Enable multiple models switchable on runtime

[jooray]

I would like to be able to switch models. For example when a local model does not perform well, I might want to switch to a remote model, which costs API keys. My use-case would be ollama and probably qwen3-235b through venice.ai.

That would require the change to configuration file format (you can't have more than one openrouter: key) and a new command /model

[alvinunreal]

Profiles would be proper a fix I guess for stuff like this - will add an issue.

Meanwhile what you could do is to do an alias command, which sets certain env variables, such as alias qwen-tmuxai="TMUXAI_OPENROUTER_API_KEY=xxx TMUXAI_OPENROUTER_MODEL=xx tmuxai"

It's possible to set different model runtime too: /config set openrouter.model

[jooray]

Yes, but that would not switch the model for running session. My idea is - if one model can't do something well, try a different one in session. And the different model might be at some other endpoint.

[alvinunreal]

Right yes, that should be done with profiles than - still to design best approach how to manage/use them.