Fix model selector showing wrong model in tabs#7784
Conversation
Fix the model selector display being incorrectly synchronized across all tabs when changing a model in one tab. The bug occurred because the global ModelAndProviderContext state was being updated whenever any session's model changed, causing all tabs to display the same model name regardless of their actual session models. Changes: - Make model display per-session: each tab now shows its own session's model - Pass sessionModel/sessionProvider as props through the component tree instead of reading from global context - Add local override state in ModelsBottomBar for immediate UI feedback before session re-fetches - Only update the config default when changeModel() is called without a sessionId (new chats use defaults, existing chats don't affect defaults) Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
![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: 37fe03f133
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".
ui/desktop/src/components/settings/models/subcomponents/SwitchModelModal.tsx
Outdated
Show resolved
Hide resolved
- Use session model/provider for token-limit lookup in ChatInput instead of always reading from global config, so tabs with session-specific models get accurate context-window limits. - Make changeModel() return a boolean success indicator so onModelSelected callback only fires after a successful switch, preventing the UI from showing an override for a model change that failed. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f298d1774a
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".
- Check updateAgentProvider response for errors so changeModel() correctly returns false on API failure, preventing stale UI overrides. - Include configModel/configProvider in token-limit refresh deps so Hub (no-session) ChatInput reloads limits after a config model change. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a2d73aa303
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".
…imit refresh after in-session model changes Moves the modelOverride state from ModelsBottomBar up to ChatInput, creating effectiveModel/effectiveProvider that both CostTracker, loadProviderDetails, and ModelsBottomBar consume. This ensures cost tracking and token limits update immediately after an in-session model change. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f3d469b88e
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".
ui/desktop/src/components/settings/models/bottom_bar/ModelsBottomBar.tsx
Outdated
Show resolved
Hide resolved
When no session exists (Hub, new empty chat), sessionModel/sessionProvider are null. Re-add the useModelAndProvider fallback so the bottom bar shows the configured default model instead of "Select Model". Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5d6a1f3e6d
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".
ui/desktop/src/components/settings/models/subcomponents/SwitchModelModal.tsx
Show resolved
Hide resolved
The predefined model initialization was reading GOOSE_MODEL from config, which could preselect the wrong model when a tab's model differs from the global default. Use the already-computed currentModel (which prefers session model over config) instead. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3ae3522d5a
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".
ui/desktop/src/components/settings/models/subcomponents/SwitchModelModal.tsx
Outdated
Show resolved
Hide resolved
Split predefined model initialization into its own effect with currentModel in the dependency array, so if the modal opens before session/config data has loaded, the selection updates once it arrives. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ui/desktop/src/components/settings/models/bottom_bar/ModelsBottomBar.tsx
Outdated
Show resolved
Hide resolved
- CostTracker: make model/provider props required (non-optional) - ModelsBottomBar: make onModelChanged required (always set by ChatInput) - useCostTracking: early-return when model/provider are missing instead of proceeding with null values Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 26a4954516
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".
In Hub / no-session paths, sessionModel/sessionProvider are always null so the override was never cleared. Now also check configModel/configProvider when there's no active session, preventing stale override from blocking config-driven model updates. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When opening a past chat, session model data loads asynchronously. During that window, both ModelsBottomBar and loadProviderDetails were falling back to the config default model, causing a brief flash of the wrong model. Now only fall back to config when there is genuinely no session (sessionId is null); when a session exists but data hasn't arrived, wait rather than showing/loading the wrong model. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… fallback Use opacity-0 on the model label while waiting for session data to load, so the bot icon stays visible but no incorrect model name flashes. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… chats The sessionId guard was too aggressive — new chats have a sessionId but should still show the config default model. Instead, track whether session model data has resolved: for new chats (!sessionId or sessionModel already set on mount) it's immediately resolved; for resumed chats it resolves once sessionModel arrives, hiding the label only during that brief window. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous sessionModelResolved heuristic couldn't distinguish a new chat (session loaded but no model_config) from a resumed chat still loading. Now BaseChat passes sessionLoaded (session !== undefined) through ChatInput to ModelsBottomBar, so: - New chat: session loads quickly with no model → shows config default - Resumed chat: hides label only until session fetch completes Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1d985cac3c
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".
ui/desktop/src/components/settings/models/subcomponents/SwitchModelModal.tsx
Show resolved
Hide resolved
In manual (non-predefined) mode, provider and model useState values are set at mount time, which may be before session data loads. Add an effect to update them when currentModel/currentProvider resolve, but only if the user hasn't already changed them. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0c12fe10bb
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".
effectiveModel/effectiveProvider were null in Hub (no session, no override), causing CostTracker to receive null and hide pricing. Add configModel/configProvider as the final fallback in the chain. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8cc0ff6123
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".
ui/desktop/src/components/settings/models/subcomponents/SwitchModelModal.tsx
Outdated
Show resolved
Hide resolved
…ovider When the modal opens with initialProvider different from the active provider, model is intentionally set to empty so the user picks one. The sync effect was overriding this by repopulating model from currentModel. Guard against this by skipping sync in the forced-provider flow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b0fb74eb06
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".
ui/desktop/src/components/settings/models/subcomponents/SwitchModelModal.tsx
Outdated
Show resolved
Hide resolved
The sync effect was firing whenever model was empty, including after the user intentionally cleared it by changing provider. Use a ref to ensure it only fires once (on initial session data arrival) and doesn't interfere with subsequent user-driven state changes. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* main: (270 commits) test(acp): align provider and server test parity (#7822) fix(acp): register MCP extensions when resuming a session (#7806) fix(goose): load .gitignore in prompt_manager for hint file filtering (#7795) fix: remap max_completion_tokens to max_tokens for OpenAI-compatible providers (#7765) fix(openai): preserve Responses API tool call/output linkage (#7759) chore(deps): bump @hono/node-server from 1.19.9 to 1.19.11 in /evals/open-model-gym/mcp-harness (#7687) fix: return ContextLengthExceeded when prompt exceeds effective KV cache size (#7815) feat: MCP Roots support (#7790) fix(google): use `includeThoughts/part.thought` for thinking handling (#7593) refactor: simplify tokenizer initialization — remove unnecessary Result wrapper (#7744) Fix model selector showing wrong model in tabs (#7784) Stop collecting goosed stderr after startup (#7814) fix: avoid word splitting by space for windows shell commands (#7781) (#7810) Simplify and make it not break on linux (#7813) Add preferred microphone selection (#7805) Remove dependency on posthog-rs (#7811) feat: load hints in nested subdirs (#7772) feat(acp): add read tool and delegate filesystem I/O to ACP clients (#7668) Support secret interpolation in streamable HTTP extension URLs (#7782) More logging for command injection classifier model training (#7779) ...
2 participants
Summary
Fixed the model selector display being incorrectly synchronized across all tabs when changing a model in one tab. The bug occurred because the global
ModelAndProviderContextstate was being updated whenever any session's model changed, causing all tabs to display the same model name regardless of their actual session models.Type of Change
Key Changes
sessionModel/sessionProvideras props through the component tree instead of reading from global contextModelsBottomBarfor immediate UI feedback before session re-fetcheschangeModel()is called without asessionId(existing chats don't affect defaults for new chats)Testing
Related Issues
Fixes #7402
AI Assistance