Giving --model-metadata-file <file.json> does not seem to load that file

[paul-gauthier]

I have a models metadata file in ~/.cache/aider-model-metadata.json, which looks something like this:

{

    "foobar-model": {

        "input_cost_per_token": 0,

        "litellm_provider": "openai",

        "max_input_tokens": 90000,

        "max_output_tokens": 4096,

        "max_tokens": 90000,

        "mode": "chat",

        "output_cost_per_token": 0

    }

}

There's more model than one though. I run aider with this command:

aider \

  --model foobar-model \

  --model-metadata-file ~/.cache/aider-model-metadata.json \

  --config ~/.config/aider/aider.conf.yml

v0.73.0:

When running on 0.73.0, everything works fine. Using /models I can see the model:

/models foobar



Models which match "foobar":

- foobar-model

- openai/foobar-model

v0.74.0:

When running on 0.74.0, I see that the model isn't there when doing /models:

/models foobar



No models match "foobar".

I get this error when trying to use the model:

litellm.BadRequestError: LLM Provider NOT provided. Pass in the LLM provider you are trying to call.

Both versions report the same settings, relevant lines from aider -v:

  - model: foobar-model

  - model_metadata_file: /Users/oskar/.cache/aider-model-metadata.json

Let me know what other information I can provide.

Originally posted by @oskarkook in #2928

[paul-gauthier]

Thanks for trying aider and filing this issue.

You need to include a litellm_provider field in the model file. This tells litellm how to communicate with the model's API. So:

{

    "foobar-model": {

        "input_cost_per_token": 0,

        "litellm_provider": "openai",

        "max_input_tokens": 90000,

        "max_output_tokens": 4096,

        "max_tokens": 90000,

        "mode": "chat",

        "output_cost_per_token": 0,

        "litellm_provider": "openai"
    }

}

[oskarkook]

But now you just have litellm_provider twice in there?