The Execution Environment in Polar TEP allows the headless execution of Jupyter Notebooks. This setup uses an additional API layer on the Polar TEP platform to enable the execution of arbitrary algorithms packaged as Jupyter notebooks in a headless, i.e., non-interactive, manner.
Any Jupyter notebook able to be run in the Polar TEP workspace may also be executed via the headless API, leveraging the same capabilities (libraries, injected environment variables) and respecting the same user resource constraints (quotas) as through direct invocation via the JupyterLab interface.
To use the Execution Environment, a notebook needs to fulfil the following requirements:
- Input parameters within the notebook must be tagged to allow customised execution, e.g. the prepared notebook cell must be tagged as "parameters" to be seen as input.
- The API call to be triggered must contain the following:
- the notebook filename (relative to the configured directory like /extra)
- all input parameters encoded in JSON
- access token retrievable from the user workspace
The resulting notebook contains the output of all executed cells and is stored in the directory job-output in the configured directory along with any output stored by the notebook.
We can also support conda-store and docker containers.
Please get in touch if you’d like to leverage these capabilities and we will configure your workspace as required.