Move reasoning effort control to composer - #394
Conversation
![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: 6c95407aac
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".
| await handler?.(currentMessage.trim(), files, pastedContent, { | ||
| ...(selectedModel ? { model: selectedModel.modelId || selectedModel.id } : {}), | ||
| ...(selectedModel?.providerId ? { provider: selectedModel.providerId } : {}), | ||
| reasoningEffort: selectedReasoningEffort, |
There was a problem hiding this comment.
Because this field is unconditional and the state initializes to medium, merely sending a message requests reasoning even when the user never selected an effort. For any profile without the reasoning capability—including the built-in OpenAI and DashScope profiles in provider/catalog.rs—require_provider_capability rejects the request, so every ordinary composer submission fails. Provide a provider-default/off state or only attach the effort when the selected profile supports it.
Useful? React with 👍 / 👎.
1 participant
Summary
Why
Reasoning effort is model- and request-specific. Keeping it in Agent Defaults could silently apply a stale global value to unrelated providers or models.
User impact
Users choose effort next to the active model. Medium is the first-use fallback, and later choices are restored locally. Unsupported provider/model combinations fail explicitly rather than being silently downgraded.
Validation
npm test(78 files, 518 tests)npm run buildcargo checkgit diff --check