Model matching and costing with litellm proxy

[zv-i]

Issue

There is this issue already logged for aider: Aider-AI#3323

When having multiple models from different providers sitting on a litellm proxy we have to use the openai api. At the moment with cecli (and aider) I have to maintain a .cecli.model.settings.yml and .cecli.model.metadata.json file in order to map the names of the models because of the need for an openai prefix.

For example with gemini I have to do this in that model settings file:
"openai/gemini/gemini-3.5-flash": {
"max_tokens": 65535,
"max_input_tokens": 1048576,
...
}

Is there some feature in cecli that can resolve this automatically? Is there scope for looking into a code contribution to help with this?
Ideally this should be in upstream, but there seems to be no appetite for it, leading to cecli's existence. While I could use a script to do this, in a team setup it creates an extra point of friction.

Version and model info

cecli 0.100.8
various models: gemini, anthropic

[dwash96]

This makes sense, I can add something in the 0.100.10 patch version for this

[dwash96]

Adding a v0.100.10rc4 pre release version that will have this if you want to use it now. I dogfood my own changes for a bit with these pre-releases