Sketch plan

[alecandido]

The overall idea is to just work locally, and then use a CLI to execute various operations remotely, possibly directly on a given queue.

E.g., for a Qibocal calibration run, contained in the scrip mycal.py, the CLI should

• take the script as input
• connect to the configured cluster (config file/env var/option/...) through SSH
  • using the available SSH key, if registered
• pull the repo with all submodules, and the submodules recursively
• poetry install on the login node
  • convenient for the QPU nodes, less convenient for simulation and powerful machines - where the installation happens could be configured
• dispatch a shell script on the queue, which is
  1. activating the environment
  2. setting the required variables (e.g. $QIBOLAB_PLATFORMS)
  3. running python mycal.py