PhotoPrism - Personal Photo Management powered by Go and Google TensorFlow

[jdaviescoates]

@doodlemania2 I think perhaps you're confusing me with @marcusquinn who has sponsored various packages

I'm barely breaking even and so have not sponsored any!

But I also think @nebulon will perhaps be up for packaging this once it's actually release - he's previously met with the developers because I think they all live in Berlin.

So, um, great! But, it hasn't been released yet, I'm not sure there is any bounty, and I think perhaps no bounty will be necessary.

[doodlemania2]

@jdaviescoates ah - no worries - i'm interested in it too, just didn't know if there was a bounty out there, I know there's lots of those flying around

[nebulon]

Right, we have to wait for a first upstream release in order to get going with the packaging.

[doodlemania2]

@nebulon tell em to hurry! hehe

[jdaviescoates]

@doodlemania2 I just noticed they hit their MVP milestone 2 days ago https://github.com/photoprism/photoprism/milestone/1

And they've got some docs here too:
https://docs.photoprism.org/getting-started/

[nebulon]

Theresa from the PhotoPrism team contacted me yesterday about next steps to get a package out. Lets see how far we can get this week.

[lucidfox]

Any updates on this... would love to have this on Cloudron. Don't mean to nudge, just curious how it's going with packaging.

[nebulon]

While we are in contact with the upstream devs, I have not managed to build a working package yet, since installation of tensorflow appears to be very particular.

[lucidfox]

Ok, thanks much for the update. Wish I could help with the packaging, but I've got zero skills or experience to contribute to that. Would be amazing to have this on Cloudron at some point, it seems to be the most promising Google Photos alternative.

[jdaviescoates]

@nebulon said in PhotoPrism - Personal Photo Management powered by Go and Google TensorFlow:

have not managed to build a working package yet, since installation of tensorflow appears to be very particular.

Are you still working on this?

Manage to work out how to install tensorflow?

Thanks!

[A Former User]

@nebulon Any ideas when it could be available?

[moonmeister]

@nebulon I'd be interested in sponsoring this to happen. Not sure how much $$ that'd be but if you or someone else is interested please let me know.

[girish]

@moonmeister Might also be interesting if you contact the photoprism team itself for packaging this. If they are interested, we can help them out (but you can pay them alone). AFAIK, they are also a small team and any help them get would be appreciated by them.

[moonmeister]

@girish will do

[moonmeister]

Okay, I decided to take a crack at packaging this and was surprisingly successful.

I've detailed some of the more technical things over on the PhotoPrism Github: https://github.com/photoprism/photoprism/issues/1284

For anyone who wants to expriment the docker image is moonmeister/cloudron-photoprism:0.1.3 and the configs can be found here https://github.com/moonmeister/photoprism/tree/feat/cloudron-image/docker/cloudron

PLEASE NOTE: "Working" is currently defined as "You can log in". Basic features like, uploading photos are still broken.

@girish I'm currious what you need when this is all working to publish. Meaning, do you need a dedicated repo with the manifest and Dockerfiles and all associated info? Can this be anywhere or does it need to be in the cloudron gitlab? Or can this stuff live in the PhotoPrism repository in a subfolder?

I think that's everything I have for now. Anything else I should know? thanks.

[nebulon]

@moonmeister ah impressive, so you were able to deal with tensorflow using our base image? Looking at your Dockerfiles you seem to temporarily rely on moonmeister/photoprism-dev:test2 is this something which could be put into that Dockerfile for completeness to have version pinned releases in the package?

[moonmeister]

@nebulon Yes, If you checkout my GitHub issue on the PhotoPrism I address that issue. Basically, PhotoPrism doesn't release builds, they release a docker image. That image is built on Ubuntu 12.10. To make the build compatible with 12.04 I had to rebuild their dev image which is what that image is. Before I duplicated all that I was waiting to see if they had any feedback to optimize the process.

[moonmeister]

@nebulon As to tensorflow...I just modified their production Dockerfile. I assume it works as of yet.

[moonmeister]

@nebulon so photoprism is running on 20.10 due to bugs in some dependencies on 20.04. Is it possible to get a 20.10 base image or do I need to figure out backport those dependencies?

See https://github.com/photoprism/photoprism/issues/1284#issuecomment-841406953 for context.

UPDATE: photoprism folks informed me they tried compiling ffmpeg for 20.04 but it was a huge nightmare and they gave up. A 20.10 21.04 base image sounds much easier if y'all can provide that.

[robi]

@moonmeister why not take the binary from 20.10 and use it on 20.04? Or use a statically compiled one that works anywhere?