Instance-Level Self-Hosted Model serving configuration
When a customer self-hosts a model, the existence of this model along with it's connectivity information needs to be declared to GitLab Rails. GitLab Rails is responsible for exposing this information to the AI Gateway.
Once a model has been declared, it can be assigned to specific GitLab Duo functions as described in Configuration option to select model (#454323 - closed)
This configuration is added in the instance-level configuration settings.
See also Self-Hosted Model Deployment Blueprint (!148599 - merged)
Screenshots
Taken from the MR Draft: Instance-Level Self-Hosted Model serving... (!151505 - closed)
Edited by Sean Carroll