Error when running KNIME workflow in Jupyter Notbook

Hi everyone,

First of all, I like the interation of KNIME into Jupyter Notebook and vice versa very much :slight_smile: . I tried to run a KNIME workflow from Jupyter, similar as it is explained in your recent YouTube video (https://www.youtube.com/watch?v=aQqqlXEZ9Ag). However, when finally running the code

with knime.Workflow(workflow_path=workflow,workspace_path=workspace) as wf:
    wf.data_table_inputs[0] = df
    wf.execute()

I get the error message “Output from KNIME not found”.

I am stuck at this point couldn’t find a solution online… Maybe somebody here can help me?

Thanks in advance!

P.S. I attached a snapshot of my simple KNIME workflow.
image

2 Likes

Hi fabi_an –

In addition to the error message, in Jupyter notebooks it should also give you a red/pink box containing more detailed information captured from stdout/stderr. This is how Jupyter, by default, shows messages sent to the logging module’s logging facility. If you have instead configured logging to be sent to a file or somewhere else (possibly via logging.config(...) or via Jupyter’s controls for logging) then you will not receive an attention-grabbing pink box because it is already being sent somewhere else. Here is how it looks for me when I trigger a problem:

This may not be the root of the problem you are seeing, but I keep making this same mistake: I forget to close my workflow in the KNIME GUI before trying to execute it from my Jupyter notebook. When KNIME has a workflow open for editing in the GUI, it locks the workflow to protect against something else modifying and thereby corrupting it. If this is the case, simply close that one workflow in KNIME and try executing it from Jupyter/Python again.

Hope this helps,

Davin

2 Likes

Hi @potts,

Can the Jupyter integration also be done in R?

Thanks,
Haystack

Hi @Haystack –

The question from @fabi_an was really a question about the behavior of the knimepy Python package. It just happened to be used from a Jupyter notebook.

I suggest asking your question again in a new thread so that the developers who worked on the Jupyter integration might see it.

Thanks,

Davin