RTXteam / RTX

Software repo for Team Expander Agent (Oregon State U., Institute for Systems Biology, and Penn State U.)
https://arax.ncats.io/
MIT License
33 stars 21 forks source link

SmartAPI challenges #2144

Closed edeutsch closed 1 year ago

edeutsch commented 1 year ago

Hi everyone, I'm just stumbling back into things after the trip. I did a git pull in devED and saw:

 code/ARAX/ARAXQuery/ARAX_query.py                     |  1 +
 code/ARAX/ARAXQuery/ARAX_query_tracker.py             |  2 ++
 code/ARAX/ARAXQuery/Expand/kg2_querier.py             | 27 +++++++++++++++++----------
 code/UI/interactive/rtx.js                            | 55 +++++++++++++++++++++++++++++--------------------------
 code/UI/interactive/rtx.version                       |  2 +-
 code/reasoningtool/kg-construction/QueryNCBIeUtils.py | 88 +++++++++++++++++++++++++++++++++++++++++++++++++---------------------------------------
 deploy/arax/scripts/download_database.sh              | 16 +++++++++++++++-
 deploy/arax/templates/deployment.yaml                 |  6 +++++-
 deploy/arax/values.yaml                               | 15 +++++++++++----
 deploy/kg2/scripts/download_database.sh               |  2 +-
 requirements.txt                                      |  1 +

and restarted the service. It restarted okay (although seemed quite slow to start) but when I do our example 1 query, I see:

image

I've not seen the "KP does not have a development TRAPI 1.4 endpoint" message before and it is weird that we claim that MolePro does not support this edge. This is a classic MolePro query I think?

Anyone know what might be going on? https://arax.ncats.io/devED/?r=171450

It works much better in CI: https://arax.ci.transltr.io/?r=171451

I just tried running on /beta with the same problem as devED. It seems that the KP selector system is broken on multiple endpoints on arax.ncats.io. Does anyone know anything about this?

edeutsch commented 1 year ago

it does seem like a lot of services are missing development maturity: https://arax.ncats.io/devED/?smartapi=1

Is our code confused or is there a systemic problem with many Translator services not having an development maturity?

isbluis commented 1 year ago

I am not sure if this is related, but there seem to be two different KPs with the infores:molepro identifier right now:

image

It does appear that none of the automat KPs have a dev endpoint: https://arax.ncats.io/index.html?smartapi=1

edeutsch commented 1 year ago

aha, good eye! I wonder if there would be a good way to detect such duplication and flag it red! with animated blood spattering, even.

isbluis commented 1 year ago

Something like this? https://arax.ncats.io/devLM/?smartapi=3

image

edeutsch commented 1 year ago

sure, looks good.

There's a chance that we may have to accept this as the new normal, in which case we'll have to change our code, which I assume somewhere assumes that there is only one entry per infores key per TRAPI version or something..

isbluis commented 1 year ago

This seems counterintuitive, since, among other issues, it would obfuscate provenance.

In any case, let me know if/when this should be pushed to master.

edeutsch commented 1 year ago

I would agree, but I can imagine arguments against..

push it!

isbluis commented 1 year ago

Awful Prairie dog pushed!

edeutsch commented 1 year ago

great, thanks! deployed to instances tracking master

isbluis commented 1 year ago

Looks like MolePro is no longer duplicated:

image

And the Automat ones also have a 1.4 dev endpoint.

Are you able to run your test again?

edeutsch commented 1 year ago

sure, but I see somewhat unusual results.

at devED, I see: https://arax.ncats.io/devED/?r=172032 image

But at CI, I see: https://arax.ci.transltr.io/?r=172031 image

(no gelinea!!)

I wonder if a restart is needed to pick up changes to SmartAPI? I assumed that this would automatically refresh after 10 minutes? I suppose it possible that the refresher died on CI.

isbluis commented 1 year ago

gelinea only has a dev endpoint, and no staging -- is this what CI is?

image

edeutsch commented 1 year ago

ah yes, you are right, exactly!

edeutsch commented 1 year ago

With all that considered, this has been fixed, closing.