Azure openai deployments hard-routed to model

[Willem-J-an]

I'm provided with an Azure openai endpoint which i'm using with gp.nvim.

                openai = {
                    secret = { "pass", "show", "OPENAI" },
                    endpoint = "https://myurl/openai/deployments/gpt-4.1/chat/completions?api-version=2024-03-01-preview",
                }

The problem is the model is referenced in the deployment, and it effectively ignores the model specified in the "agents" section. This means I can only change models by changing the deployment name, and I can't choose a different agent to use on the fly.

Not sure if others have this problem as well? Potential solution direction could be to allow specification of a format in the endpoint like endpoint = "https://abc/openai/deployments/{{model_name}}/chat/completions", then replace the model_name on the fly with the value from the agents?

Or allow specification of multiple named endpoints, and link the endpoint to an agent in config?