🤖 feat: persist thinking level preference per model #945
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.
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".
| window.addEventListener("storage", handleStorage); | ||
| return () => window.removeEventListener("storage", handleStorage); |
There was a problem hiding this comment.
The new ThinkingProvider only subscribes to the native storage event, but updatePersistedState (used by the command palette’s onSetThinkingLevel) emits a custom storage event and does not trigger storage in the same tab. As a result, changing thinking level from the palette updates localStorage but never updates the context state, so UI controls and useSendMessageOptions keep using the stale level until a model change or reload. Subscribing to the custom event (as the previous usePersistedState implementation did) is needed to keep same-tab changes in sync.
Useful? React with 👍 / 👎.
ammar-agent
force-pushed
the
thinking-level-per-model
branch
from
b386164 to
51a62e3
Compare
Thinking level is now stored per model instead of per workspace:
- When switching models, the thinking level for that model is loaded
- When changing thinking level, it's saved for the current model
- This allows different models to have different thinking preferences
Storage key changes:
- thinkingLevel:{modelName} - main thinking level per model
- lastActiveThinking:{modelName} - for toggle keybind restoration
Removed:
- lastThinkingByModel:{model} (consolidated into main storage)
- Workspace-based thinking level sync during workspace creation
The toggle keybind (Ctrl+Shift+T) still remembers the last active
(non-off) level for quick on/off toggling.
ammar-agent
force-pushed
the
thinking-level-per-model
branch
from
4f1d6ca to
0220fe1
Compare
1 participant
Thinking level is stored per workspace again, with project seeding for new workspaces.
thinkingLevel:{scopeId})Tests:
make static-checkGenerated with
mux