Google Gemini provider offers only 2 models, with no way to add others from the UI

[jsavaliya9747]

What happened

After configuring the google-gemini provider, the model picker offers exactly two models:

• gemini-3-1-pro-preview
• gemini-3-6-flash

In Settings, the provider's "Available to add" list is empty, so there is no way to reach any other Gemini model from the UI.

Why

Three things combine:

1. The shipped catalog has only two Gemini entries. packages/trueforge/catalog/model-catalog.yaml:128 lists 2 models for google-gemini, where comparable providers get more — openai 5, anthropic 6, together 7.

2. The UI can only add models already in the catalog. ModelSettings.tsx:312 builds the "Available to add" list as catalogModels.filter(...), so for Gemini it is empty once both catalog models are configured.

3. The configure form has no model field. ConfigureModelProviderForm.tsx collects only API key and base URL for well-known providers. Only the custom type has a model editor (CustomModelProviderForm.tsx).

This is a catalog/UI limit, not an API limit

The API already accepts any model ID: ConfiguredModelSchema (packages/trueforge/src/schemas/modelProvider.ts:32) takes model_id as a free-form string with no allowlist, and GET /models flattens whatever the stored manifest holds. The backend would serve any Gemini model — the catalog just does not list them.

Workarounds

• Point MODEL_CATALOG_PATH at a custom YAML to replace the shipped catalog (packages/trueforge/src/catalog/ModelCatalog.ts:23).
• Or define a custom provider, which does expose a model editor.

Both mean giving up the built-in Gemini provider entry.

Suggested fix

Either would resolve this:

• Expand the google-gemini catalog entries to cover the current lineup, bringing it in line with openai/anthropic.
• Or allow well-known providers to accept a free-text model_id in the UI, as custom already does. This also stops the catalog needing an update every time a provider ships a new model.

Environment

• Commit 1ec19f0f (main), server 0.2.0-rc.2
• Running locally via docker compose up --build

[jayesh9747]

@chiragjn can i pick this up ?

[jayesh9747]

/assign

[jayesh9747]

need option in setting as well to add model or disable model :

[harshitsaxena214]

hey I’d like to take this up. I can work on adding the option to add/disable models from Settings and support additional Gemini model IDs.

[jayesh9747]

@harshitsaxena214 hey bro , i am already working on that

[chiragjn]

I would suggest besides auto discovery which would take longer to review, we can just expand the catalog with newer models as a smaller PR?

[jsavaliya9747]

@chiragjn sure.