The thing here is that the CLI lists available services in the help message:
blackfish run --help
...
Commands:
speech-recognition Start service MODEL.
text-generate Start service MODEL.
A user will reasonably assume that the listed services are available. So perhaps it is better to simply include logic to check that the required image is available when the user tries to run the service:
blackfish run ... speech-recognition
If the image isn't available, then we return a message. The check should be performed within the service endpoint so that we can also return an error to the UI.
For slurm profiles, we need to check for images in the remote cache (default) and home (backup) directories.
For local profiles, we need to check for images available from Docker or images in the cache and home directories, depending on the container provider available (Docker or Apptainer).
The thing here is that the CLI lists available services in the help message:
A user will reasonably assume that the listed services are available. So perhaps it is better to simply include logic to check that the required image is available when the user tries to run the service:
If the image isn't available, then we return a message. The check should be performed within the
service
endpoint so that we can also return an error to the UI.For
slurm
profiles, we need to check for images in the remote cache (default) and home (backup) directories.For
local
profiles, we need to check for images available from Docker or images in the cache and home directories, depending on the container provider available (Docker or Apptainer).