fix: register LiteLLM with refresh_only inventory so model picker lists models - #10489
Conversation
…ts models LiteLLM was registered without an inventory registration, making supports_refresh=false and skipping every refresh with DoesNotSupportRefresh. Combined with empty static known_models, the model picker stayed permanently empty. Register it with refresh_only() like Databricks so fetch_supported_models is called and proxy models appear in the picker. Fixes aaif-goose#10473
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 70f44ee979
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4e4c806bbf
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0b453a06a0
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 85e4d332e4
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| config | ||
| .get_param::<serde_json::Value>("LITELLM_HOST") | ||
| .is_ok() | ||
| || config | ||
| .get_secret::<serde_json::Value>("LITELLM_API_KEY") | ||
| .is_ok() |
There was a problem hiding this comment.
When LITELLM_HOST or LITELLM_API_KEY is present but empty (for example from an .env placeholder like LITELLM_API_KEY=), these is_ok() checks still mark LiteLLM as configured. LiteLLMProvider::from_env treats an empty API key as NoAuth, and an empty host is used as the base URL, so the model picker will start inventory refreshes against an unauthenticated/default or invalid endpoint instead of reporting the provider as unconfigured. Use the existing normalized helpers (or deserialize to String and require non-empty) before returning true.
Useful? React with 👍 / 👎.
michaelneale
left a comment
michaelneale
left a comment
There was a problem hiding this comment.
nice one @Abhijay007
2 participants
fixes: #10473
Summary
LiteLLM's model picker was permanently empty due to 3 bugs: it was registered without inventory (so /model/info was never called), lacked a proper configured check (keyless proxies were skipped), and didn't implement skip_canonical_filtering (so proxy models got filtered out).
Testing
manual and unit
Screenshots/Demos (for UX changes)