When we create new python files, or modify the ones we are executing, there is no button to save the file and the user is forced to use menu or CTRL+S

Here is the build log

$ pushd packages/enterprise_scheduler_extension && yarn install && popd
~/build/ai-workspace/ai-workspace/packages/enterprise_scheduler_extension ~/build/ai-workspace/ai-workspace
...
error Couldn't find package "@wdp/common-canvas@^6.1.9" required by "@aiworkspace/[email protected]" on the "npm" registry.

The notebook pipeline visual editor enables
building pipelines and executing on remote
runtimes such as Kubeflow Pipelines.

Co-authored-by: Martha Cryan [email protected]
Co-authored-by: ALAN K. CHIN [email protected]

• switched over to lerna to manage packages
• relocated extension over to packages dir
• ensure lerna is installed "npm install -g lerna" and then run "lerna bootstrap" to install scheduler extension which should then pull in the jupyterlab packages as well.

Still need to update scheduler.py with more appropriate log msgs

This should be similar to notebook cell, where [*] indicates that the cell is running and [number] indicates it has been done

• Hard codes the image used for now

• Adds a button to save a current pipeline, which opens a dialog to ask for the filename (doesn't check for validity of filename yet). Adds .pipeline to the filename then saves to the jupyterlab filesystem
• Adds a button to open a pipeline - must have a file selected in the filebrowser. Currently doesn't check for validity of filename / content.
• Adds a button to clear the canvas

Please review but don't merge yet

CMD+Z works as undo on Jupyter notebook cells
But not on the python editor

• "New pipeline" toolbar icon was too big

• Relies on an issue being fixed in phosphor, see this jupyterlab PR for more information.

Fixes #40

./npmrc file was not being created or placed in the correct directory
Fixes #8

The pipelines editor was originally created as a separate
extension on it's own repository. This commit moves the
initial implementation to the ai-workspace repo inside
the packages directory.

Co-authored-by: Martha Cryan [email protected]
Co-authored-by: Luciano Resende [email protected]

The utils.ts file will eventually be removed and called from jupyterlab

small changes to the TS build configs

also changed node package build command from using jlpm to yarn

After a successful pipeline submission, open experiment/run window as a lab widget

Add a init.py to a notebook folder
Create a file helloworld.py with the following contents

class Helloworld:
    def say_hi(self, name):
        print('Hi {}'.format(name))

Create a notebook and import that file (it works)

from helloworld import Helloworld
a = Helloworld()
a.say_hi('John')

Do the same from a python file

Error : ModuleNotFoundError - No module named 'helloworld'

When you try from .helloworld import Helloworld you will get

Error : ModuleNotFoundError - No module named '__main__.util'; '__main__' is not a package

The following sample code results in nothing when running from python runner

import io
import requests
import pandas as pd

def df_from_url(url):
    data = requests.get(url).content
    df = pd.read_csv(io.StringIO(data.decode('utf-8')))
    return df

df = df_from_url('http://samplecsvs.s3.amazonaws.com/Sacramentorealestatetransactions.csv')
df['price'].mean()

• Moved kfp container logic over to NotebookOp
• updated tar to package files into archive with root folder (i.e packages all files in foo/* but not foo/ itself)
• See https://github.ibm.com/ai-workspace/kfp-notebook/pull/1 for needed NotebookOp Class and bootstrapper.py script

• Store node data (full path to notebook file, docker image to be used, etc) in app_data within the flow in Common Canvas

Here is how to make it auto-enabled

Adds integration with JupyterLab file browser and
add selected notebook(s) files, including information
regarding the workdir, to the active pipeline.

• Fixes #33

• Adds command to context menu to "Open Notebook" when a node is right-clicked
• When multiple notebooks are selected, uses the command "Open Notebooks" in the context menu, and opens both of them.

• Use common properties (documentation here) to allow users to specify properties such as docker image to use for execution of that notebook. Properties should be specific to each node.

testing

• Add python icon to file editor tab
• Added python file launcher
• Added python file option to file menu