Python VetiverAPI deployment issues to. Posit connect

[rhea2909]

Hello I am facing packaging issues on Posit workbench vs posit connect while deploying a vetiver python model. I am using
vetiver.deploy_rsconnect( connect_server, board=board, pin_name=model_name, )
I want it to use my requirements.txt file but it still uses the default env.
The error i get is this:
ERROR: No matching distribution found for cloud-init==23.1.2 2024/11/19 19:31:49.530226724 pip install failed with exit code 1

[rhea2909]

I have a custom transformer in my model and i tried to do it in the way that is described here: https://github.com/rstudio/vetiver-python/blob/main/docs/custom_code.qmd
But still after running the rsconnect deploy function the error is the same on Posit connect which says: AttributeError: Can't get attribute 'InteractionTransformer' on <module 'main' from '/opt/rstudio-connect/python/connect_fastapi_runtime.py'>

[isabelizimm]

Hi there! Thank you so much for the report. For using the requirements.txt, you can pass it in using something like vetiver.deploy_rsconnect( connect_server, board=board, pin_name=model_name, extra_files=['requirements.txt'])

Are you able to share some more information on the setup of your deployed files for the AttributeError? Eg, what files are being deployed, and what each one is doing?

[rhea2909]

i am deploying app.py, custom_transformers.py and requirements.txt. The app.py is reading the python model pipeline pin from posit connect and wrapping it up in VetiverAPI. app.py is importing my custom_transformers script (from custom_transformers import....)
Still on posit connect I get the same attribute error

[isabelizimm]

Ah, a few things might be happening here! The two most seen ways to get that error are:

1. The original custom_transformer may have been pinned from inside a Jupyter Notebook. Because pickling only saves a reference for how to read a class, not the source code, a custom model transformer pinned from a Jupyter notebook will not resolve to the same location as custom_transformers.py. To fix this, repin your model/transformer from inside the custom_transformers.py script.
2. You may not be supplying the custom transformer to the model. When deploying, you'll want to add in the transformer to the extra_files so that it can be read back later.vetiver.deploy_rsconnect( connect_server, board=board, pin_name=model_name, extra_files=['custom_transformers.py', 'requirements.txt'])

Let me know if either of these fixes helped out!