Study how to support credentials in `kedro viz build`

[ravi-kumar-pilla]

Description

kedro viz build fails on CI and local when credentials are involved

Context

• Even though Kedro-Viz doesn’t actually need dataset credentials, if they’re declared in catalog.yml, kedro viz build fails in CI/CD because credentials.yml is gitignored.
• Fake credentials often still trigger “invalid credentials” errors.
• This adds unnecessary friction when publishing via GitHub Actions or other pipelines.

Steps to Reproduce

1. Try placing credentials key in your catalog.yml
2. Try running kedro viz build

Expected Result

kedro viz build should work

Actual Result

Fails with ERROR: Failed to build/deploy Kedro-Viz : "Unable to find credentials

ERROR: Failed to build/deploy Kedro-Viz :
unable to connect to account for Must provide either a connection_string or account_name with credentials!!.
Failed to instantiate dataset 'xxx_data' of type 'kedro_datasets.pandas.excel_dataset.ExcelDataset'.

-- If you received an error, place it here.

-- Separate them if you have more than one.

Your Environment

• Kedro version used (pip show kedro or kedro -V):
• Python version used (python -V):
• Operating system and version:

[Vincent-Liagre-QB]

@ravi-kumar-pilla the description of the issue is not accurate. Even after adding an empty or a dummy crednetials.yml file it fails

[ravi-kumar-pilla]

HI @Vincent-Liagre-QB , did you add the credentials file to the GitHub repository for where you run the github action ?

[Vincent-Liagre-QB]

1. W/o any file :

ERROR: Failed to build/deploy Kedro-Viz : "Unable to find credentials 'azure_blob_storage': check your data catalog and credentials configuration. See https://kedro.readthedocs.io/en/stable/kedro.io.DataCatalog.html for an example." 

2. With empty credentials.yml : same error

3. With dummy credentials.yml :

ERROR: Failed to build/deploy Kedro-Viz : 
unable to connect to account for Must provide either a connection_string or account_name with credentials!!.
Failed to instantiate dataset 'xxx_data' of type 'kedro_datasets.pandas.excel_dataset.ExcelDataset'. 

Dummy credentials :

azure_blob_storage:
  account_name: ""
  account_key: ""
  client_id: "" 
  client_secret: "" 
  tenant_id: ""

[Vincent-Liagre-QB]

@ravi-kumar-pilla yes I added the credentials file as part of the GH Action workflow but notes that I have the same issue / behavior when doing this on my computer ; so its not a GH action issue, really a kedro-viz issue

[ravi-kumar-pilla]

okay, I was trying locally with dummy credentials mentioned here and it works fine locally as there is a file in place. But it is interesting to see it does not work locally for you. I will try with azure too and update the ticket. Did you happen to try it with aws ?

[ravi-kumar-pilla]

Hi @astrojuanlu , Root cause is KedroViz gets catalog using context.catalog which internally has CatalogConfigResolver which tries to resolve credentials. By default kedro viz build and kedro viz deploy have previews as false. So the issue is not with data previews but catalog resolution. It would be nice to resolve credentials when we materialize the dataset and not during catalog resolution. Thank you

[rashidakanchwala]

@ravi-kumar-pilla , do you think this issue will be fixed with the new KedroDataCatalog?