Can we have a proxy/fake model?

[EthraZa]

I’d like to have a proxy model that points to whichever model is currently loaded, and if no model is loaded, falls back to loading a specific default one.

To clarify further:
I have the Hermes Agent running, and every time it needs to perform a task, it loads the default model—gpt-oss-20b—to work with.
However, I also want to use a specific model (e.g., Qwen3-Coder) for working on pi.dev. Since my machine can only load one model at a time, if Hermes kicks in while I’m using another model (like Qwen3-Coder), it will load its own model (gpt-oss-20b), displacing mine—which is problematic.
So, I’d like to configure Hermes to use a “fake” proxy model on llama-swap that instead of loading one, simply points to whichever model already loaded (e.g., Qwen3-Coder). If no model is currently loaded, then it should load the default one (e.g., gpt-oss-20b).

Is that possible?