me to do that yet 🥲 is it still in progress since 2021? why was this closed?

Nothing was closed. 🧐 What do you mean?

i meant removed from in-progress

I'm not sure we were really actively working on it to begin with. Were not really using GitHub projects 😅

ok so, for this to get fixed we need to find a way to put the active docker function inside the appwrite docker instance right? this can be done with multi-stage builds which allows you to change the base image for that to happen, during function deployment there should be a script running to put the function container inside the appwrite instance and checking if there are any added docker images to be used if those are already added continue with the multi-stage change if not the script should add the image to the Dockerfile and commit the changes, im gonna go rest and check on these tomorrow..

Appwrite functions already run in a docker container. The problem is the container OS doesn't have pre-built binaries of these python packages and it doesn't have tools (like gcc) to build the packages.

🤔 why is there need for gcc you can compile every package like that? Is it not wasted resources to have each function container with the same dependencies created when you can create and deploy your own container, wouldnt it better to have it required to put the function container along with the other deployed container that has the dependencies?

Sorry I don't understand your suggestion about putting function container along with other deployed container

i was specificaly talking to this where the first stage is the alpine image with the dependencies and the second one the appwrite function correct me if im wrong https://docs.docker.com/build/building/multi-stage/

We don't build an image for your function. we use this image: https://github.com/open-runtimes/open-runtimes/blob/main/runtimes/python-3.10/Dockerfile

Do you by any chance know where this is used? i would assume when the function gets executed, btw is there any resources on the methodology of open-runtimes?

This image is spun up, the code is copied in, and then your code is executed

@Steven i used another machine to make calls and pass the results on a function should i close this thread?

If you think it's solved, you can mark it as solved.

[SOLVED] Adding technologies to appwrite