Ganimede

Store, version, edit and execute notebooks in sandboxes and integrate them directly via REST interfaces.

Use cases

• Ability to write machine learning logic and expose them to systems as rest api
• Write Jupyter nb locally and run them in a centralised powerful machine to reduce cost
• Create framework to directly connect Jupyter notebook to other systems

Requirements

• docker
• redis

Stack

• Redis
• FastAPI
• Papermill
• Jupyter
• Poetry
• Docker

Build

• Clone the repo
• Run poetry install
• Run run.py or scripts\launch.sh or cd docker;docker-compose up -d

Deployment

• Clone the repo and in docker folder, run docker-compose build. The docker image will be build
• Push to registry or use your custom publishing method to publish the image

API Docs

• Start the application
• Go to localhost:8000/docs for swagger and localhost:8000/redoc for redoc

Main APIs

Jupyter Notebook

• define for defining projects and its dependencies
• store for storing notebook and associated files
• run for executing a jupyter file
• html to get a rendered page of executed notebook
• output to get the output of Jupyter execution in json format
• plain_text to get the plain text output

Jupyter Notebook Editor

• edit for editing notebook in a sandbox
• view for viewing notebook in sandbox and can run it, but not save the changes

Jupyter Updater

• update for storing the next version of notebook from edit endpoint

TODO

• Provide live environment for editing and running jupyter
• Custom transformations for jupyter output
• Scheduled cleanup of created jupyter docker containers
• Change container implementation to podman or other rootless systems