wrangler detects it

Your worker has access to the following bindings:
- Constellation Projects:
  - CLASSIFIER: removed

but then its undefined 🤔

you made your own constellation project and changed the project_id to it as well, I assume, right?

that error looks like exactly the error you would get if the binding is null I believe, just not sure why it would be

oh, are you manually hiding the ID of your constellation project and it's not actually saying the classifier is "removed"? lol, that confused me

yup yup

i tried creating it through cli and dashboard, same result

I just realized now, as part of the setup process, they have you locally install the beta version of wrangler

npm install wrangler@beta --save-dev

, not sure if this is still required or not, but might be worth a try. If you npm install'd when you cloned my repo though, you should already have it

yep tried that too

welp you're doomed

i think wrangler and i dont get along very well

a definite possibility

a colleague tried to repeat the processes and has the same issue on his machine

I don't think this could even the error, but you did change out the model id in the source code as well, right?

yup

env is undefined at all stages of the worker

not undefined, its just an empty object

One other possibility, on Wrangler 3.x the default is local testing, which won't work, has to be

npx wrangler dev --remote

or just publishing it and testing it live

oh my god

certainly took the long way around to get there. It looks like the default instructions do just say to use

npx wrangler dev

, not sure if they were before 3.x came out, or it should work locally but doesn't (which is consistent with my experience)

yeah it's not supported locally. You'd have to use remote, which was default in wrangler 2, but not in 3. 😅

is this after zipping/compression?

no clue 🤷 I just know they upped the limit from 10mb to 50mb

alright thanks for the info 👍

The inference is quite fast and 50mb is plenty to run many models, I was wondering if there is any plan to allow some sort of python pre-processing before constellation picks the input and the model 🤔 . We have some models that rely on pillow, numpy & pandas to process inputs but, porting or finding an alternative in js/ts can be troublesome and makes moving from other clouds to constellation quite challenging

So like supported Python-to-JS transpilation?

allowing to upload a python function that is executed before the model would fix most struggles i think

Yeah, but to where? CF doesn't currently have a Python-compatible runtime, so it would have to be transpiled(or run elsewhere)

that'd be something for the team to figure out 😅

||Unfortunately we cannot be bothered to maintain supercloud-python. We recommend converting your Python pre-processor to JavaScript. Thank you, and have a *super* day!|| This is a /s, if it isn't clear.