nuclio-jupyter 0.4.3

Convert Jupyter notebook to nuclio

nuclio Jupyter Export

Convert Jupyter notebook to Python code that can run as nuclio handler

Installing

pip install nuclio-jupyter

Install in Notebook

Run the following in a cell

# nuclio: ignore
!pip install nuclio-jupyter

Usage

When developing, import Event from nucilo and use it to generate a mock request events.

import nuclio

This will inject a context object into the notebook namespace.

Then

# nuclio: ignore
from nuclio import Event

event = Event(body='Hello Nuclio')
# your code goes here

In the cell you'd like to become the handler, added the magic %%nuclio handler If there's a specific line you'd like to be the returned one - added # nuclio:return at the end of it.

Cells containing # nuclio: ignore comment will be omitted in the export process.

Now choose File/Download as/Nuclio in Jupyter notebook (you might need to mark the notebook as Trusted in order for the Nuclio option to show)

Or you can run

jupyter nbconvert --to nuclio example.ipynb

This will create example.zip with your code wrapped in handler function and function.yaml nuclio configuration file.

Example

Will generate a zip file with

handler.py

# Generated by nuclio.export.NuclioExporter on 2018-12-10 08:26:39.086907

import nuclio

def handler(context, event):
   msg = 'Hello ' + event.body
   return msg # nuclio: return

and function.yaml

# Generated by nuclio.export.NuclioExporter on 2018-12-10 08:26:39.095522
apiVersion: nuclio.io/v1
kind: Function
metadata: {}
spec:
 runtime: python:3.6
 handler: handler:handler
 env:
   - name: API_KEY
     value: '1234'
   - name: NUM_JOBS
     value: '2'
   - name: PASSWORD
     value: t0ps3cr3t
   - name: USER
     value: guazio
build:
 commands:
 - apt-get install -y libyaml-dev

Other Notebooks

We currently don't support Jupyter Lab, Google Colaboratory, Kaggle Notebooks and others.

You can use nuclio.print_handler_code or nuclio.save_handler_code to print/save the handler code and then copy&paste it to the nuclio dashboard.

Try It Out

You can build a docker image and try it out

Build

$ docker build -t jupyter-nuclio .

Run

$ docker run -p 8888:8888 jupyter-nuclio

Then open your browser at http://localhost:8888

Developing

We're using pipenv as package manager. To install dependencies run

$ pipenv sync -d

To run the tests run

$ pipenv run python -m pytest -v tests

Licence

Apache 2.0 (see LICENSE.txt)