Check if server can be deployed on ibmidocker about streamlit-template

[timosachsenberg]

Happy to explain the plan a bit more in detail once @enetz is back.

[enetz]

I have gotten access to ibmidocker from our IT admin just before my vacation. I should be able to set up a streamlit service, that is accessible from the local network, in the next few days.
Making it public (accessible from outside our local network) will require asking the IT admin to set up proxies for each tool with its own port. Which we can do, when the tools are robust enough for external users.

[axelwalter]

Workspaces without activity in the last seven days will be removed via a daily cron job.

Deployment and updating apps on ibmidocker works. Currently running the template app and UmetaFlow GUI.

[axelwalter]

A detailed tutorial how to host apps can be found here.

Template App and UmetaFlow are accessible as abi-services.

https://abi-services.cs.uni-tuebingen.de/streamlit-template/
https://abi-services.cs.uni-tuebingen.de/umetaflow/

[timosachsenberg]

I think @JeeH-K and @enetz would be great candidates for that because deploying their tools as web app would profit from that

[timosachsenberg]

I updated #4 to be only Docker based. should make it much easier to build custom docker container for feature branches.

[timosachsenberg]

@enetz and @JeeH-K is it possible to display and browse spectra (e.g., by clicking on table rows of identifications / masses)?
I think this would be another important visualization we would need.

[JeeH-K]

Hi Timo

@enetz and @JeeH-K is it possible to display and browse spectra (e.g., by clicking on table rows of identifications / masses)? I think this would be another important visualization we would need.

Yes, displaying and browsing spectra part is (partially) done in FLASHViewer right now. It has a scan table that interacts with deconvolved spectra. When a row in the scan table is clicked, a deconvolved spectrum and a mass table are drawn/updated.

[timosachsenberg]

let's move the web apps forward and meet after this sprint to discuss state/limitations.

[poshul]

How is this looking?

[poshul]

Sounds good. Thanks!

[timosachsenberg]

If the apache is the proxy server, we can probably do this ourself? Would be good if we find a solution that doesn't require contacting an admin every time

[enetz]

This works now in general, one app is running on the local network.

Two points I will contact the admin about:

1. Can we do the proxy setup ourselves
2. Broader access to the ibmidocker machine for the OpenMS team:
   • give everyone access through their WSI accounts, or
   • make one collective OpenMS WSI account to manage apps? (This option might be less secure, since login info for this WSI account would have to be somewhat public within our core team)

[timosachsenberg]

ok that is nice. Thanks! Would be great if you could stay in contact with the admin and push that forward.

[timosachsenberg]

any update here?

[timosachsenberg]

please meet organize yourself as a "task force"

[axelwalter]

Deploying multiple OpenMS apps is now possible with the streamlit-deployment repository using docker-compose.

Anyone with access to the ibmidocker (needs WSI account) can manage the docker containers. Workspaces and user data are stored within docker volumes. I will host them there next week and we can test in the local network before asking to open ports.

One task left is to make sure user data gets cleaned up after a while with a cron job.

[timosachsenberg]

This is mostly done. I would suggest polishing them, so we can have an official presentation.
Things that come to my mind:

• user experience: have a quickstart section that describes what steps users need to do on first run
• potentially an admin section (think about what to do there - e.g. cleanup, looking up ids)
• use pyOpenMS logo ?

[axelwalter]

• add app testing API which was recently introduced in streamlit 1.28