upvote
Code review could also be run as an unattended/batched task though, possibly with at least some use of on-prem inference (which excels at this). That would be a major saving compared to the usual cloud inference scenario.
reply
with which models, though?
reply