Repository navigation
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![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: 1ab2bda818
ℹ️ About Codex in GitHub
Your team has set up Codex to 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 👍.
Codex can also answer questions or update the PR. Try commenting "@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: f509704622
ℹ️ About Codex in GitHub
Your team has set up Codex to 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 👍.
Codex can also answer questions or update the PR. Try commenting "@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: 427c6749f4
ℹ️ About Codex in GitHub
Your team has set up Codex to 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 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
codegraph-librechat
Bot
added
the
🗺️ Backend Infra
Summary
A custom endpoint pointed at a known provider (OpenRouter, xAI, OpenAI) never showed the composer reasoning effort control unless
librechat.yamldeclaredcustomParamsfor it. The client treats custom endpoints as definition-driven because their model catalogs belong to the deployment, so without areasoningFormat, a reasoningparamDefinitionsentry or a knowndefaultParamsEndpointit resolves no reasoning setting at all.loadCustomEndpointsConfignow declaresreasoningFormat: reasoning_effortfor an endpoint whosebaseURLhost is OpenRouter, xAI or OpenAI. The composer, the send path and the server-side override validation already readcustomParams.reasoningFormat, so all three pick it up with no client change. The match is on the base URL host only, never the endpoint name or icon, so a local server named "xAI" gets nothing.Anything the admin configured wins: a native
provider, a non-defaultdefaultParamsEndpoint, anyreasoningFormat(includingdisabled, which still hides the control), or a reasoningparamDefinitionsentry. Hosts whose request shape I could not confirm (Groq, Deepseek, Mistral) are deliberately left out and covered by a test.Type of change
Testing
Tested environments/configuration:
Custom endpoint configs only, exercised through
loadCustomEndpointsConfig. Not run against a live provider or in a browser.Automated tests:
Added cases to
packages/api/src/endpoints/custom/config.spec.ts: known hosts getreasoning_effort; a name or icon alone does not; Groq and Mistral hosts do not; an explicitreasoningFormat, reasoningparamDefinitions, non-defaultdefaultParamsEndpointor nativeprovideris left untouched. The host case failed first withExpected: "reasoning_effort", Received: undefined. Passing after the change:config.spec.tsandproviders.spec.ts(32 tests),parameterSettings.spec.ts(45 tests).lc lintis clean andtsc --noEmitpasses inpackages/apiandpackages/data-provider.Screenshots / recordings
Not captured. The effort control now appears for these endpoints, but it was not checked in a running app.
Risk / compatibility
Every custom endpoint on those hosts now shows the control by default, with no new config field;
reasoningFormat: disabledis the existing way to turn it off. The flatreasoning_effortshape for xAI is based on secondary sources, not xAI's own documentation, and was not tested against a live API. Endpoints on other hosts, and self-hosted ones, still needcustomParams.Checklist